The End of Data Entry: Why Your Business Logic is Moving to Agents

1
00:00:00,000 --> 00:00:02,800
Enterprise Software has one job for the last 20 years.

2
00:00:02,800 --> 00:00:04,680
We believe that job was storage.

3
00:00:04,680 --> 00:00:07,400
Store the customer record, store the sales opportunity,

4
00:00:07,400 --> 00:00:09,600
store the invoice, store the PO.

5
00:00:09,600 --> 00:00:12,200
We spend billions trying to build a single source of truth

6
00:00:12,200 --> 00:00:13,880
where data was organized and accessible.

7
00:00:13,880 --> 00:00:15,160
We called it a system of record,

8
00:00:15,160 --> 00:00:17,880
but something changed, and nobody announced it.

9
00:00:17,880 --> 00:00:20,480
Your Enterprise Software isn't just storing data anymore.

10
00:00:20,480 --> 00:00:23,280
It's executing processes, it's making decisions,

11
00:00:23,280 --> 00:00:25,120
and it's orchestrating work across systems

12
00:00:25,120 --> 00:00:26,880
that were never designed to talk to each other.

13
00:00:26,880 --> 00:00:28,120
The problem is simple.

14
00:00:28,120 --> 00:00:30,000
We haven't actually shifted how we staff,

15
00:00:30,000 --> 00:00:32,680
how we build, or how we think about what those systems do.

16
00:00:32,680 --> 00:00:35,600
We're still hiring people to be integration layers.

17
00:00:35,600 --> 00:00:37,360
A sales rep reads an email from a prospect

18
00:00:37,360 --> 00:00:39,480
and then spends 20 minutes entering that information

19
00:00:39,480 --> 00:00:43,400
into the CRM because the system can't do it automatically.

20
00:00:43,400 --> 00:00:45,000
A finance controller receives an invoice

21
00:00:45,000 --> 00:00:47,200
and then manually matches it to a purchase order

22
00:00:47,200 --> 00:00:48,320
and a general ledger entry

23
00:00:48,320 --> 00:00:50,840
because three systems can't communicate.

24
00:00:50,840 --> 00:00:53,080
A procurement manager gets an email from a supplier

25
00:00:53,080 --> 00:00:54,560
saying a delivery is delayed,

26
00:00:54,560 --> 00:00:57,040
so they check the warehouse system, update the purchase order

27
00:00:57,040 --> 00:00:59,560
manually and send a notification downstream.

28
00:00:59,560 --> 00:01:01,360
These aren't jobs, they're workarounds.

29
00:01:01,360 --> 00:01:03,560
And the cost isn't just in the time they consume,

30
00:01:03,560 --> 00:01:04,960
it's in the speed they destroy.

31
00:01:04,960 --> 00:01:07,600
A lead takes three days to move from an email inbox

32
00:01:07,600 --> 00:01:09,080
into your sales pipeline

33
00:01:09,080 --> 00:01:11,040
because a human has to be the translator.

34
00:01:11,040 --> 00:01:12,720
A month and close takes five days

35
00:01:12,720 --> 00:01:14,880
because finance teams are matching pennies

36
00:01:14,880 --> 00:01:16,360
instead of analyzing patterns.

37
00:01:16,360 --> 00:01:19,120
A supplier communication takes days instead of seconds

38
00:01:19,120 --> 00:01:21,480
because someone has to read an email, check a system,

39
00:01:21,480 --> 00:01:22,640
and push the update.

40
00:01:22,640 --> 00:01:24,680
This is what happens when you build a system of record

41
00:01:24,680 --> 00:01:25,960
instead of a system of action.

42
00:01:25,960 --> 00:01:27,080
That's changing now.

43
00:01:27,080 --> 00:01:29,960
Not with new features or dashboards, but with agents.

44
00:01:29,960 --> 00:01:33,080
Software that observes, reasons and acts autonomously

45
00:01:33,080 --> 00:01:34,840
within guardrails you define.

46
00:01:34,840 --> 00:01:37,720
These agents are grounded in your actual business data

47
00:01:37,720 --> 00:01:39,680
rather than generic AI knowledge.

48
00:01:39,680 --> 00:01:42,280
They respect your security roles, your approval workflows

49
00:01:42,280 --> 00:01:43,520
and your compliance requirements

50
00:01:43,520 --> 00:01:46,600
while executing the work humans were forced to do manually.

51
00:01:46,600 --> 00:01:48,720
Organizations that move toward agente workflows

52
00:01:48,720 --> 00:01:50,880
now will compress their operational cycles.

53
00:01:50,880 --> 00:01:52,560
Everyone else will fall behind.

54
00:01:52,560 --> 00:01:54,320
The system of record is broken.

55
00:01:54,320 --> 00:01:56,200
20 years ago the promise was clean.

56
00:01:56,200 --> 00:01:58,520
Software would be the single source of truth.

57
00:01:58,520 --> 00:02:00,560
The CRM would store every customer interaction.

58
00:02:00,560 --> 00:02:02,480
The ERP would store every transaction

59
00:02:02,480 --> 00:02:04,880
and we'd finally say goodbye to spreadsheets and guessing.

60
00:02:04,880 --> 00:02:06,440
Data would flow in, get organized

61
00:02:06,440 --> 00:02:08,560
and everyone would have access to the same facts.

62
00:02:08,560 --> 00:02:11,240
It was a good promise, but here's what actually happened.

63
00:02:11,240 --> 00:02:13,920
Data lives in the system, work happens everywhere else.

64
00:02:13,920 --> 00:02:16,160
The invoice gets scanned and stored in the system

65
00:02:16,160 --> 00:02:18,160
but the finance manager works in email

66
00:02:18,160 --> 00:02:20,960
forwarding it to accounting to ask about a discrepancy.

67
00:02:20,960 --> 00:02:23,080
The customer record gets created in the CRM

68
00:02:23,080 --> 00:02:25,000
but the sales rep is working in email

69
00:02:25,000 --> 00:02:27,720
asking for approval from their manager before they reach out.

70
00:02:27,720 --> 00:02:29,440
The PO gets entered into the ERP

71
00:02:29,440 --> 00:02:31,520
but the procurement manager is working in email

72
00:02:31,520 --> 00:02:34,160
negotiating the delivery date directly with the supplier

73
00:02:34,160 --> 00:02:36,200
instead of updating the system first.

74
00:02:36,200 --> 00:02:37,760
The system became a filing cabinet,

75
00:02:37,760 --> 00:02:39,520
a really expensive filing cabinet

76
00:02:39,520 --> 00:02:41,160
and someone has to be the librarian.

77
00:02:41,160 --> 00:02:42,760
That librarian costs money.

78
00:02:42,760 --> 00:02:45,240
A typical enterprise sales rep spends two hours a day

79
00:02:45,240 --> 00:02:48,160
entering data that the system could capture automatically

80
00:02:48,160 --> 00:02:51,400
like a lead from LinkedIn, a conversation note from a call

81
00:02:51,400 --> 00:02:53,320
or an email attachment with a proposal.

82
00:02:53,320 --> 00:02:57,320
Two hours every day at roughly $150,000 a year fully loaded

83
00:02:57,320 --> 00:03:00,640
that $75,000 a year per rep just to be a manual data relay

84
00:03:00,640 --> 00:03:01,800
between systems.

85
00:03:01,800 --> 00:03:03,800
Multiply that across a sales team of 50.

86
00:03:03,800 --> 00:03:07,160
That's $3.75 million a year in pure integration tax

87
00:03:07,160 --> 00:03:09,080
but the real cost isn't the salary.

88
00:03:09,080 --> 00:03:10,080
It's the speed.

89
00:03:10,080 --> 00:03:14,480
A lead arrives as an email and a human has to read it

90
00:03:14,480 --> 00:03:17,000
and rich it, score it and enter it into the CRM.

91
00:03:17,000 --> 00:03:19,240
That's a day, maybe three if the person is busy.

92
00:03:19,240 --> 00:03:20,640
By the time it hits the rep's desk

93
00:03:20,640 --> 00:03:22,800
the prospect is already called three competitors.

94
00:03:22,800 --> 00:03:25,840
A supplier sends an email saying they can't deliver on time

95
00:03:25,840 --> 00:03:28,080
so a human has to read it, check the warehouse,

96
00:03:28,080 --> 00:03:30,360
understand the impact and update the purchase order.

97
00:03:30,360 --> 00:03:32,360
While they're doing that, production is still scheduled

98
00:03:32,360 --> 00:03:34,320
based on outdated information.

99
00:03:34,320 --> 00:03:36,360
A customer calls with a billing question

100
00:03:36,360 --> 00:03:38,560
and the support agent roots it based on the subject line

101
00:03:38,560 --> 00:03:41,360
instead of context so it bounces between three departments

102
00:03:41,360 --> 00:03:44,240
before anyone figures out the real issue is a contract renewal

103
00:03:44,240 --> 00:03:45,720
that's about to expire.

104
00:03:45,720 --> 00:03:48,000
The system stores the data but humans move it

105
00:03:48,000 --> 00:03:49,080
and humans are slow.

106
00:03:50,000 --> 00:03:52,560
This model is breaking now because AI can do what humans

107
00:03:52,560 --> 00:03:54,240
have been doing, observe, reason and act

108
00:03:54,240 --> 00:03:56,080
but without the delay, without the error

109
00:03:56,080 --> 00:03:57,280
and without the cost.

110
00:03:57,280 --> 00:04:00,560
Not chatbots, not generic LLMs trained on the internet

111
00:04:00,560 --> 00:04:03,160
but agents grounded in your actual business data

112
00:04:03,160 --> 00:04:04,960
constrained by your security roles

113
00:04:04,960 --> 00:04:07,160
and orchestrated through your existing systems.

114
00:04:07,160 --> 00:04:09,400
The shift from system of record to system of action

115
00:04:09,400 --> 00:04:10,240
isn't coming.

116
00:04:10,240 --> 00:04:11,200
It's already here.

117
00:04:11,200 --> 00:04:12,760
The agent shift explained.

118
00:04:12,760 --> 00:04:14,120
There's a fundamentally different way

119
00:04:14,120 --> 00:04:16,440
to build what enterprise software actually does.

120
00:04:16,440 --> 00:04:18,440
Most applications are just storage layers.

121
00:04:18,440 --> 00:04:19,960
You move data in, you move data

122
00:04:19,960 --> 00:04:21,840
out but imagine software that actually

123
00:04:21,840 --> 00:04:23,040
participates in the work.

124
00:04:23,040 --> 00:04:25,080
It observes what is happening in real time.

125
00:04:25,080 --> 00:04:27,680
It reads an incoming email and understands the meaning.

126
00:04:27,680 --> 00:04:29,320
It pulls context from your systems

127
00:04:29,320 --> 00:04:31,640
and makes a decision about what happens next.

128
00:04:31,640 --> 00:04:32,800
Then it takes action.

129
00:04:32,800 --> 00:04:34,800
All of this happens without asking for permission

130
00:04:34,800 --> 00:04:35,880
at every single step.

131
00:04:35,880 --> 00:04:37,120
That is what a gentick means.

132
00:04:37,120 --> 00:04:38,640
It is not sentient or creative.

133
00:04:38,640 --> 00:04:40,840
It is not replacing your strategic thinking.

134
00:04:40,840 --> 00:04:43,840
An agent is software designed for a specific business workflow.

135
00:04:43,840 --> 00:04:46,760
It perceives an event, reasons about it using your data

136
00:04:46,760 --> 00:04:48,760
and orchestrates actions across your systems.

137
00:04:48,760 --> 00:04:50,480
It is autonomous within guardrails.

138
00:04:50,480 --> 00:04:52,240
It is deterministic within boundaries.

139
00:04:52,240 --> 00:04:55,520
Every agentic workflow follows the same three-part pattern.

140
00:04:55,520 --> 00:04:57,200
First, an event happens.

141
00:04:57,200 --> 00:04:59,800
A prospect sends an email to your company mailbox.

142
00:04:59,800 --> 00:05:01,360
A supplier sends a purchase order.

143
00:05:01,360 --> 00:05:02,800
A customer opens a support ticket.

144
00:05:02,800 --> 00:05:04,760
These are real events happening right now.

145
00:05:04,760 --> 00:05:05,800
The event is the trigger.

146
00:05:05,800 --> 00:05:07,600
Second, the agent reasons about it.

147
00:05:07,600 --> 00:05:08,720
It pulls context.

148
00:05:08,720 --> 00:05:10,600
For a sales prospect, the agent queries

149
00:05:10,600 --> 00:05:12,680
your data verse to see if this company has talked to you

150
00:05:12,680 --> 00:05:13,400
before.

151
00:05:13,400 --> 00:05:15,800
It checks linked in to see their size and funding.

152
00:05:15,800 --> 00:05:17,640
It compares them against your history

153
00:05:17,640 --> 00:05:20,400
to see if they are a genuine opportunity or just a tire

154
00:05:20,400 --> 00:05:20,960
kicker.

155
00:05:20,960 --> 00:05:22,720
For a purchase order, it checks your inventory

156
00:05:22,720 --> 00:05:23,960
and your contract terms.

157
00:05:23,960 --> 00:05:26,280
For a support ticket, it pulls the full history

158
00:05:26,280 --> 00:05:27,920
and payment record.

159
00:05:27,920 --> 00:05:29,920
The reasoning layer answers one question.

160
00:05:29,920 --> 00:05:31,960
What is actually happening here?

161
00:05:31,960 --> 00:05:33,920
Third, the agent orchestrates actions.

162
00:05:33,920 --> 00:05:35,640
It does not just auto-qualify a prospect.

163
00:05:35,640 --> 00:05:38,480
Instead, it prepares intelligence so your sales rep

164
00:05:38,480 --> 00:05:40,760
can make a smart decision in seconds.

165
00:05:40,760 --> 00:05:43,160
For an incoming PO, it flags discrepancies

166
00:05:43,160 --> 00:05:44,520
before they hit your system.

167
00:05:44,520 --> 00:05:46,480
For a support ticket, it roots the customer

168
00:05:46,480 --> 00:05:48,760
to the right team with the context already loaded.

169
00:05:48,760 --> 00:05:51,080
For month and close, it identifies exceptions

170
00:05:51,080 --> 00:05:53,320
and only escalates the ones that need a human.

171
00:05:53,320 --> 00:05:54,800
The reason this works is data verse.

172
00:05:54,800 --> 00:05:57,320
Your agents do not operate on generic AI knowledge.

173
00:05:57,320 --> 00:06:00,080
They do not guess, based on what the internet taught them.

174
00:06:00,080 --> 00:06:02,120
They reason using your actual business data.

175
00:06:02,120 --> 00:06:04,800
This includes your customer records, your transaction history,

176
00:06:04,800 --> 00:06:06,440
and your financial policies.

177
00:06:06,440 --> 00:06:09,560
Data verse is the platform underneath Dynamics 365.

178
00:06:09,560 --> 00:06:11,720
It stores your accounts and your invoices.

179
00:06:11,720 --> 00:06:13,360
When an agent queries data verse,

180
00:06:13,360 --> 00:06:17,400
it uses the same authoritative source your users access manually.

181
00:06:17,400 --> 00:06:19,360
The agent knows what your business knows.

182
00:06:19,360 --> 00:06:21,840
And the agent operates within the same security boundaries

183
00:06:21,840 --> 00:06:22,840
your employees do.

184
00:06:22,840 --> 00:06:24,320
This is a critical guardrail.

185
00:06:24,320 --> 00:06:25,960
Agents are not god mode automation.

186
00:06:25,960 --> 00:06:28,080
They do not bypass your approval workflows.

187
00:06:28,080 --> 00:06:29,200
They cannot see data.

188
00:06:29,200 --> 00:06:30,760
They are not authorized to see.

189
00:06:30,760 --> 00:06:33,120
An agent assigned to a sales role can only access

190
00:06:33,120 --> 00:06:34,800
what a sales user can access.

191
00:06:34,800 --> 00:06:37,440
High value decisions still require a human sign off.

192
00:06:37,440 --> 00:06:39,600
Every action is logged and audited.

193
00:06:39,600 --> 00:06:41,240
The agent respects your governance.

194
00:06:41,240 --> 00:06:43,320
The real world impact tells the story.

195
00:06:43,320 --> 00:06:45,320
US Ventures deployed reconciliation agents

196
00:06:45,320 --> 00:06:47,800
and cut their month and close time by 80%.

197
00:06:47,800 --> 00:06:49,720
They moved from five days down to one.

198
00:06:49,720 --> 00:06:51,960
Lifetime products implemented procurement agents

199
00:06:51,960 --> 00:06:54,280
and reduced their workload by 20%.

200
00:06:54,280 --> 00:06:56,520
They handle the same volume with fewer people

201
00:06:56,520 --> 00:06:58,240
because agents handle the routine matching

202
00:06:58,240 --> 00:07:00,000
that used to eat half a buyer's day.

203
00:07:00,000 --> 00:07:01,400
This is not theoretical.

204
00:07:01,400 --> 00:07:02,920
Organizations are seeing these results

205
00:07:02,920 --> 00:07:04,840
because they moved from humans moving data

206
00:07:04,840 --> 00:07:07,120
to software executing the process.

207
00:07:07,120 --> 00:07:09,080
Why generic AI fails in enterprise?

208
00:07:09,080 --> 00:07:10,520
You have probably tested a chatbot.

209
00:07:10,520 --> 00:07:12,360
You asked it about a product or a policy.

210
00:07:12,360 --> 00:07:13,480
It sounded intelligent.

211
00:07:13,480 --> 00:07:15,040
It gave you a structured answer.

212
00:07:15,040 --> 00:07:16,320
It even cited sources.

213
00:07:16,320 --> 00:07:17,800
Then you asked it for real knowledge.

214
00:07:17,800 --> 00:07:18,720
And it made things up.

215
00:07:18,720 --> 00:07:19,920
That is the chatbot trap.

216
00:07:19,920 --> 00:07:22,120
It reveals the gap between generic AI

217
00:07:22,120 --> 00:07:24,240
and agents grounded in your business.

218
00:07:24,240 --> 00:07:26,920
A model trained on the internet can talk about your business.

219
00:07:26,920 --> 00:07:28,880
It can discuss sales or supply chains,

220
00:07:28,880 --> 00:07:30,360
but it does not understand your business.

221
00:07:30,360 --> 00:07:32,040
It does not know your standard payment term

222
00:07:32,040 --> 00:07:33,960
is net 30 unless you type it in.

223
00:07:33,960 --> 00:07:35,280
It does not know your top customer

224
00:07:35,280 --> 00:07:37,560
has a 15% discount on big orders.

225
00:07:37,560 --> 00:07:39,520
It does not know your approval thresholds

226
00:07:39,520 --> 00:07:41,040
for managers versus directors.

227
00:07:41,040 --> 00:07:43,080
It does not know the nuance of your relationships

228
00:07:43,080 --> 00:07:44,680
or the specifics of your contracts.

229
00:07:44,680 --> 00:07:46,840
More critically, it does not know what actually matters

230
00:07:46,840 --> 00:07:48,160
in your context.

231
00:07:48,160 --> 00:07:50,280
A customer calls about a billing question.

232
00:07:50,280 --> 00:07:52,640
A generic chatbot just roots it to billing.

233
00:07:52,640 --> 00:07:54,880
But a Dynamics 365 agent reads the history.

234
00:07:54,880 --> 00:07:57,080
It sees they have called six times this month.

235
00:07:57,080 --> 00:07:59,000
It realizes the real problem is a contract

236
00:07:59,000 --> 00:08:01,720
about to expire and roots them to account management instead.

237
00:08:01,720 --> 00:08:03,160
The difference is not the AI model.

238
00:08:03,160 --> 00:08:05,080
The difference is whether the system has access

239
00:08:05,080 --> 00:08:06,480
to your actual business context.

240
00:08:06,480 --> 00:08:07,800
This is the context problem.

241
00:08:07,800 --> 00:08:09,160
At scale, it is fatal.

242
00:08:09,160 --> 00:08:11,160
Generic AI has no access to your data.

243
00:08:11,160 --> 00:08:12,800
It cannot check your supplier contracts

244
00:08:12,800 --> 00:08:14,360
or see your approval workflows.

245
00:08:14,360 --> 00:08:15,960
It does not understand your hierarchy.

246
00:08:15,960 --> 00:08:17,640
Every piece of context must be typed

247
00:08:17,640 --> 00:08:19,480
into the conversation in real time.

248
00:08:19,480 --> 00:08:21,880
That means you're either feeding it information manually

249
00:08:21,880 --> 00:08:23,280
or it is operating blind.

250
00:08:23,280 --> 00:08:25,160
There is another gap nobody talks about.

251
00:08:25,160 --> 00:08:26,640
The execution gap.

252
00:08:26,640 --> 00:08:29,560
Chatbots suggest agents execute.

253
00:08:29,560 --> 00:08:32,200
A chatbot can recommend that you create a purchase order,

254
00:08:32,200 --> 00:08:34,560
but it cannot actually create that PO in your ERP.

255
00:08:34,560 --> 00:08:35,800
It cannot root it for approval

256
00:08:35,800 --> 00:08:37,520
or notify your procurement manager.

257
00:08:37,520 --> 00:08:40,240
It suggests next steps and then you have to do the work.

258
00:08:40,240 --> 00:08:43,040
A Dynamics 365 agent does not suggest it executes.

259
00:08:43,040 --> 00:08:44,200
It creates the record.

260
00:08:44,200 --> 00:08:45,280
It roots the workflow.

261
00:08:45,280 --> 00:08:46,200
It logs the action.

262
00:08:46,200 --> 00:08:47,760
The agent is connected to your systems

263
00:08:47,760 --> 00:08:49,520
while the chatbot is disconnected.

264
00:08:49,520 --> 00:08:51,360
Then there is compliance.

265
00:08:51,360 --> 00:08:53,960
Generic AI does not understand your security model.

266
00:08:53,960 --> 00:08:55,400
It does not know that a rep in Denver

267
00:08:55,400 --> 00:08:57,640
should only see opportunities in her territory.

268
00:08:57,640 --> 00:08:59,240
It does not know that finance members

269
00:08:59,240 --> 00:09:01,920
cannot approve payments over $10,000.

270
00:09:01,920 --> 00:09:03,760
It does not understand your data loss policies.

271
00:09:03,760 --> 00:09:06,600
It responds as if all information is equal for everyone.

272
00:09:06,600 --> 00:09:09,080
In regulated industries like healthcare or finance,

273
00:09:09,080 --> 00:09:10,960
that is a massive liability.

274
00:09:10,960 --> 00:09:14,560
Dynamics 365 agents operate inside your security framework.

275
00:09:14,560 --> 00:09:16,720
They are bound by the same roles as your users.

276
00:09:16,720 --> 00:09:18,240
If an agent is in a sales role,

277
00:09:18,240 --> 00:09:20,400
it only acts on what a sales user can see.

278
00:09:20,400 --> 00:09:21,760
Every decision is auditable.

279
00:09:21,760 --> 00:09:23,280
Governance is not an add-on.

280
00:09:23,280 --> 00:09:24,760
It is built into how the agent works.

281
00:09:24,760 --> 00:09:26,720
The cost of getting this wrong is concrete.

282
00:09:26,720 --> 00:09:28,080
A lead gets sent to the wrong rep

283
00:09:28,080 --> 00:09:30,920
because the system did not understand territory rules.

284
00:09:30,920 --> 00:09:32,960
That rep calls a customer they should not touch

285
00:09:32,960 --> 00:09:34,320
and you lose the account.

286
00:09:34,320 --> 00:09:36,320
A payment gets approved that violates your policy

287
00:09:36,320 --> 00:09:38,600
because the chatbot did not check limits.

288
00:09:38,600 --> 00:09:40,840
An employee gets data they are not supposed to see

289
00:09:40,840 --> 00:09:43,000
because the AI did not enforce security.

290
00:09:43,000 --> 00:09:44,160
These are the recurring patterns

291
00:09:44,160 --> 00:09:47,440
when you treat enterprise software like an internet chatbot.

292
00:09:47,440 --> 00:09:50,960
Dynamics 365 agents are different because they are not generic.

293
00:09:50,960 --> 00:09:52,000
They are grounded.

294
00:09:52,000 --> 00:09:54,520
Sales qualification agent, the intent engine.

295
00:09:54,520 --> 00:09:56,200
Let's move from theory into practice.

296
00:09:56,200 --> 00:09:59,080
We'll start with sales because the problem is the most visible

297
00:09:59,080 --> 00:10:00,920
and the agent's impact is immediate.

298
00:10:00,920 --> 00:10:02,920
Here's how sales actually works right now.

299
00:10:02,920 --> 00:10:05,320
A prospect emails your generic company address

300
00:10:05,320 --> 00:10:07,040
or they fill out a form on your website

301
00:10:07,040 --> 00:10:09,160
or they respond to a LinkedIn outreach campaign.

302
00:10:09,160 --> 00:10:11,560
Your sales development team gets a notification.

303
00:10:11,560 --> 00:10:12,360
Then what happens?

304
00:10:12,360 --> 00:10:13,240
They read the email.

305
00:10:13,240 --> 00:10:14,480
They Google the company.

306
00:10:14,480 --> 00:10:17,120
They check if you've worked with this organization before.

307
00:10:17,120 --> 00:10:19,080
They try to figure out if this is a real opportunity

308
00:10:19,080 --> 00:10:20,440
or someone who's just curious.

309
00:10:20,440 --> 00:10:21,720
But here's the problem.

310
00:10:21,720 --> 00:10:22,880
They can't know for certain

311
00:10:22,880 --> 00:10:25,400
so they assume it might be real, which means they chase it.

312
00:10:25,400 --> 00:10:27,280
A sales rep spends four hours on a prospect

313
00:10:27,280 --> 00:10:28,840
who turns out to be a tire kicker.

314
00:10:28,840 --> 00:10:30,320
They do discovery calls.

315
00:10:30,320 --> 00:10:31,840
They put together a proposal

316
00:10:31,840 --> 00:10:33,880
and they check in twice before they realize

317
00:10:33,880 --> 00:10:36,880
this prospect has no budget, no authority and no timeline.

318
00:10:36,880 --> 00:10:39,560
Or worse, they've already committed to a competitor.

319
00:10:39,560 --> 00:10:41,400
The rep just spent four hours on a deal

320
00:10:41,400 --> 00:10:43,400
with a 15% close likelihood

321
00:10:43,400 --> 00:10:44,960
while the three deals in their pipeline

322
00:10:44,960 --> 00:10:48,040
with 70% likelihood are sitting cold

323
00:10:48,040 --> 00:10:51,200
because the rep is busy chasing low probability noise.

324
00:10:51,200 --> 00:10:52,040
That's the flaw.

325
00:10:52,040 --> 00:10:56,280
By the time a rep realizes a lead is low quality,

326
00:10:56,280 --> 00:10:58,360
they've already invested significant time.

327
00:10:58,360 --> 00:11:01,120
They made this judgment call based on incomplete information

328
00:11:01,120 --> 00:11:03,360
because they can only do so much research manually

329
00:11:03,360 --> 00:11:06,320
in the window before they need to decide whether to engage.

330
00:11:06,320 --> 00:11:09,040
A sales qualification agent changes this entirely.

331
00:11:09,040 --> 00:11:09,920
Here's what it does.

332
00:11:09,920 --> 00:11:12,840
It researches the prospect before your rep ever sees them.

333
00:11:12,840 --> 00:11:15,640
It pulls the prospects linked in profile, company size,

334
00:11:15,640 --> 00:11:17,440
funding status, industry classification,

335
00:11:17,440 --> 00:11:18,600
and employee count.

336
00:11:18,600 --> 00:11:20,080
It checks your historical data

337
00:11:20,080 --> 00:11:22,200
to see if you've worked with this company before

338
00:11:22,200 --> 00:11:23,360
and what the outcome was.

339
00:11:23,360 --> 00:11:24,840
It looks at your wind patterns.

340
00:11:24,840 --> 00:11:27,680
You sell enterprise software to mid market tech companies

341
00:11:27,680 --> 00:11:29,480
in the education space.

342
00:11:29,480 --> 00:11:32,320
This prospect is a 200% SaaS company in Edtech.

343
00:11:32,320 --> 00:11:33,160
That's a match.

344
00:11:33,160 --> 00:11:35,360
Or there are 5,000 person manufacturing company

345
00:11:35,360 --> 00:11:36,400
in industrial automation.

346
00:11:36,400 --> 00:11:37,440
That's not your pattern.

347
00:11:37,440 --> 00:11:39,400
It checks previous interactions to see if someone

348
00:11:39,400 --> 00:11:41,680
from your company emailed this person if they responded

349
00:11:41,680 --> 00:11:44,160
or if they ever actually scheduled a call.

350
00:11:44,160 --> 00:11:46,560
The agent scores the lead against all of this context

351
00:11:46,560 --> 00:11:47,400
in seconds.

352
00:11:47,400 --> 00:11:49,360
The lead arrives via email and the agent compares

353
00:11:49,360 --> 00:11:51,360
the prospect against your ideal customer profile,

354
00:11:51,360 --> 00:11:54,360
your historical wind data, and your current account coverage.

355
00:11:54,360 --> 00:11:57,200
Then the agent enriches the lead record in Dynamics 365

356
00:11:57,200 --> 00:11:59,760
with this intelligence and flags it as hot,

357
00:11:59,760 --> 00:12:01,640
research further or pass.

358
00:12:01,640 --> 00:12:02,920
But here's the thing.

359
00:12:02,920 --> 00:12:04,520
The agent doesn't auto-qualify.

360
00:12:04,520 --> 00:12:06,760
It doesn't automatically assign the lead to a rep

361
00:12:06,760 --> 00:12:08,040
and declare it closed.

362
00:12:08,040 --> 00:12:10,200
The qualification decision stays with your team.

363
00:12:10,200 --> 00:12:12,160
What the agent does is prepare intelligence

364
00:12:12,160 --> 00:12:14,600
so your rep makes a smarter decision faster.

365
00:12:14,600 --> 00:12:16,920
Instead of a rep spending four hours on discovery

366
00:12:16,920 --> 00:12:18,440
to realize its low quality,

367
00:12:18,440 --> 00:12:21,160
they spend 10 minutes reviewing what the agent discovered

368
00:12:21,160 --> 00:12:23,880
and deciding to either move forward or pass.

369
00:12:23,880 --> 00:12:25,680
If the agent flagged it as hot,

370
00:12:25,680 --> 00:12:28,840
the rep has context about who this prospect is, what they do,

371
00:12:28,840 --> 00:12:31,440
and whether your company has worked with them before.

372
00:12:31,440 --> 00:12:33,800
If the agent flagged it as research further,

373
00:12:33,800 --> 00:12:35,920
the rep knows what additional information is missing

374
00:12:35,920 --> 00:12:38,080
and where to focus their early conversations.

375
00:12:38,080 --> 00:12:39,400
The outcome is concrete.

376
00:12:39,400 --> 00:12:40,800
Your rep spend time on prospects

377
00:12:40,800 --> 00:12:42,800
with 70% plus conversion likelihood

378
00:12:42,800 --> 00:12:44,480
instead of 15% likelihood.

379
00:12:44,480 --> 00:12:45,880
They get to the first meeting faster

380
00:12:45,880 --> 00:12:48,280
because they're not wasting time on tire kickers.

381
00:12:48,280 --> 00:12:51,200
They walk into that first meeting with context already prepared,

382
00:12:51,200 --> 00:12:54,120
including the prospect's industry, their company size,

383
00:12:54,120 --> 00:12:55,600
and why they reached out.

384
00:12:55,600 --> 00:12:57,520
That changes the conversation immediately.

385
00:12:57,520 --> 00:13:00,080
You're not starting with, so tell me about your business.

386
00:13:00,080 --> 00:13:02,120
You're starting with, I saw that your company

387
00:13:02,120 --> 00:13:04,840
recently raised Series B in the edtech space.

388
00:13:04,840 --> 00:13:07,560
How is that shaping your approach to this problem?

389
00:13:07,560 --> 00:13:09,560
Sales qualification isn't a new concept,

390
00:13:09,560 --> 00:13:12,080
but qualifying at scale using manual research

391
00:13:12,080 --> 00:13:13,360
was always a bottleneck.

392
00:13:13,360 --> 00:13:15,440
The sales qualification agent removes that bottleneck

393
00:13:15,440 --> 00:13:18,040
by doing the research automatically in context

394
00:13:18,040 --> 00:13:19,640
against your actual business data.

395
00:13:19,640 --> 00:13:23,040
That's the power of an agent grounded in dataverse.

396
00:13:23,040 --> 00:13:24,800
Account reconciliation agent,

397
00:13:24,800 --> 00:13:26,160
the autonomous auditor.

398
00:13:26,160 --> 00:13:28,560
Finance teams face a different kind of manual work

399
00:13:28,560 --> 00:13:30,880
and it's where agents deliver the fastest ROI.

400
00:13:30,880 --> 00:13:34,120
Month and close is happening right now in thousands of organizations.

401
00:13:34,120 --> 00:13:35,440
And here's what it looks like.

402
00:13:35,440 --> 00:13:37,640
Accountants and controllers are sitting at their desks,

403
00:13:37,640 --> 00:13:38,920
looking at three windows.

404
00:13:38,920 --> 00:13:41,040
A sub-ledger, a general ledger, and a spreadsheet.

405
00:13:41,040 --> 00:13:43,760
They're matching invoice amounts to purchase order amounts

406
00:13:43,760 --> 00:13:45,560
to general ledger entries.

407
00:13:45,560 --> 00:13:46,920
They're looking for discrepancies.

408
00:13:46,920 --> 00:13:51,320
A $10,000 invoice came in, but the PO says $9,850.

409
00:13:51,320 --> 00:13:51,840
Why?

410
00:13:51,840 --> 00:13:53,640
Wrong GL code, quantity variance,

411
00:13:53,640 --> 00:13:55,520
early payment discount that wasn't applied,

412
00:13:55,520 --> 00:13:57,880
timing difference, they have to investigate.

413
00:13:57,880 --> 00:14:00,120
Then they create a journal entry to adjust the mismatch,

414
00:14:00,120 --> 00:14:01,240
then they move to the next one.

415
00:14:01,240 --> 00:14:02,600
This is the reconciliation tax.

416
00:14:02,600 --> 00:14:04,760
Finance teams spend four days a month on this work

417
00:14:04,760 --> 00:14:07,280
for a team of five people that's 20 days of human effort

418
00:14:07,280 --> 00:14:10,080
per month every month doing work that a machine should be doing.

419
00:14:10,080 --> 00:14:11,320
The human cost is staggering.

420
00:14:11,320 --> 00:14:15,560
You're paying a controller $120,000 a year to find and fix

421
00:14:15,560 --> 00:14:17,480
0.47 discrepancies.

422
00:14:17,480 --> 00:14:18,960
That controller has a finance degree,

423
00:14:18,960 --> 00:14:20,560
they pass the CPA exam,

424
00:14:20,560 --> 00:14:23,520
and they understand audit strategy and risk mitigation.

425
00:14:23,520 --> 00:14:26,280
Instead, they're looking for typos, they're matching numbers.

426
00:14:26,280 --> 00:14:28,480
They're being paid six figures to do data entry work

427
00:14:28,480 --> 00:14:30,440
that should have been automated a decade ago.

428
00:14:30,440 --> 00:14:32,960
But the real cost isn't the salary, it's the speed.

429
00:14:32,960 --> 00:14:34,720
Your month end clause takes five days

430
00:14:34,720 --> 00:14:37,120
because your finance team is stuck in reconciliation.

431
00:14:37,120 --> 00:14:39,160
They can't close the books until every discrepancies

432
00:14:39,160 --> 00:14:41,280
investigated every journal entry is approved,

433
00:14:41,280 --> 00:14:42,920
and every number is confirmed.

434
00:14:42,920 --> 00:14:45,160
That delay means you don't have accurate financials

435
00:14:45,160 --> 00:14:47,080
until the fifth day of the following month.

436
00:14:47,080 --> 00:14:49,320
Your CFO can't report results to the board.

437
00:14:49,320 --> 00:14:52,680
Your finance team can't move to the next month's planning.

438
00:14:52,680 --> 00:14:54,680
Your working capital decisions are delayed

439
00:14:54,680 --> 00:14:57,880
because you don't have clarity on what actually happened last month.

440
00:14:57,880 --> 00:15:00,000
An account reconciliation agent changes this.

441
00:15:00,000 --> 00:15:01,000
Here's what it does.

442
00:15:01,000 --> 00:15:03,080
It identifies mismatches automatically.

443
00:15:03,080 --> 00:15:05,840
It doesn't just flag invoice doesn't match PO.

444
00:15:05,840 --> 00:15:08,920
It analyzes the discrepancy and identifies the root cause.

445
00:15:08,920 --> 00:15:10,520
If there's a wrong GL code,

446
00:15:10,520 --> 00:15:12,280
the agent checks the invoice description

447
00:15:12,280 --> 00:15:14,120
against the purchase order specification

448
00:15:14,120 --> 00:15:15,600
and proposes the correction.

449
00:15:15,600 --> 00:15:17,000
If there's a quantity variance,

450
00:15:17,000 --> 00:15:18,440
the agent checks the receiving record

451
00:15:18,440 --> 00:15:20,600
to see if fewer units were actually received.

452
00:15:20,600 --> 00:15:21,840
If there's a timing difference,

453
00:15:21,840 --> 00:15:23,320
the agent checks the invoice date

454
00:15:23,320 --> 00:15:26,120
against the delivery date to determine whether this is a normal lag

455
00:15:26,120 --> 00:15:27,160
or an anomaly.

456
00:15:27,160 --> 00:15:28,440
The month end close starts

457
00:15:28,440 --> 00:15:30,560
and the agent pulls data from your data verse,

458
00:15:30,560 --> 00:15:32,360
including the sub-ledger, the general ledger,

459
00:15:32,360 --> 00:15:35,560
the invoice records, and your vendo master with payment terms.

460
00:15:35,560 --> 00:15:37,840
The agent applies your business rules automatically.

461
00:15:37,840 --> 00:15:41,600
For example, invoices under $500 are reconciled automatically

462
00:15:41,600 --> 00:15:43,520
if the variance is under 2%,

463
00:15:43,520 --> 00:15:46,960
while invoices over $500 require manual review

464
00:15:46,960 --> 00:15:48,600
if the variance exceeds 1%.

465
00:15:48,600 --> 00:15:51,720
The agent runs these rules across your entire invoice population

466
00:15:51,720 --> 00:15:52,920
in seconds.

467
00:15:52,920 --> 00:15:54,000
Then orchestration happens,

468
00:15:54,000 --> 00:15:55,480
the agent creates a draft journal entry

469
00:15:55,480 --> 00:15:56,920
for each exception that identifies.

470
00:15:56,920 --> 00:15:58,400
It doesn't post the entry.

471
00:15:58,400 --> 00:16:00,160
That still requires human approval,

472
00:16:00,160 --> 00:16:02,480
but it prepares the entry with the correct accounts,

473
00:16:02,480 --> 00:16:05,280
the correct amounts, and the reasoning behind the correction.

474
00:16:05,280 --> 00:16:07,120
Your controller reviews the proposed entry,

475
00:16:07,120 --> 00:16:09,080
confirms its correct, and approves it.

476
00:16:09,080 --> 00:16:11,400
That takes 30 seconds instead of 30 minutes.

477
00:16:11,400 --> 00:16:12,840
The agent handles the mechanics,

478
00:16:12,840 --> 00:16:14,280
your controller handles the judgment.

479
00:16:14,280 --> 00:16:16,320
This is the critical difference from a report.

480
00:16:16,320 --> 00:16:18,560
A reconciliation report shows you the problem.

481
00:16:18,560 --> 00:16:21,080
An account reconciliation agent shows you the problem

482
00:16:21,080 --> 00:16:22,520
and proposes the solution.

483
00:16:22,520 --> 00:16:25,880
Invoice 1, 2, 3, 4, 5 does not match PO67892

484
00:16:25,880 --> 00:16:28,400
because the vendor applied a 5% volume discount

485
00:16:28,400 --> 00:16:30,000
per your contract terms.

486
00:16:30,000 --> 00:16:31,480
Proposed journal entry,

487
00:16:31,480 --> 00:16:33,520
debit accounts payable $350,

488
00:16:33,520 --> 00:16:36,200
credit vendor discount expense $350.

489
00:16:36,200 --> 00:16:37,680
The measurable impact is direct.

490
00:16:37,680 --> 00:16:39,960
US Ventures deployed reconciliation agents

491
00:16:39,960 --> 00:16:42,880
and cut their reconciliation cycle by 80%.

492
00:16:42,880 --> 00:16:45,680
Their month and clothes compressed from five days to one day.

493
00:16:45,680 --> 00:16:47,320
That means their financial teams are available

494
00:16:47,320 --> 00:16:49,840
for strategic work instead of reconciliation tax.

495
00:16:49,840 --> 00:16:51,960
That means their CFO has numbers four days earlier.

496
00:16:51,960 --> 00:16:54,760
That means working capital decisions happen faster.

497
00:16:54,760 --> 00:16:56,880
That's not an efficiency gain in some abstract sense.

498
00:16:56,880 --> 00:16:59,560
That's a material change to how fast the business operates.

499
00:16:59,560 --> 00:17:02,360
Customer intent agent, the relationship memory.

500
00:17:02,360 --> 00:17:04,160
Customer support operates on a lie.

501
00:17:04,160 --> 00:17:06,120
The lie is that you know what a customer needs

502
00:17:06,120 --> 00:17:07,600
based on how they describe it.

503
00:17:07,600 --> 00:17:10,280
A customer calls and says, "I have a billing question."

504
00:17:10,280 --> 00:17:11,920
The ticket gets routed to billing.

505
00:17:11,920 --> 00:17:14,480
A customer writes, "I can't access my account."

506
00:17:14,480 --> 00:17:15,560
The ticket goes to IT.

507
00:17:15,560 --> 00:17:17,840
A customer says, "I need to update my contract."

508
00:17:17,840 --> 00:17:18,800
It lands in legal.

509
00:17:18,800 --> 00:17:20,040
The system reads the subject line

510
00:17:20,040 --> 00:17:21,080
and decides where to send it.

511
00:17:21,080 --> 00:17:21,920
But here's the problem.

512
00:17:21,920 --> 00:17:24,680
The customer's subject line isn't the actual issue.

513
00:17:24,680 --> 00:17:27,320
What typically happens is a customer calls about a billing issue

514
00:17:27,320 --> 00:17:29,040
and the billing team pulls up the account.

515
00:17:29,040 --> 00:17:30,360
They see the charges correct.

516
00:17:30,360 --> 00:17:31,600
They explain the policy.

517
00:17:31,600 --> 00:17:32,720
The customer gets frustrated

518
00:17:32,720 --> 00:17:35,440
because they don't actually care about the billing question.

519
00:17:35,440 --> 00:17:38,360
The real issue is their contract renewal is in 90 days.

520
00:17:38,360 --> 00:17:41,400
They're trying to negotiate the price down before they resign.

521
00:17:41,400 --> 00:17:43,320
They brought up billing as a proxy for,

522
00:17:43,320 --> 00:17:46,080
"I'm thinking about leaving unless this gets cheaper."

523
00:17:46,080 --> 00:17:48,640
The billing team has no visibility into that context.

524
00:17:48,640 --> 00:17:49,960
So they solve the wrong problem.

525
00:17:49,960 --> 00:17:52,920
The customer escalates the issue bounces to account management.

526
00:17:52,920 --> 00:17:54,160
By now you've lost two days

527
00:17:54,160 --> 00:17:56,360
and the customer is already comparison shopping.

528
00:17:56,360 --> 00:17:57,720
This is the routing problem.

529
00:17:57,720 --> 00:17:59,040
And it costs more than you think.

530
00:17:59,040 --> 00:18:00,800
The cost isn't just in resolution time.

531
00:18:00,800 --> 00:18:02,040
It's in the customer experience.

532
00:18:02,040 --> 00:18:03,800
The first agent interaction sets the tone

533
00:18:03,800 --> 00:18:05,160
for the entire resolution.

534
00:18:05,160 --> 00:18:06,720
If a customer starts out frustrated

535
00:18:06,720 --> 00:18:08,120
and gets sent to the wrong person,

536
00:18:08,120 --> 00:18:09,560
that frustration compounds.

537
00:18:09,560 --> 00:18:11,280
They have to re-explain the issue.

538
00:18:11,280 --> 00:18:13,040
The new person has to get context.

539
00:18:13,040 --> 00:18:15,080
By the time someone with real authority shows up,

540
00:18:15,080 --> 00:18:17,360
the customer is already thinking about switching vendors.

541
00:18:17,360 --> 00:18:19,600
If that first interaction had gone to the right person

542
00:18:19,600 --> 00:18:21,480
with full context already loaded,

543
00:18:21,480 --> 00:18:22,680
the conversation is different.

544
00:18:22,680 --> 00:18:25,200
It's, "I see your contract renewal is coming up."

545
00:18:25,200 --> 00:18:28,080
Let's talk about what matters to you in the next term.

546
00:18:28,080 --> 00:18:31,200
Instead of, "Wait, you actually needed account management?"

547
00:18:31,200 --> 00:18:32,360
Let me transfer you.

548
00:18:32,360 --> 00:18:35,360
A customer intent agent reads the situation differently.

549
00:18:35,360 --> 00:18:36,880
When a customer initiates contact

550
00:18:36,880 --> 00:18:39,440
and email a phone call, a support portal submission,

551
00:18:39,440 --> 00:18:41,280
the agent doesn't just read the subject line.

552
00:18:41,280 --> 00:18:43,360
It reads the entire interaction history.

553
00:18:43,360 --> 00:18:44,680
What have they called about before?

554
00:18:44,680 --> 00:18:45,760
What patterns do you see?

555
00:18:45,760 --> 00:18:47,400
The agent analyzes email tone.

556
00:18:47,400 --> 00:18:48,240
Is this frustrated?

557
00:18:48,240 --> 00:18:48,880
Curious?

558
00:18:48,880 --> 00:18:50,160
Urgent?

559
00:18:50,160 --> 00:18:51,880
The agent checks contract status.

560
00:18:51,880 --> 00:18:53,480
When does this renewal happen?

561
00:18:53,480 --> 00:18:55,000
What's their payment history?

562
00:18:55,000 --> 00:18:56,680
The agent looks at previous tickets.

563
00:18:56,680 --> 00:18:58,400
Have they called with similar issues?

564
00:18:58,400 --> 00:18:59,760
Did those get resolved?

565
00:18:59,760 --> 00:19:02,360
The agent cross-references open support cases.

566
00:19:02,360 --> 00:19:05,640
Is there an ongoing issue that this call relates to?

567
00:19:05,640 --> 00:19:06,920
The event is straightforward.

568
00:19:06,920 --> 00:19:08,640
Customer initiates contact.

569
00:19:08,640 --> 00:19:10,840
The reasoning phase pulls everything the agent knows

570
00:19:10,840 --> 00:19:12,600
about this customer into context.

571
00:19:12,600 --> 00:19:14,840
The orchestration phase roots them to the right team

572
00:19:14,840 --> 00:19:16,880
with all of that intelligence preloaded.

573
00:19:16,880 --> 00:19:18,760
Account management doesn't get a transfer.

574
00:19:18,760 --> 00:19:19,920
They get a notification.

575
00:19:19,920 --> 00:19:22,200
Customer ABC is calling about a billing question,

576
00:19:22,200 --> 00:19:24,040
but their contract renews in 90 days

577
00:19:24,040 --> 00:19:26,480
and they've been asking about pricing for three weeks.

578
00:19:26,480 --> 00:19:29,400
They're likely exploring alternatives, root to me.

579
00:19:29,400 --> 00:19:30,400
Here's what changes.

580
00:19:30,400 --> 00:19:32,240
The account manager picks up already knowing

581
00:19:32,240 --> 00:19:33,440
this isn't a billing call.

582
00:19:33,440 --> 00:19:35,040
It's a renewal negotiation.

583
00:19:35,040 --> 00:19:36,600
Suddenly, the conversation is strategic

584
00:19:36,600 --> 00:19:37,880
instead of transactional.

585
00:19:37,880 --> 00:19:38,920
The customer feels understood

586
00:19:38,920 --> 00:19:40,880
because the person on the other end already knows

587
00:19:40,880 --> 00:19:41,960
their situation.

588
00:19:41,960 --> 00:19:43,400
You're not starting from scratch.

589
00:19:43,400 --> 00:19:44,840
You're starting from context.

590
00:19:44,840 --> 00:19:47,400
First contact resolution is the metric that matters here.

591
00:19:47,400 --> 00:19:49,600
Instead of a customer getting rooted three times,

592
00:19:49,600 --> 00:19:51,680
they reach the right person on the first contact.

593
00:19:51,680 --> 00:19:52,800
That person has context.

594
00:19:52,800 --> 00:19:54,560
The resolution happens in one conversation.

595
00:19:54,560 --> 00:19:56,120
The customer stays satisfied.

596
00:19:56,120 --> 00:19:58,440
The company avoids the cost and churn risk

597
00:19:58,440 --> 00:19:59,800
of a mis-routed ticket.

598
00:19:59,800 --> 00:20:02,680
This is where the real power of customer intelligence lives.

599
00:20:02,680 --> 00:20:04,720
It's not in knowing what customers say they need.

600
00:20:04,720 --> 00:20:06,720
It's in understanding what they actually need

601
00:20:06,720 --> 00:20:08,600
based on the full relationship history.

602
00:20:08,600 --> 00:20:11,120
A customer intent agent does that automatically.

603
00:20:11,120 --> 00:20:13,680
It makes every first interaction and informed interaction.

604
00:20:13,680 --> 00:20:16,240
And that changes customer outcomes fundamentally.

605
00:20:16,240 --> 00:20:19,760
Supplier communications agent, the supply chain orchestrator,

606
00:20:19,760 --> 00:20:22,240
procurement operates on a fundamentally different cadence

607
00:20:22,240 --> 00:20:23,520
than sales or finance.

608
00:20:23,520 --> 00:20:25,360
You don't get one supplier email a day.

609
00:20:25,360 --> 00:20:27,040
You get 50.

610
00:20:27,040 --> 00:20:28,120
A supplier sends.

611
00:20:28,120 --> 00:20:29,560
We can't deliver on Tuesday.

612
00:20:29,560 --> 00:20:32,160
Friday is the earliest we can ship.

613
00:20:32,160 --> 00:20:33,480
An account sends.

614
00:20:33,480 --> 00:20:36,120
Your PO 45892 has been received.

615
00:20:36,120 --> 00:20:38,440
But we need clarification on the address.

616
00:20:38,440 --> 00:20:40,280
A logistics provider updates you.

617
00:20:40,280 --> 00:20:42,360
Your shipment is delayed due to weather.

618
00:20:42,360 --> 00:20:44,960
A manufacturer confirms we've expedited your order

619
00:20:44,960 --> 00:20:46,800
and it will arrive three days early.

620
00:20:46,800 --> 00:20:49,360
Each of these emails requires the same sequence.

621
00:20:49,360 --> 00:20:51,840
Read it, check the system, understand the impact,

622
00:20:51,840 --> 00:20:54,040
take action, notify downstream teams.

623
00:20:54,040 --> 00:20:55,960
And it happens dozens of times a week.

624
00:20:55,960 --> 00:20:57,320
The communication tax in procurement

625
00:20:57,320 --> 00:20:59,440
is different from the data entry tax in sales.

626
00:20:59,440 --> 00:21:00,880
It's not about volume of records.

627
00:21:00,880 --> 00:21:02,920
It's about the velocity of decision making.

628
00:21:02,920 --> 00:21:05,160
Every supplier email that doesn't get processed

629
00:21:05,160 --> 00:21:07,120
immediately creates a cascading problem.

630
00:21:07,120 --> 00:21:09,240
Your warehouse doesn't know the shipment is delayed

631
00:21:09,240 --> 00:21:11,200
so it schedules receiving incorrectly.

632
00:21:11,200 --> 00:21:13,520
Your production line doesn't know the component order

633
00:21:13,520 --> 00:21:15,400
is expedited so you have double inventory.

634
00:21:15,400 --> 00:21:17,840
Your project doesn't know the material won't arrive on time

635
00:21:17,840 --> 00:21:19,960
so you're still running on the original timeline.

636
00:21:19,960 --> 00:21:22,880
The person doing this work is trapped between two bad options.

637
00:21:22,880 --> 00:21:25,520
If it's a buyer, someone with authority and context,

638
00:21:25,520 --> 00:21:29,480
you're paying $100,000 a year to read emails and update systems.

639
00:21:29,480 --> 00:21:32,320
If it's an admin, someone cheaper without supply chain knowledge,

640
00:21:32,320 --> 00:21:34,040
they're making guesses about impact.

641
00:21:34,040 --> 00:21:35,480
They escalate everything out of caution

642
00:21:35,480 --> 00:21:37,560
because they don't understand the consequences.

643
00:21:37,560 --> 00:21:40,200
A supplier communications agent operates differently.

644
00:21:40,200 --> 00:21:41,360
Here's the mechanism.

645
00:21:41,360 --> 00:21:43,720
An email arrives from a supplier.

646
00:21:43,720 --> 00:21:46,080
Your shipment of component X is delayed by one week

647
00:21:46,080 --> 00:21:48,200
due to quality issues.

648
00:21:48,200 --> 00:21:49,680
The agent reads it.

649
00:21:49,680 --> 00:21:51,000
But it doesn't just log the delay.

650
00:21:51,000 --> 00:21:53,560
It understands what that delay means for your operation.

651
00:21:53,560 --> 00:21:55,000
The event is the email itself.

652
00:21:55,000 --> 00:21:57,200
The reasoning phase is where the intelligence lives.

653
00:21:57,200 --> 00:21:59,800
The agent checks your current inventory of component X.

654
00:21:59,800 --> 00:22:00,960
How much do you have on hand?

655
00:22:00,960 --> 00:22:02,080
What's your consumption rate?

656
00:22:02,080 --> 00:22:04,360
When would you run out if nothing else arrives?

657
00:22:04,360 --> 00:22:06,280
The agent pulls your safety stock policy.

658
00:22:06,280 --> 00:22:09,360
For critical components, you maintain 30 days of supply.

659
00:22:09,360 --> 00:22:12,120
For non-critical items, you're comfortable with 15.

660
00:22:12,120 --> 00:22:14,440
Based on the classification and your current inventory,

661
00:22:14,440 --> 00:22:16,040
how urgent is this delay?

662
00:22:16,040 --> 00:22:17,960
The agent checks your production schedule.

663
00:22:17,960 --> 00:22:20,280
Do you have an active work order that needs component X

664
00:22:20,280 --> 00:22:21,040
in the next week?

665
00:22:21,040 --> 00:22:23,320
Is there a customer commitment that depends on this?

666
00:22:23,320 --> 00:22:25,120
The agent understands impact.

667
00:22:25,120 --> 00:22:26,840
A week delay on a common fastener

668
00:22:26,840 --> 00:22:28,760
affects your production time line slightly.

669
00:22:28,760 --> 00:22:30,880
A week delay on a custom manufactured module

670
00:22:30,880 --> 00:22:33,760
designed for an active customer order is a critical issue.

671
00:22:33,760 --> 00:22:35,200
The agent knows the difference

672
00:22:35,200 --> 00:22:37,560
because it has access to your operational data.

673
00:22:37,560 --> 00:22:38,600
Then orchestration.

674
00:22:38,600 --> 00:22:41,000
The agent updates the PO in your system automatically.

675
00:22:41,000 --> 00:22:43,080
It creates a notification to your warehouse

676
00:22:43,080 --> 00:22:44,840
that this delivery is delayed.

677
00:22:44,840 --> 00:22:47,320
If the delay is non-critical, the component arrives

678
00:22:47,320 --> 00:22:50,320
Thursday instead of Monday, and you have 60 days of supply.

679
00:22:50,320 --> 00:22:51,920
The agent handles it autonomously.

680
00:22:51,920 --> 00:22:52,880
The record is updated.

681
00:22:52,880 --> 00:22:54,200
The warehouse is notified.

682
00:22:54,200 --> 00:22:56,400
The downstream team knows the new arrival date.

683
00:22:56,400 --> 00:22:59,080
If the delay is critical, you have only five days of supply

684
00:22:59,080 --> 00:23:01,120
and component X is needed for an active order.

685
00:23:01,120 --> 00:23:03,240
The agent flags it for immediate escalation.

686
00:23:03,240 --> 00:23:04,880
It goes to your procurement manager

687
00:23:04,880 --> 00:23:06,280
and your operations lead.

688
00:23:06,280 --> 00:23:07,640
The guard rail is precise.

689
00:23:07,640 --> 00:23:09,640
Routine updates happen automatically.

690
00:23:09,640 --> 00:23:11,480
High-risk changes require approval.

691
00:23:11,480 --> 00:23:13,760
A three-day delay on a standardized commodity.

692
00:23:13,760 --> 00:23:14,840
The agent updates it.

693
00:23:14,840 --> 00:23:16,360
A redesign request from a supplier

694
00:23:16,360 --> 00:23:18,080
that changes the specifications.

695
00:23:18,080 --> 00:23:19,440
The agent escalates it.

696
00:23:19,440 --> 00:23:22,400
A price increase that affects your cost model escalated.

697
00:23:22,400 --> 00:23:24,120
An expedited delivery that adds cost

698
00:23:24,120 --> 00:23:27,120
but solves a production problem, escalated with the cost impact

699
00:23:27,120 --> 00:23:28,840
clearly stated so your manager can approve

700
00:23:28,840 --> 00:23:30,280
or decline based on the math.

701
00:23:30,280 --> 00:23:33,200
The outcome is a compression of decision velocity.

702
00:23:33,200 --> 00:23:35,520
Supplier communications that used to take a day or two

703
00:23:35,520 --> 00:23:37,760
to process now happen in seconds.

704
00:23:37,760 --> 00:23:39,480
Your production schedule stays synchronized

705
00:23:39,480 --> 00:23:41,800
with supplier reality because you're not operating

706
00:23:41,800 --> 00:23:43,280
on yesterday's information.

707
00:23:43,280 --> 00:23:45,880
When a shipment delay actually does require human judgment,

708
00:23:45,880 --> 00:23:48,280
it arrives on the right desk already analyzed.

709
00:23:48,280 --> 00:23:50,000
The business impact is clearly stated.

710
00:23:50,000 --> 00:23:52,440
Your procurement team isn't wasting time reading emails.

711
00:23:52,440 --> 00:23:54,560
They are making strategic decisions about exceptions.

712
00:23:54,560 --> 00:23:56,000
Your operation isn't held up waiting

713
00:23:56,000 --> 00:23:58,400
for someone to process a routine supplier update.

714
00:23:58,400 --> 00:24:00,240
The system is processing it automatically.

715
00:24:00,240 --> 00:24:03,040
Now, field service agent, the context carrier.

716
00:24:03,040 --> 00:24:05,240
Field service operates on a completely different problem

717
00:24:05,240 --> 00:24:07,640
than sales, finance, or even supply chain.

718
00:24:07,640 --> 00:24:09,320
The problem isn't data entry volume.

719
00:24:09,320 --> 00:24:10,800
It's not reconciliation cycles.

720
00:24:10,800 --> 00:24:13,760
It's fragmentation of context across a distributed workforce.

721
00:24:13,760 --> 00:24:16,400
Your field technicians are literally spread across geography.

722
00:24:16,400 --> 00:24:18,440
Driving to customer sites, working in conditions

723
00:24:18,440 --> 00:24:20,800
where they might not have reliable network access.

724
00:24:20,800 --> 00:24:22,120
They're making decisions on the fly

725
00:24:22,120 --> 00:24:23,600
with incomplete information.

726
00:24:23,600 --> 00:24:25,840
And those decisions have a direct profit impact

727
00:24:25,840 --> 00:24:28,640
that sales or finance teams rarely experience.

728
00:24:28,640 --> 00:24:30,440
But here's the problem.

729
00:24:30,440 --> 00:24:32,080
A work order gets created.

730
00:24:32,080 --> 00:24:34,200
A technician's mobile app notifies them.

731
00:24:34,200 --> 00:24:36,560
Work order at Boo-O-45612.

732
00:24:36,560 --> 00:24:39,120
Equipment failure at customer location, service level,

733
00:24:39,120 --> 00:24:40,320
four hour response.

734
00:24:40,320 --> 00:24:43,040
The technician sees a ticket number and an address.

735
00:24:43,040 --> 00:24:44,200
That's their context.

736
00:24:44,200 --> 00:24:45,640
They drive to the customer site.

737
00:24:45,640 --> 00:24:46,160
They arrive.

738
00:24:46,160 --> 00:24:47,400
They look at the equipment.

739
00:24:47,400 --> 00:24:48,360
It's an HVAC unit.

740
00:24:48,360 --> 00:24:50,920
The serial number is XYZ2847.

741
00:24:50,920 --> 00:24:52,320
But the technician has no idea what's

742
00:24:52,320 --> 00:24:53,920
been done to this unit before.

743
00:24:53,920 --> 00:24:55,680
Has it been serviced in the last six months?

744
00:24:55,680 --> 00:24:57,800
Was a similar failure reported last year?

745
00:24:57,800 --> 00:24:58,800
What parts were used?

746
00:24:58,800 --> 00:24:59,920
What's the warranty status?

747
00:24:59,920 --> 00:25:01,800
The technician starts diagnosing from scratch.

748
00:25:01,800 --> 00:25:03,760
They spend 30 minutes pulling the cover,

749
00:25:03,760 --> 00:25:06,080
checking the circuit board and testing components

750
00:25:06,080 --> 00:25:08,640
before they finally find a loose connection that was already

751
00:25:08,640 --> 00:25:11,440
tightened six months ago during the last service call.

752
00:25:11,440 --> 00:25:13,160
Because the technician just spent 30 minutes

753
00:25:13,160 --> 00:25:15,160
rediscovering a fix that's already documented

754
00:25:15,160 --> 00:25:17,840
in your system, you've already lost money on the labor.

755
00:25:17,840 --> 00:25:19,720
Or they identify a component failure

756
00:25:19,720 --> 00:25:21,480
that requires a part that's not on the truck, which

757
00:25:21,480 --> 00:25:23,000
means they have to drive back to the warehouse

758
00:25:23,000 --> 00:25:25,240
to get it and then drive back to the site.

759
00:25:25,240 --> 00:25:26,800
Total cost to resolve a problem that

760
00:25:26,800 --> 00:25:28,480
could have been fixed in one visit.

761
00:25:28,480 --> 00:25:30,520
Two trips plus three hours of technician time

762
00:25:30,520 --> 00:25:31,800
plus customer frustration.

763
00:25:31,800 --> 00:25:33,240
This is the information silo.

764
00:25:33,240 --> 00:25:34,240
And it's expensive.

765
00:25:34,240 --> 00:25:35,680
Field service margins are thin.

766
00:25:35,680 --> 00:25:37,160
A technician's time costs money.

767
00:25:37,160 --> 00:25:38,040
A vehicle costs money.

768
00:25:38,040 --> 00:25:39,360
Fuel costs money.

769
00:25:39,360 --> 00:25:41,040
And when a technician has to make two trips

770
00:25:41,040 --> 00:25:43,200
because they didn't have context the first time,

771
00:25:43,200 --> 00:25:45,480
you've just doubled the cost of that job.

772
00:25:45,480 --> 00:25:48,360
If that happens 20% of the time across your field service

773
00:25:48,360 --> 00:25:50,920
operation, you're looking at significant profit loss

774
00:25:50,920 --> 00:25:53,240
that could have been prevented by a technician

775
00:25:53,240 --> 00:25:55,640
walking onto a job site with full context.

776
00:25:55,640 --> 00:25:57,840
A field service agent solves this differently.

777
00:25:57,840 --> 00:25:59,680
When a work order gets created and assigned

778
00:25:59,680 --> 00:26:02,800
to a technician, the agent doesn't just create a notification

779
00:26:02,800 --> 00:26:03,720
with a ticket number.

780
00:26:03,720 --> 00:26:06,280
It synthesizes everything you know about this equipment

781
00:26:06,280 --> 00:26:09,240
and this situation into a brief that the technician can review

782
00:26:09,240 --> 00:26:11,800
on their mobile device before they ever leave the truck.

783
00:26:11,800 --> 00:26:13,760
The event trigger is the work order dispatch.

784
00:26:13,760 --> 00:26:16,240
The reasoning phase is where the intelligence accumulates.

785
00:26:16,240 --> 00:26:18,680
The agent pulls the equipment service history

786
00:26:18,680 --> 00:26:21,240
to see how many times it has been serviced, what the issues

787
00:26:21,240 --> 00:26:22,920
were, and how they were resolved.

788
00:26:22,920 --> 00:26:24,520
It checks for patterns to see if there

789
00:26:24,520 --> 00:26:26,880
are recurring failures or a component that

790
00:26:26,880 --> 00:26:28,560
fails every 18 months.

791
00:26:28,560 --> 00:26:31,760
The agent pulls IoT data if the equipment is connected.

792
00:26:31,760 --> 00:26:34,160
Temperature, pressure, uptime statistics,

793
00:26:34,160 --> 00:26:36,680
it looks for sensor readings that point to a specific failure

794
00:26:36,680 --> 00:26:37,240
mode.

795
00:26:37,240 --> 00:26:38,720
The agent checks parts inventory.

796
00:26:38,720 --> 00:26:41,320
If this equipment typically needs component X when

797
00:26:41,320 --> 00:26:43,560
this issue occurs is component X in stock,

798
00:26:43,560 --> 00:26:45,400
if not, can it be delivered to the job site

799
00:26:45,400 --> 00:26:47,480
or does the technician need to know to order it remotely

800
00:26:47,480 --> 00:26:49,560
before arriving, then orchestration?

801
00:26:49,560 --> 00:26:51,440
The agent prepares a mobile optimized briefing

802
00:26:51,440 --> 00:26:52,560
for the technician.

803
00:26:52,560 --> 00:26:54,160
Not a report, a briefing.

804
00:26:54,160 --> 00:26:56,600
It's formatted for someone who's literally in their vehicle

805
00:26:56,600 --> 00:26:58,680
or walking up to a customer's equipment.

806
00:26:58,680 --> 00:27:03,160
Equipment, HVAC XYZ2847, last service, six months ago.

807
00:27:03,160 --> 00:27:05,160
Lose connection repaired, current issue,

808
00:27:05,160 --> 00:27:06,840
similar symptoms reported.

809
00:27:06,840 --> 00:27:10,160
Recommendation, check, capacitor connection first.

810
00:27:10,160 --> 00:27:12,440
80% probability this is the same issue.

811
00:27:12,440 --> 00:27:14,080
If confirmed, no parts needed.

812
00:27:14,080 --> 00:27:16,400
If different, component cooling fan is in stock

813
00:27:16,400 --> 00:27:18,400
and will be staged at your next location.

814
00:27:18,400 --> 00:27:19,960
The agent also alerts the warehouse

815
00:27:19,960 --> 00:27:21,720
that a cooling fan might be needed.

816
00:27:21,720 --> 00:27:23,360
So if the technician does need it,

817
00:27:23,360 --> 00:27:24,840
it's already pulled and ready for pickup

818
00:27:24,840 --> 00:27:26,240
on the way to the next job.

819
00:27:26,240 --> 00:27:28,080
So what's actually happening is the technician

820
00:27:28,080 --> 00:27:30,520
walks up to the equipment with context already loaded.

821
00:27:30,520 --> 00:27:33,240
They know what was done before, but they know what probably failed.

822
00:27:33,240 --> 00:27:34,960
They know what parts are available.

823
00:27:34,960 --> 00:27:37,560
They complete the job on the first visit in most cases.

824
00:27:37,560 --> 00:27:40,400
First time fix rate improves, cost per job drops.

825
00:27:40,400 --> 00:27:42,760
Customer satisfaction increases because they're not waiting

826
00:27:42,760 --> 00:27:44,360
for a second technician visit.

827
00:27:44,360 --> 00:27:46,160
Your margins improve because you're not wasting

828
00:27:46,160 --> 00:27:47,840
technician time on rediscovery.

829
00:27:47,840 --> 00:27:49,400
Field service margins are thin.

830
00:27:49,400 --> 00:27:50,920
Every hour saved is profit.

831
00:27:50,920 --> 00:27:53,040
A field service agent gives technicians the context

832
00:27:53,040 --> 00:27:55,520
they need to work smarter, not just faster.

833
00:27:56,320 --> 00:27:59,120
The Dataverse Foundation, why context matters?

834
00:27:59,120 --> 00:28:01,480
All of these agents work because they're rooted in something

835
00:28:01,480 --> 00:28:03,480
generic AI can never access.

836
00:28:03,480 --> 00:28:06,080
Your business data, specifically Dataverse.

837
00:28:06,080 --> 00:28:09,200
It's the data platform that lives underneath Dynamics 365.

838
00:28:09,200 --> 00:28:10,880
Not a database in the traditional sense

839
00:28:10,880 --> 00:28:12,880
where you query tables and extract records.

840
00:28:12,880 --> 00:28:15,080
It's a structured, governed relational platform

841
00:28:15,080 --> 00:28:18,200
designed to store the entities that make your business run.

842
00:28:18,200 --> 00:28:20,680
Your accounts, your contacts, your orders, your invoices,

843
00:28:20,680 --> 00:28:22,840
your purchase orders, your cases, your work orders,

844
00:28:22,840 --> 00:28:25,320
your opportunities, all of it lives in Dataverse.

845
00:28:25,320 --> 00:28:27,720
This matters enormously for how agents work.

846
00:28:27,720 --> 00:28:30,440
When the sales qualification agent researchers are prospect,

847
00:28:30,440 --> 00:28:31,920
it's not calling out to the internet

848
00:28:31,920 --> 00:28:34,080
or hallucinating based on general knowledge.

849
00:28:34,080 --> 00:28:35,360
It's querying Dataverse.

850
00:28:35,360 --> 00:28:37,840
It's checking if you've worked with this company before

851
00:28:37,840 --> 00:28:40,400
what industry they're in, how long they've been a customer

852
00:28:40,400 --> 00:28:42,240
and what their payment history looks like.

853
00:28:42,240 --> 00:28:44,640
It's checking your account records, your contact records,

854
00:28:44,640 --> 00:28:46,200
and your opportunity history.

855
00:28:46,200 --> 00:28:47,840
When the account reconciliation agent

856
00:28:47,840 --> 00:28:50,040
analyzes a discrepancy, it's not guessing.

857
00:28:50,040 --> 00:28:51,680
It's pulling subledges from Dataverse,

858
00:28:51,680 --> 00:28:53,440
comparing them to GL entries,

859
00:28:53,440 --> 00:28:55,400
checking vendor master records for payment terms

860
00:28:55,400 --> 00:28:57,200
and checking your tolerance rules.

861
00:28:57,200 --> 00:28:59,480
When the customer intent agent understands why a customer

862
00:28:59,480 --> 00:29:01,920
is calling, it's not inferring from tone alone.

863
00:29:01,920 --> 00:29:04,040
It's querying the entire interaction history

864
00:29:04,040 --> 00:29:06,320
stored in Dataverse, previous cases,

865
00:29:06,320 --> 00:29:10,240
previous emails, contract status, payment records.

866
00:29:10,240 --> 00:29:12,320
The agent doesn't have to guess what your business logic is.

867
00:29:12,320 --> 00:29:14,120
It reads it from Dataverse.

868
00:29:14,120 --> 00:29:15,480
And here's what most people miss.

869
00:29:15,480 --> 00:29:17,240
Security alignment is built in.

870
00:29:17,240 --> 00:29:20,160
Your users access Dataverse with specific permissions.

871
00:29:20,160 --> 00:29:21,760
Sarah in the Denver Sales Office

872
00:29:21,760 --> 00:29:23,560
can see opportunities in her territory,

873
00:29:23,560 --> 00:29:26,160
but she can't see opportunities in the Seattle region

874
00:29:26,160 --> 00:29:28,520
because row-level security handles that.

875
00:29:28,520 --> 00:29:30,520
Mike in Finance can create journal entries,

876
00:29:30,520 --> 00:29:33,240
but can't delete them because of privilege-based access control.

877
00:29:33,240 --> 00:29:35,320
These aren't features bolted onto Dataverse.

878
00:29:35,320 --> 00:29:37,440
They're foundational to how Data Access works.

879
00:29:37,440 --> 00:29:39,840
Agents inherit those same security constraints.

880
00:29:39,840 --> 00:29:42,440
An agent that runs under the sales security role

881
00:29:42,440 --> 00:29:45,960
can only see and act on what a sales user can see and act on.

882
00:29:45,960 --> 00:29:47,920
It can't circumvent row-level security.

883
00:29:47,920 --> 00:29:49,240
It can't escalate privileges.

884
00:29:49,240 --> 00:29:51,480
It respects the same boundaries your employees do.

885
00:29:51,480 --> 00:29:53,600
This isn't a special agent security mode.

886
00:29:53,600 --> 00:29:54,880
It's the same security model.

887
00:29:54,880 --> 00:29:56,680
The agent operates within Dataverse

888
00:29:56,680 --> 00:29:59,000
using the same authentication and authorization layer

889
00:29:59,000 --> 00:29:59,720
as your users.

890
00:29:59,720 --> 00:30:02,520
That means you're not building a parallel governance structure.

891
00:30:02,520 --> 00:30:04,360
You're using the governance you already have.

892
00:30:04,360 --> 00:30:06,960
The integration pattern is simple because it's unified.

893
00:30:06,960 --> 00:30:09,280
You don't build custom connectors for each agent.

894
00:30:09,280 --> 00:30:11,840
You don't have API sprawl where every agent needs

895
00:30:11,840 --> 00:30:13,280
its own integration point.

896
00:30:13,280 --> 00:30:15,360
Dataverse is the universal interface.

897
00:30:15,360 --> 00:30:18,440
The sales qualification agent, the account reconciliation agent,

898
00:30:18,440 --> 00:30:20,040
and the customer intent agent,

899
00:30:20,040 --> 00:30:22,440
all query Dataverse using the same mechanisms.

900
00:30:22,440 --> 00:30:24,000
They all respect the same security.

901
00:30:24,000 --> 00:30:26,440
They all read from the same authoritative source.

902
00:30:26,440 --> 00:30:28,800
This creates something that's more subtle but more powerful.

903
00:30:28,800 --> 00:30:29,560
Coordination.

904
00:30:29,560 --> 00:30:31,840
As you add more agents to your organization,

905
00:30:31,840 --> 00:30:34,160
they're not isolated systems doing their own thing.

906
00:30:34,160 --> 00:30:36,720
They're all operating from the same Data Foundation.

907
00:30:36,720 --> 00:30:38,760
The sales qualification agent enriches a lead

908
00:30:38,760 --> 00:30:41,720
with company information and writes it back to Dataverse.

909
00:30:41,720 --> 00:30:45,000
The account reconciliation agent reads that same account record

910
00:30:45,000 --> 00:30:48,400
and uses the company size to determine which tolerance rules apply.

911
00:30:48,400 --> 00:30:50,840
The customer intent agent pulls the accounts contract status

912
00:30:50,840 --> 00:30:52,400
from the same Dataverse record.

913
00:30:52,400 --> 00:30:55,120
They're not orchestrating through APIs and web hooks.

914
00:30:55,120 --> 00:30:57,200
They're coordinating through shared context.

915
00:30:57,200 --> 00:30:58,680
When one agent updates data,

916
00:30:58,680 --> 00:31:01,040
the next agent immediately sees the updated state

917
00:31:01,040 --> 00:31:03,280
because they're all working from the same source of truth.

918
00:31:03,280 --> 00:31:04,320
That's Dataverse's role.

919
00:31:04,320 --> 00:31:05,480
It's not just a database.

920
00:31:05,480 --> 00:31:08,000
It's the operating system for agent workflows.

921
00:31:08,000 --> 00:31:10,640
It's where business logic lives, security is enforced,

922
00:31:10,640 --> 00:31:12,400
and all agents share context.

923
00:31:12,400 --> 00:31:13,560
Without Dataverse grounding,

924
00:31:13,560 --> 00:31:15,280
these agents would be chatbots.

925
00:31:15,280 --> 00:31:17,160
With it, there are autonomous participants

926
00:31:17,160 --> 00:31:19,240
in your actual business processes.

927
00:31:19,240 --> 00:31:21,320
The event reasoning orchestration framework.

928
00:31:21,320 --> 00:31:22,800
We've looked at how this pattern works

929
00:31:22,800 --> 00:31:24,160
across five different agents.

930
00:31:24,160 --> 00:31:26,680
Now, let's look at what's actually happening underneath.

931
00:31:26,680 --> 00:31:29,200
Every agent follows the same three-part architecture.

932
00:31:29,200 --> 00:31:33,080
It's the foundational pattern for every agent workflow in Dynamics 365.

933
00:31:33,080 --> 00:31:34,640
It isn't just for sales or finance.

934
00:31:34,640 --> 00:31:36,040
Once you understand this framework,

935
00:31:36,040 --> 00:31:38,400
you can apply it to any process in your company.

936
00:31:38,400 --> 00:31:39,880
The first part is the event.

937
00:31:39,880 --> 00:31:42,480
Something happens in the real world that needs a response.

938
00:31:42,480 --> 00:31:43,960
An email arrives from a prospect

939
00:31:43,960 --> 00:31:46,160
or a purchase order shows up in the system.

940
00:31:46,160 --> 00:31:47,960
A customer starts a support ticket

941
00:31:47,960 --> 00:31:50,000
or a supplier sends a status update.

942
00:31:50,000 --> 00:31:52,160
Maybe a work order is created in the field

943
00:31:52,160 --> 00:31:54,240
or the month-end closed process triggers.

944
00:31:54,240 --> 00:31:55,680
These are not scheduled tasks.

945
00:31:55,680 --> 00:31:57,960
You don't manually click a button to start them.

946
00:31:57,960 --> 00:32:00,040
The event is the exact moment something changes

947
00:32:00,040 --> 00:32:01,120
in your business.

948
00:32:01,120 --> 00:32:03,480
That change requires a decision and an action.

949
00:32:03,480 --> 00:32:05,600
This is a shift from how traditional automation works.

950
00:32:05,600 --> 00:32:07,080
Most workflows are scheduled.

951
00:32:07,080 --> 00:32:09,040
You run a nightly batch for reconciliation

952
00:32:09,040 --> 00:32:10,440
or a weekly job for lead scoring.

953
00:32:10,440 --> 00:32:11,840
It's all batch-oriented.

954
00:32:11,840 --> 00:32:13,320
An agent doesn't work that way.

955
00:32:13,320 --> 00:32:14,720
It is event-driven.

956
00:32:14,720 --> 00:32:17,120
The moment a prospect emails the agent responds.

957
00:32:17,120 --> 00:32:19,520
The moment a delivery date changes, the agent reacts.

958
00:32:19,520 --> 00:32:21,640
This is continuous operation.

959
00:32:21,640 --> 00:32:24,120
It's why agents compress your operational cycles,

960
00:32:24,120 --> 00:32:25,880
but the event is just the trigger.

961
00:32:25,880 --> 00:32:28,000
What matters is what happens next.

962
00:32:28,000 --> 00:32:30,040
The second part is the reasoning phase.

963
00:32:30,040 --> 00:32:32,040
The agent analyzes the event in context.

964
00:32:32,040 --> 00:32:33,800
This is where the intelligence layer lives.

965
00:32:33,800 --> 00:32:35,520
The agent doesn't just react blindly.

966
00:32:35,520 --> 00:32:38,240
It pulls data, checks rules, and compares patterns.

967
00:32:38,240 --> 00:32:39,880
For the sales qualification agent,

968
00:32:39,880 --> 00:32:41,480
reasoning means checking your wind data

969
00:32:41,480 --> 00:32:43,000
to see if a prospect fits.

970
00:32:43,000 --> 00:32:44,320
For the reconciliation agent,

971
00:32:44,320 --> 00:32:46,320
it means pulling sub-ledges and comparing amounts

972
00:32:46,320 --> 00:32:47,240
against your rules.

973
00:32:47,240 --> 00:32:48,800
For the customer intent agent,

974
00:32:48,800 --> 00:32:50,480
it means reading the entire history

975
00:32:50,480 --> 00:32:52,320
to see what's driving the contact.

976
00:32:52,320 --> 00:32:54,160
For the supplier communications agent,

977
00:32:54,160 --> 00:32:58,040
it means checking inventory to see if a delay is actually a crisis.

978
00:32:58,040 --> 00:33:01,000
This reasoning layer is where AI really contributes.

979
00:33:01,000 --> 00:33:02,600
It isn't just simple pattern matching.

980
00:33:02,600 --> 00:33:04,720
A rule engine can match an invoice

981
00:33:04,720 --> 00:33:06,800
if the amount is within 2% of the order,

982
00:33:06,800 --> 00:33:08,680
but what if the amount is off by 4%,

983
00:33:08,680 --> 00:33:10,560
a rule engine just stops and asks for help.

984
00:33:10,560 --> 00:33:11,960
An agent reasons about it.

985
00:33:11,960 --> 00:33:14,040
It checks the purchase order for a documented variance.

986
00:33:14,040 --> 00:33:15,920
It looks at vendor terms for a discount.

987
00:33:15,920 --> 00:33:18,800
It compares dates to see if this is just a timing difference.

988
00:33:18,800 --> 00:33:21,920
The agent uses language reasoning to understand context.

989
00:33:21,920 --> 00:33:23,560
It isn't just following hard rules.

990
00:33:23,560 --> 00:33:25,040
It reasons the way a human would,

991
00:33:25,040 --> 00:33:27,640
but it does it instantly against your entire data set.

992
00:33:27,640 --> 00:33:30,160
This distinction is important because the world is messy.

993
00:33:30,160 --> 00:33:31,800
Your business operates in ambiguity.

994
00:33:31,800 --> 00:33:34,080
Sometimes a reconciliation error is just a typo.

995
00:33:34,080 --> 00:33:35,760
Sometimes it's a legitimate variance.

996
00:33:35,760 --> 00:33:37,440
An agent can tell the difference.

997
00:33:37,440 --> 00:33:39,280
A lead might look like a tire kicker,

998
00:33:39,280 --> 00:33:42,320
but there might be a prospect you're already chasing elsewhere.

999
00:33:42,320 --> 00:33:43,960
The agent surfaces that context.

1000
00:33:43,960 --> 00:33:46,400
Reasoning allows automation to handle real complexity

1001
00:33:46,400 --> 00:33:48,080
instead of just perfect scenarios.

1002
00:33:48,080 --> 00:33:49,520
The third part is orchestration.

1003
00:33:49,520 --> 00:33:51,800
Once the agent has reasoned and made a decision,

1004
00:33:51,800 --> 00:33:53,720
it executes across your systems.

1005
00:33:53,720 --> 00:33:56,920
It creates a record, sensor notification, or updates a status.

1006
00:33:56,920 --> 00:34:00,080
It might escalate to a human or call an approval workflow.

1007
00:34:00,080 --> 00:34:02,520
For the sales agent, orchestration means enriching the lead

1008
00:34:02,520 --> 00:34:03,960
and flagging it for the team.

1009
00:34:03,960 --> 00:34:05,160
For the reconciliation agent,

1010
00:34:05,160 --> 00:34:07,640
it means creating a draft journal entry.

1011
00:34:07,640 --> 00:34:09,040
For the customer intent agent,

1012
00:34:09,040 --> 00:34:12,400
it means rooting the ticket with the context already loaded.

1013
00:34:12,400 --> 00:34:14,280
Orchestration respects your governance.

1014
00:34:14,280 --> 00:34:16,120
An agent doesn't bypass your security

1015
00:34:16,120 --> 00:34:17,480
or escalate its own privileges.

1016
00:34:17,480 --> 00:34:18,480
It can't see data.

1017
00:34:18,480 --> 00:34:20,000
It isn't authorized to see.

1018
00:34:20,000 --> 00:34:22,880
High-risk actions still require a human to sign off.

1019
00:34:22,880 --> 00:34:25,080
Low-risk routine actions happen on their own.

1020
00:34:25,080 --> 00:34:27,800
This guardrail structure makes orchestration safe.

1021
00:34:27,800 --> 00:34:29,920
The agent stays within the boundaries you define.

1022
00:34:29,920 --> 00:34:31,960
This pattern matters because it is repeatable.

1023
00:34:31,960 --> 00:34:33,560
You aren't building custom automation

1024
00:34:33,560 --> 00:34:34,880
for every single process.

1025
00:34:34,880 --> 00:34:36,480
You're applying the same event reasoning

1026
00:34:36,480 --> 00:34:39,200
orchestration framework to whatever comes next.

1027
00:34:39,200 --> 00:34:41,800
When a new process needs automation, you use this structure.

1028
00:34:41,800 --> 00:34:44,640
The governance is consistent and the tooling is the same.

1029
00:34:44,640 --> 00:34:47,360
That is how you scale agents across an entire organization.

1030
00:34:47,360 --> 00:34:50,520
The governance reality, agent 365 and the agent hub,

1031
00:34:50,520 --> 00:34:53,000
there is a problem lurking underneath everything we've discussed.

1032
00:34:53,000 --> 00:34:54,360
It isn't the agents themselves.

1033
00:34:54,360 --> 00:34:55,760
It's what happens when you deploy them

1034
00:34:55,760 --> 00:34:57,000
without a governance structure.

1035
00:34:57,000 --> 00:34:58,200
It's called shadow AI.

1036
00:34:58,200 --> 00:35:00,320
It is the compliance version of shadow IT.

1037
00:35:00,320 --> 00:35:02,040
Someone in procurement builds an agent

1038
00:35:02,040 --> 00:35:04,160
to update the ERP from supplier emails.

1039
00:35:04,160 --> 00:35:05,520
It works great for them.

1040
00:35:05,520 --> 00:35:08,160
Then finance builds an agent for expense reports.

1041
00:35:08,160 --> 00:35:09,920
Then sales builds one for lead routing.

1042
00:35:09,920 --> 00:35:11,240
Each team solves their own problem.

1043
00:35:11,240 --> 00:35:12,640
But now you have five agents running

1044
00:35:12,640 --> 00:35:13,760
and nobody knows what they're doing.

1045
00:35:13,760 --> 00:35:15,040
Nobody knows who authorised them

1046
00:35:15,040 --> 00:35:16,640
or if they're making good decisions.

1047
00:35:16,640 --> 00:35:19,560
The CFO asks which agents are running in the environment.

1048
00:35:19,560 --> 00:35:21,240
The security team can't answer.

1049
00:35:21,240 --> 00:35:22,600
The compliance officer doesn't know

1050
00:35:22,600 --> 00:35:24,360
if they're violating audit rules.

1051
00:35:24,360 --> 00:35:25,840
The controller doesn't know if an agent

1052
00:35:25,840 --> 00:35:27,680
just created a massive compliance gap.

1053
00:35:27,680 --> 00:35:29,120
This is the shadow AI problem.

1054
00:35:29,120 --> 00:35:31,760
It's real and it's why agent 365 matters.

1055
00:35:31,760 --> 00:35:34,720
Agent 365 is the identity and lifecycle management system

1056
00:35:34,720 --> 00:35:35,320
for agents.

1057
00:35:35,320 --> 00:35:36,800
It treats agents like employees.

1058
00:35:36,800 --> 00:35:39,320
They aren't autonomous software running unsupervised.

1059
00:35:39,320 --> 00:35:41,400
They are managed identities with defined roles

1060
00:35:41,400 --> 00:35:42,600
and accountability.

1061
00:35:42,600 --> 00:35:44,000
Every agent has a sponsor.

1062
00:35:44,000 --> 00:35:46,000
That's a person or a team responsible

1063
00:35:46,000 --> 00:35:47,480
for what the agent does.

1064
00:35:47,480 --> 00:35:49,920
The agent gets security roles just like a user.

1065
00:35:49,920 --> 00:35:51,880
Its actions are logged and auditable.

1066
00:35:51,880 --> 00:35:55,080
It has a defined life cycle from creation to retirement.

1067
00:35:55,080 --> 00:35:57,680
You don't just build an agent and let it run forever.

1068
00:35:57,680 --> 00:35:59,120
It has an owner and oversight.

1069
00:35:59,120 --> 00:36:01,280
This changes how safely you can scale.

1070
00:36:01,280 --> 00:36:04,240
The agent hub in Dynamics 365 is where this governance lives.

1071
00:36:04,240 --> 00:36:05,400
It's a central dashboard

1072
00:36:05,400 --> 00:36:07,760
that shows every agent running in your environment.

1073
00:36:07,760 --> 00:36:09,400
It isn't a manual spreadsheet.

1074
00:36:09,400 --> 00:36:10,400
It's a live inventory.

1075
00:36:10,400 --> 00:36:12,600
You see what each agent does and who sponsors it.

1076
00:36:12,600 --> 00:36:14,920
You see the security roles and how often it triggers.

1077
00:36:14,920 --> 00:36:17,080
You see what percentage of decisions go to humans.

1078
00:36:17,080 --> 00:36:19,200
You can see if performance is getting better or worse.

1079
00:36:19,200 --> 00:36:20,760
You aren't flying blind anymore.

1080
00:36:20,760 --> 00:36:23,360
This matters because you need to know if an agent is drifting.

1081
00:36:23,360 --> 00:36:25,080
When an agent launches, it might escalate

1082
00:36:25,080 --> 00:36:27,160
20% of decisions for review.

1083
00:36:27,160 --> 00:36:29,000
That's normal, while it learns your business.

1084
00:36:29,000 --> 00:36:31,400
Three months later, that might drop to 5%.

1085
00:36:31,400 --> 00:36:33,400
That's good news because the agent is improving,

1086
00:36:33,400 --> 00:36:36,080
but then you might notice it's processing 40% more exceptions

1087
00:36:36,080 --> 00:36:36,960
than last month.

1088
00:36:36,960 --> 00:36:38,520
That isn't a gain in confidence.

1089
00:36:38,520 --> 00:36:40,000
That is drift.

1090
00:36:40,000 --> 00:36:42,440
Something changed in your data or your supplier behavior.

1091
00:36:42,440 --> 00:36:43,920
The agent isn't adapting correctly.

1092
00:36:43,920 --> 00:36:46,800
The agent hub shows you that signal before it becomes a crisis.

1093
00:36:46,800 --> 00:36:49,360
High-risk actions still require a human signature.

1094
00:36:49,360 --> 00:36:50,960
This is baked into the orchestration.

1095
00:36:50,960 --> 00:36:53,520
A reconciliation agent can propose a journal entry,

1096
00:36:53,520 --> 00:36:55,760
but it can't post it if the amount is too high.

1097
00:36:55,760 --> 00:36:57,640
A procurement agent can update delivery dates

1098
00:36:57,640 --> 00:36:59,320
but it can't change a contract.

1099
00:36:59,320 --> 00:37:01,520
The approval workflow isn't a separate system.

1100
00:37:01,520 --> 00:37:02,960
It's part of how the agent operates.

1101
00:37:02,960 --> 00:37:04,080
The agent knows its limits.

1102
00:37:04,080 --> 00:37:06,760
It works on its own until it hits a boundary you've set.

1103
00:37:06,760 --> 00:37:10,560
The audit trail is the proof every decision is logged.

1104
00:37:10,560 --> 00:37:13,400
The system records the outcome and the reasoning behind it.

1105
00:37:13,400 --> 00:37:15,000
When an agent roots a customer,

1106
00:37:15,000 --> 00:37:17,920
the log shows the history it analyzed and the rule it applied.

1107
00:37:17,920 --> 00:37:19,440
When it proposes a journal entry,

1108
00:37:19,440 --> 00:37:22,360
the log shows the data sources and the business rules.

1109
00:37:22,360 --> 00:37:25,240
If an auditor asks why something happened, you don't have to guess.

1110
00:37:25,240 --> 00:37:26,480
The decision chain is right there.

1111
00:37:26,480 --> 00:37:29,000
This governance is what makes agents enterprise grade.

1112
00:37:29,000 --> 00:37:32,360
It moves them from an experiment to a real part of the business.

1113
00:37:32,360 --> 00:37:35,440
Model context protocol, the language agents speak.

1114
00:37:35,440 --> 00:37:38,160
There is a technical layer underneath everything we have discussed

1115
00:37:38,160 --> 00:37:39,960
that makes the whole system work together.

1116
00:37:39,960 --> 00:37:43,120
It is called model context protocol or MCP.

1117
00:37:43,120 --> 00:37:46,200
It is the language agents use to talk to your enterprise systems.

1118
00:37:46,200 --> 00:37:47,760
But here is the problem it solves.

1119
00:37:47,760 --> 00:37:51,480
If you have five agents that need to access your data and trigger actions,

1120
00:37:51,480 --> 00:37:53,160
do you build five separate integrations?

1121
00:37:53,160 --> 00:37:56,240
Do you write custom code for every single agent to talk to dataverse?

1122
00:37:56,240 --> 00:37:59,440
Do you create five different API layers that scales like a nightmare?

1123
00:37:59,440 --> 00:38:02,240
By the time you have 20 agents across your organization,

1124
00:38:02,240 --> 00:38:04,200
you have 20 custom integration points.

1125
00:38:04,200 --> 00:38:05,280
Every single one is brittle.

1126
00:38:05,280 --> 00:38:07,040
Every single one is a security gap.

1127
00:38:07,040 --> 00:38:09,520
And every single one requires manual maintenance,

1128
00:38:09,520 --> 00:38:11,600
the moment your business logic changes.

1129
00:38:11,600 --> 00:38:13,400
MCP inverts that problem.

1130
00:38:13,400 --> 00:38:15,720
Instead of agents calling custom APIs,

1131
00:38:15,720 --> 00:38:17,280
MCP is a standard protocol.

1132
00:38:17,280 --> 00:38:19,240
Agents use it to describe what they need.

1133
00:38:19,240 --> 00:38:21,680
And your systems respond in a consistent way.

1134
00:38:21,680 --> 00:38:25,440
Without MCP, you are building a new phone line every time you want to talk to someone.

1135
00:38:25,440 --> 00:38:29,840
With MCP, you are using a standardized communication protocol that anyone can use.

1136
00:38:29,840 --> 00:38:33,720
An agent says, "I need a list of active purchase orders from the last 30 days."

1137
00:38:33,720 --> 00:38:37,320
The MCP server hears that request and translates it into a dataverse query.

1138
00:38:37,320 --> 00:38:39,800
The agent does not have to know how to query dataverse.

1139
00:38:39,800 --> 00:38:42,440
And it does not need to understand your data schema.

1140
00:38:42,440 --> 00:38:44,640
It describes what it needs in plain language.

1141
00:38:44,640 --> 00:38:46,760
And the MCP server handles the translation.

1142
00:38:46,760 --> 00:38:50,520
This matters for scaling because it decouples agents from your infrastructure.

1143
00:38:50,520 --> 00:38:53,920
The sales qualification agent does not care whether you run dataverse.

1144
00:38:53,920 --> 00:38:57,440
SAP, Oracle, or a custom system that describes what it needs,

1145
00:38:57,440 --> 00:38:59,600
like company information or contact records.

1146
00:38:59,600 --> 00:39:03,440
And the MCP server translates that into whatever backend you are using.

1147
00:39:03,440 --> 00:39:06,400
You are not writing new code for every agent you deploy.

1148
00:39:06,400 --> 00:39:10,320
You are simply configuring which data and which operations the MCP server exposes.

1149
00:39:10,320 --> 00:39:11,560
That is a massive difference.

1150
00:39:11,560 --> 00:39:12,800
Configuration scales.

1151
00:39:12,800 --> 00:39:14,360
Custom code does not.

1152
00:39:14,360 --> 00:39:15,880
So what does this look like in practice?

1153
00:39:15,880 --> 00:39:19,640
Imagine you are building a new agent for a process that does not exist yet.

1154
00:39:19,640 --> 00:39:21,400
In the old model, you would say.

1155
00:39:21,400 --> 00:39:24,600
I need to build a custom API that this agent can call.

1156
00:39:24,600 --> 00:39:25,720
That takes development time.

1157
00:39:25,720 --> 00:39:28,640
Someone has to write code, test it, and deploy it.

1158
00:39:28,640 --> 00:39:30,880
With MCP, you just say.

1159
00:39:30,880 --> 00:39:32,680
Here is what my agent needs to do.

1160
00:39:32,680 --> 00:39:37,280
It needs to read inventory levels, check production schedules, and pull supplier contracts.

1161
00:39:37,280 --> 00:39:39,640
The MCP server already has access to all of that.

1162
00:39:39,640 --> 00:39:44,360
You configure which data the agent can see and which operations it can perform.

1163
00:39:44,360 --> 00:39:45,200
And then you are done.

1164
00:39:45,200 --> 00:39:49,480
No new code, no new API, just configuration.

1165
00:39:49,480 --> 00:39:52,960
The security implication is straightforward because it is inherited.

1166
00:39:52,960 --> 00:39:56,640
An MCP server enforces the same security roles as dataverse.

1167
00:39:56,640 --> 00:40:00,240
If a user is restricted from seeing certain data due to row-level security,

1168
00:40:00,240 --> 00:40:03,280
an agent using that MCP server is also restricted.

1169
00:40:03,280 --> 00:40:05,760
You are not building parallel security structures.

1170
00:40:05,760 --> 00:40:07,920
You are leveraging the security you already have.

1171
00:40:07,920 --> 00:40:11,080
An agent can only access what its assigned security role allows.

1172
00:40:11,080 --> 00:40:13,280
The MCP server is the enforcement point.

1173
00:40:13,280 --> 00:40:16,040
But the rules are the same ones governing your users.

1174
00:40:16,040 --> 00:40:19,200
And there is an ecosystem angle here that matters for the future.

1175
00:40:19,200 --> 00:40:20,880
MCP is becoming an industry standard.

1176
00:40:20,880 --> 00:40:23,560
It is not proprietary to Dynamics 365.

1177
00:40:23,560 --> 00:40:28,360
Anthropic, the company behind Claude, developed it as an open standard, Microsoft is adopting it.

1178
00:40:28,360 --> 00:40:32,320
If you build agents in co-pilot studio that use MCP servers, those agents can talk to your

1179
00:40:32,320 --> 00:40:37,440
dataverse, but they can also talk to SAP systems, Oracle backends or custom applications.

1180
00:40:37,440 --> 00:40:41,320
This matters because it means agents built in Microsoft's ecosystem are not trapped in

1181
00:40:41,320 --> 00:40:42,720
Microsoft infrastructure.

1182
00:40:42,720 --> 00:40:45,480
They can orchestrate across your entire technology stack.

1183
00:40:45,480 --> 00:40:49,400
For your organization right now, MCP means you are not going to have an API explosion as

1184
00:40:49,400 --> 00:40:51,000
you deploy more agents.

1185
00:40:51,000 --> 00:40:54,280
You are going to have a configuration system that gets more powerful as more agents use

1186
00:40:54,280 --> 00:40:55,280
it.

1187
00:40:55,280 --> 00:40:58,040
Each new agent you add does not increase your technical debt.

1188
00:40:58,040 --> 00:41:00,400
It increases your operational capability.

1189
00:41:00,400 --> 00:41:01,600
The skills shift.

1190
00:41:01,600 --> 00:41:03,600
What changes when agents take over?

1191
00:41:03,600 --> 00:41:06,200
Here is what nobody talks about when they discuss automation.

1192
00:41:06,200 --> 00:41:09,680
The human beings in your organization are going to feel threatened.

1193
00:41:09,680 --> 00:41:13,520
Not because they are irrational, but because they are accurate, their job is changing.

1194
00:41:13,520 --> 00:41:16,560
The question is whether you manage that change or let it happen to them.

1195
00:41:16,560 --> 00:41:20,720
Right now, most people in your organization spend their time operating the system.

1196
00:41:20,720 --> 00:41:24,160
A sales development rep spends their day entering leads into sales force.

1197
00:41:24,160 --> 00:41:25,800
They are not thinking about sales strategy.

1198
00:41:25,800 --> 00:41:26,800
They are entering data.

1199
00:41:26,800 --> 00:41:31,000
A finance analyst spends their day matching invoices and creating journal entries.

1200
00:41:31,000 --> 00:41:33,320
They are not analyzing trends or identifying risk.

1201
00:41:33,320 --> 00:41:34,320
They are doing bookkeeping.

1202
00:41:34,320 --> 00:41:38,560
A procurement coordinator spends their week responding to supplier emails.

1203
00:41:38,560 --> 00:41:41,120
Updating spreadsheets and coordinating delivery dates.

1204
00:41:41,120 --> 00:41:44,360
They are not sourcing strategically or building vendor relationships.

1205
00:41:44,360 --> 00:41:45,920
They are managing an email inbox.

1206
00:41:45,920 --> 00:41:49,360
These are smart people doing work that does not require them to be smart.

1207
00:41:49,360 --> 00:41:52,080
When agents take over that work, the role fundamentally changes.

1208
00:41:52,080 --> 00:41:53,600
The old role was execution.

1209
00:41:53,600 --> 00:41:56,120
Do this task, do it correctly and do it fast.

1210
00:41:56,120 --> 00:41:57,720
Follow the process.

1211
00:41:57,720 --> 00:41:59,280
The new role is decision making.

1212
00:41:59,280 --> 00:42:02,360
The agent handles the execution and the human handles the exception.

1213
00:42:02,360 --> 00:42:03,680
A lead comes in.

1214
00:42:03,680 --> 00:42:07,120
The agent researches it and the human decides whether to pursue it.

1215
00:42:07,120 --> 00:42:08,920
An invoice discrepancy is flagged.

1216
00:42:08,920 --> 00:42:10,800
The agent proposes the correction.

1217
00:42:10,800 --> 00:42:13,560
And the human approves or declines based on judgment.

1218
00:42:13,560 --> 00:42:15,520
A supplier sends an urgent notification.

1219
00:42:15,520 --> 00:42:19,520
The agent escalates it and the human decides the business impact and what to do about it.

1220
00:42:19,520 --> 00:42:21,360
This is a fundamental skill shift.

1221
00:42:21,360 --> 00:42:24,160
You are not asking people to become better at their current job.

1222
00:42:24,160 --> 00:42:27,640
You are asking them to do a different job and if you do not prepare them for that shift,

1223
00:42:27,640 --> 00:42:28,800
they will resist.

1224
00:42:28,800 --> 00:42:32,600
Not because they are lazy or afraid of technology, but because you have asked them to perform

1225
00:42:32,600 --> 00:42:35,480
a completely different role without warning or training.

1226
00:42:35,480 --> 00:42:37,360
What this means for hiring is immediate.

1227
00:42:37,360 --> 00:42:40,440
You stop looking for people who are really good at operating systems.

1228
00:42:40,440 --> 00:42:44,760
You stop hiring for strong Excel skills or detail oriented data entry.

1229
00:42:44,760 --> 00:42:46,240
You start hiring for judgment.

1230
00:42:46,240 --> 00:42:49,320
You hire for the ability to reason through ambiguous situations.

1231
00:42:49,320 --> 00:42:53,720
For strategic thinking and for relationship building, these are different hiring profiles entirely.

1232
00:42:53,720 --> 00:42:57,600
A person who was an excellent reconciliation specialist because they could match numbers

1233
00:42:57,600 --> 00:43:02,960
precisely might not be the right fit for an analyst role that requires identifying patterns

1234
00:43:02,960 --> 00:43:04,760
and recommending changes.

1235
00:43:04,760 --> 00:43:08,760
Not because they are less capable but because their strengths were in a different domain.

1236
00:43:08,760 --> 00:43:11,320
The transition risk is real because it is not optional.

1237
00:43:11,320 --> 00:43:15,080
If you deploy agents without retraining your teams, you do not get smooth adoption.

1238
00:43:15,080 --> 00:43:16,440
You get resistance.

1239
00:43:16,440 --> 00:43:18,400
People feel replaced instead of augmented.

1240
00:43:18,400 --> 00:43:21,000
They lack context for what they are supposed to do now.

1241
00:43:21,000 --> 00:43:25,600
The agent is executing but nobody trains the human to manage exceptions, to make judgment

1242
00:43:25,600 --> 00:43:29,600
calls or to think strategically so they either try to do the old work that the agent is already

1243
00:43:29,600 --> 00:43:31,440
doing or they feel lost.

1244
00:43:31,440 --> 00:43:33,280
Both outcomes are bad.

1245
00:43:33,280 --> 00:43:35,080
The upside is what makes it worth doing.

1246
00:43:35,080 --> 00:43:36,840
People stop spending their intelligence on patients.

1247
00:43:36,840 --> 00:43:39,560
A reconciliation analyst can actually analyze instead of matching.

1248
00:43:39,560 --> 00:43:42,680
A sales rep can actually build relationships instead of entering data.

1249
00:43:42,680 --> 00:43:46,000
A procurement coordinator can actually source instead of answering emails.

1250
00:43:46,000 --> 00:43:47,240
This is not just time savings.

1251
00:43:47,240 --> 00:43:50,840
It is the difference between work that exhausts you and work that develops you.

1252
00:43:50,840 --> 00:43:53,480
It is the difference between a job and a career path.

1253
00:43:53,480 --> 00:43:56,680
The organizational shift that emerges is subtle but transformative.

1254
00:43:56,680 --> 00:43:57,680
Support functions.

1255
00:43:57,680 --> 00:44:00,960
Stop being cost centers and start being strategic capabilities.

1256
00:44:00,960 --> 00:44:04,240
Finance stops being the people who close the books and becomes the people who understand

1257
00:44:04,240 --> 00:44:05,640
your business financially.

1258
00:44:05,640 --> 00:44:09,720
Sales stops being the people who enter opportunities and becomes the people who understand your

1259
00:44:09,720 --> 00:44:11,640
market and your customers deeply.

1260
00:44:11,640 --> 00:44:15,760
Supply chain stops being coordination and becomes resilience and optimization.

1261
00:44:15,760 --> 00:44:17,080
These functions do not disappear.

1262
00:44:17,080 --> 00:44:18,600
They evolve upward.

1263
00:44:18,600 --> 00:44:20,080
But here is what people miss.

1264
00:44:20,080 --> 00:44:23,240
This is a change management challenge first and a technology deployment second.

1265
00:44:23,240 --> 00:44:25,840
The technology works.

1266
00:44:25,840 --> 00:44:28,720
The hard part is helping your organization through the transition.

1267
00:44:28,720 --> 00:44:31,400
Your people need to understand why their role is changing.

1268
00:44:31,400 --> 00:44:35,120
They need training for what the new role actually is and they need time to adapt if you skip

1269
00:44:35,120 --> 00:44:36,120
that work.

1270
00:44:36,120 --> 00:44:40,640
You will have adoption resistance that no amount of technology can overcome.

1271
00:44:40,640 --> 00:44:42,160
The ROI math.

1272
00:44:42,160 --> 00:44:44,080
Why this justifies the investment?

1273
00:44:44,080 --> 00:44:45,960
Let's get practical for a moment.

1274
00:44:45,960 --> 00:44:48,480
None of this matters if you can't justify the cost.

1275
00:44:48,480 --> 00:44:51,360
Implementing agents across your organization requires real investment.

1276
00:44:51,360 --> 00:44:52,680
You're buying licenses.

1277
00:44:52,680 --> 00:44:54,920
You're building agents in co-pilot studio.

1278
00:44:54,920 --> 00:44:57,520
You're dedicating time to configuration and governance.

1279
00:44:57,520 --> 00:45:00,440
You're running change management to help your teams through the transition.

1280
00:45:00,440 --> 00:45:01,840
That's real money.

1281
00:45:01,840 --> 00:45:04,080
Your CFO is asking, what's the payback?

1282
00:45:04,080 --> 00:45:05,920
When do we see this investment return?

1283
00:45:05,920 --> 00:45:09,680
The math is actually straight forward because agents affect four direct levers on your

1284
00:45:09,680 --> 00:45:10,680
P&L.

1285
00:45:10,680 --> 00:45:11,680
The first is labor cost.

1286
00:45:11,680 --> 00:45:15,280
When agent handles 40% of reconciliation work, you have options.

1287
00:45:15,280 --> 00:45:16,640
You can reduce headcount.

1288
00:45:16,640 --> 00:45:20,240
You can redeploy that capacity to higher value analytical work either way.

1289
00:45:20,240 --> 00:45:21,720
You aren't paying someone to do work.

1290
00:45:21,720 --> 00:45:23,520
The agent can do automatically.

1291
00:45:23,520 --> 00:45:28,440
If you have a five person reconciliation team and an agent handles 40% of the workload,

1292
00:45:28,440 --> 00:45:32,240
that's 1.6 full-time equivalents of capacity you've freed up.

1293
00:45:32,240 --> 00:45:37,040
At an average fully loaded cost of $120,000 per year, salary plus benefits plus overhead,

1294
00:45:37,040 --> 00:45:39,880
that's $192,000 in annual savings.

1295
00:45:39,880 --> 00:45:43,440
That's real money that flows to your bottom line or gets redeployed to work that generates

1296
00:45:43,440 --> 00:45:44,440
value.

1297
00:45:44,440 --> 00:45:45,720
The second lever is speed.

1298
00:45:45,720 --> 00:45:48,280
Fast or operational cycles compress working capital.

1299
00:45:48,280 --> 00:45:52,120
Your month and close takes five days right now and agent reduces that to one day.

1300
00:45:52,120 --> 00:45:55,400
Those four days represent working capital that's tied up because you don't have accurate

1301
00:45:55,400 --> 00:45:56,400
financials.

1302
00:45:56,400 --> 00:46:00,920
Fewer days of working capital outstanding means more cash available for operations or investment.

1303
00:46:00,920 --> 00:46:05,360
Your order to cash cycle takes 45 days because of slow customer rooting and manual invoice

1304
00:46:05,360 --> 00:46:06,600
processing.

1305
00:46:06,600 --> 00:46:08,280
You can compress that to 35 days.

1306
00:46:08,280 --> 00:46:10,320
That's 10 days of receivables freed up.

1307
00:46:10,320 --> 00:46:14,660
For a mid-market organization doing $100 million in annual revenue, 10 days of receivables at

1308
00:46:14,660 --> 00:46:18,760
roughly one month of daily revenue is a massive working capital improvement.

1309
00:46:18,760 --> 00:46:22,680
Your procurement cycle compresses because supplier communications are instant instead of

1310
00:46:22,680 --> 00:46:23,760
data laid.

1311
00:46:23,760 --> 00:46:27,840
That means you get products to market faster or deliver to customers quicker.

1312
00:46:27,840 --> 00:46:31,200
Speed translates to competitive advantage and working capital efficiency.

1313
00:46:31,200 --> 00:46:32,800
The third lever is error reduction.

1314
00:46:32,800 --> 00:46:34,440
Agents don't make typos.

1315
00:46:34,440 --> 00:46:37,920
They don't misroot a customer to the wrong department because they misread an email.

1316
00:46:37,920 --> 00:46:41,040
They don't misapply a communication that creates a production delay.

1317
00:46:41,040 --> 00:46:44,120
They don't post journal entries with reverse account codes.

1318
00:46:44,120 --> 00:46:45,280
These errors have costs.

1319
00:46:45,280 --> 00:46:47,800
A misruded customer might escalate or churn.

1320
00:46:47,800 --> 00:46:52,360
A mis-supplyer email might delay production and create customer dissatisfaction.

1321
00:46:52,360 --> 00:46:56,920
A reverse journal entry creates confusion in your financials and takes time to correct.

1322
00:46:56,920 --> 00:46:58,960
None of these costs are huge individually.

1323
00:46:58,960 --> 00:47:04,000
But across thousands of transactions monthly, error reduction compounds into measurable savings.

1324
00:47:04,000 --> 00:47:08,120
More importantly, it compounds into quality improvements that customers experience.

1325
00:47:08,120 --> 00:47:10,240
The fourth lever is compliance.

1326
00:47:10,240 --> 00:47:12,240
Agents create ordered-ready logs.

1327
00:47:12,240 --> 00:47:16,120
They apply policies consistently instead of inconsistently depending on which person

1328
00:47:16,120 --> 00:47:17,600
is handling the transaction.

1329
00:47:17,600 --> 00:47:21,360
They reduce fraud risk because there is a systematic check instead of human judgment that

1330
00:47:21,360 --> 00:47:23,000
can be influenced or overlooked.

1331
00:47:23,000 --> 00:47:25,120
These aren't dramatic year-to-year impacts.

1332
00:47:25,120 --> 00:47:27,760
But they reduce your compliance risk and audit costs.

1333
00:47:27,760 --> 00:47:32,440
In regulated industries especially, this translates to insurance savings, audit fee reduction,

1334
00:47:32,440 --> 00:47:34,920
or reduced risk of non-compliance penalties.

1335
00:47:34,920 --> 00:47:36,480
Here's where the numbers actually come from.

1336
00:47:36,480 --> 00:47:40,160
US Ventures deployed an account reconciliation agent and reduced their reconciliation cycle

1337
00:47:40,160 --> 00:47:41,560
by 80%.

1338
00:47:41,560 --> 00:47:45,160
That freed up capacity that would have cost them significant money to maintain.

1339
00:47:45,160 --> 00:47:48,840
Lifetime products cut their procurement workload by 20% through supplier communications

1340
00:47:48,840 --> 00:47:52,880
agents, which translates to headcount reduction or redeployment.

1341
00:47:52,880 --> 00:47:58,040
Microsoft's own composite business case modeling shows an average ROI of 171% over three

1342
00:47:58,040 --> 00:48:00,960
years for agentic AI deployments.

1343
00:48:00,960 --> 00:48:04,440
With organizations see payback within 12 months because the labor savings alone typically

1344
00:48:04,440 --> 00:48:07,160
covers the implementation cost in the first year.

1345
00:48:07,160 --> 00:48:11,920
But there's a hidden benefit that compounds beyond those four levers, speed compounds.

1346
00:48:11,920 --> 00:48:15,600
Faster cycles mean you can serve more customers without proportional headcount growth.

1347
00:48:15,600 --> 00:48:20,000
A sales team that can qualify leads twice as fast can pursue twice as many opportunities

1348
00:48:20,000 --> 00:48:21,400
without doubling the team.

1349
00:48:21,400 --> 00:48:25,480
A finance team that closes the books in one day instead of five has time for analysis

1350
00:48:25,480 --> 00:48:28,040
that actually moves the needle on business performance.

1351
00:48:28,040 --> 00:48:31,600
A procurement team that processes supplier communications instantly can manage more

1352
00:48:31,600 --> 00:48:34,640
vendor relationships strategically instead of operationally.

1353
00:48:34,640 --> 00:48:38,040
You're not just saving money on the current operation, you're creating capacity for growth

1354
00:48:38,040 --> 00:48:41,280
without the cost structure that growth typically requires.

1355
00:48:41,280 --> 00:48:43,240
That's where the real ROI lives.

1356
00:48:43,240 --> 00:48:46,960
The integration challenge, how agents talk to legacy systems.

1357
00:48:46,960 --> 00:48:49,400
Most of what we've discussed assumes the best case scenario.

1358
00:48:49,400 --> 00:48:53,920
Your entire operation runs on Dynamics 365, data lives in dataverse, agents query dataverse

1359
00:48:53,920 --> 00:48:56,000
and execute back into dataverse.

1360
00:48:56,000 --> 00:48:58,760
And integrated, simple, that's not your organization.

1361
00:48:58,760 --> 00:49:00,040
Your reality is more complicated.

1362
00:49:00,040 --> 00:49:03,040
You have Dynamics 365 for sales and customer service.

1363
00:49:03,040 --> 00:49:07,200
Your ERP is SAP or Oracle, something that's been running for 15 years and isn't going

1364
00:49:07,200 --> 00:49:08,520
anywhere.

1365
00:49:08,520 --> 00:49:12,800
Your warehouse management system is a specialized tool from a vendor nobody's heard of.

1366
00:49:12,800 --> 00:49:14,840
Your accounting system might be net sweet.

1367
00:49:14,840 --> 00:49:16,800
Your talent management is separate from everything else.

1368
00:49:16,800 --> 00:49:20,280
You've got islands of software and they don't talk to each other naturally.

1369
00:49:20,280 --> 00:49:24,720
People manually export data from one system, imported into another and the gaps get filled

1370
00:49:24,720 --> 00:49:26,240
by email and spreadsheets.

1371
00:49:26,240 --> 00:49:30,000
This is the heterogeneous environment where most organizations actually live.

1372
00:49:30,000 --> 00:49:32,040
The agent problem in this world is direct.

1373
00:49:32,040 --> 00:49:36,200
An agent in D365 needs to trigger an action in your legacy SAP system.

1374
00:49:36,200 --> 00:49:37,200
How does that happen?

1375
00:49:37,200 --> 00:49:40,760
The solution isn't to rip out your legacy systems and rebuild everything in Azure.

1376
00:49:40,760 --> 00:49:41,880
That's not realistic.

1377
00:49:41,880 --> 00:49:46,120
The solution is orchestration through integration platforms, power automate or logic apps become

1378
00:49:46,120 --> 00:49:47,120
the bridge layer.

1379
00:49:47,120 --> 00:49:48,200
Here's the mechanics.

1380
00:49:48,200 --> 00:49:52,040
The agent in D365 makes a decision and needs to execute an action.

1381
00:49:52,040 --> 00:49:55,920
Instead of directly calling the legacy system, the agent invokes a flow.

1382
00:49:55,920 --> 00:50:00,200
That flow understands how to translate the agent's decision into a legacy system operation.

1383
00:50:00,200 --> 00:50:02,720
The agent says update this PO delivery date.

1384
00:50:02,720 --> 00:50:06,840
The flow takes that instruction and converts it into whatever API call your SAP system

1385
00:50:06,840 --> 00:50:07,840
expects.

1386
00:50:07,840 --> 00:50:09,400
The flow handles the translation.

1387
00:50:09,400 --> 00:50:12,640
The agent doesn't need to know how your legacy ERP works.

1388
00:50:12,640 --> 00:50:16,000
This matters because it preserves your existing infrastructure investment while adding

1389
00:50:16,000 --> 00:50:17,720
agente capability on top of it.

1390
00:50:17,720 --> 00:50:23,680
A practical example, a supplier communications agent in D365 reads an incoming supplier email.

1391
00:50:23,680 --> 00:50:26,800
The email says your shipment is delayed three days.

1392
00:50:26,800 --> 00:50:31,300
The agent reasons about impact decides to update the PO delivery date and escalates to

1393
00:50:31,300 --> 00:50:33,480
procurement if the delay is critical.

1394
00:50:33,480 --> 00:50:37,160
To execute this decision, the agent needs to update your warehouse management system.

1395
00:50:37,160 --> 00:50:41,520
But your WMS is a specialized system that doesn't talk to D365 natively.

1396
00:50:41,520 --> 00:50:42,520
Here's what happens.

1397
00:50:42,520 --> 00:50:44,560
The agent calls a power automate flow.

1398
00:50:44,560 --> 00:50:47,240
The flow knows how to communicate with your WMS.

1399
00:50:47,240 --> 00:50:51,640
It extracts the relevant data from the agent's decision, formats it according to WMS API

1400
00:50:51,640 --> 00:50:53,400
requirements and makes the call.

1401
00:50:53,400 --> 00:50:54,920
The WMS gets updated.

1402
00:50:54,920 --> 00:50:58,560
The system remains orchestrated even though the agent and the target systems speak different

1403
00:50:58,560 --> 00:50:59,560
languages.

1404
00:50:59,560 --> 00:51:01,800
The governance layer is where the enterprise safeguard lives.

1405
00:51:01,800 --> 00:51:04,080
The flow doesn't just translate between systems.

1406
00:51:04,080 --> 00:51:06,040
It also enforces your business rules.

1407
00:51:06,040 --> 00:51:08,480
High value purchase order changes require approval.

1408
00:51:08,480 --> 00:51:09,840
The flow checks the PO amount.

1409
00:51:09,840 --> 00:51:12,800
If it exceeds threshold, the flow doesn't execute automatically.

1410
00:51:12,800 --> 00:51:16,120
It creates an approval request in D365 and waits for human sign-off.

1411
00:51:16,120 --> 00:51:18,760
If it's under threshold, the flow executes directly.

1412
00:51:18,760 --> 00:51:23,360
This means your approval workflows and audit requirements stay intact even as agents orchestrate

1413
00:51:23,360 --> 00:51:24,840
across heterogeneous systems.

1414
00:51:24,840 --> 00:51:26,640
The agent doesn't bypass governance.

1415
00:51:26,640 --> 00:51:28,840
The flow becomes the governance enforcement point.

1416
00:51:28,840 --> 00:51:32,120
The implication is that you don't need to make a technology bet on replacing all your

1417
00:51:32,120 --> 00:51:33,040
legacy systems.

1418
00:51:33,040 --> 00:51:36,120
You can add agente capability to D365.

1419
00:51:36,120 --> 00:51:39,720
Have those agents orchestrate across your entire technology landscape and let your legacy

1420
00:51:39,720 --> 00:51:41,440
systems run indefinitely.

1421
00:51:41,440 --> 00:51:45,520
At the same time, as those systems age out, you might consolidate more functionality into

1422
00:51:45,520 --> 00:51:46,840
D365.

1423
00:51:46,840 --> 00:51:49,840
But you're not forced into that decision because you deployed agents.

1424
00:51:49,840 --> 00:51:53,480
You've decoupled the agent capability from the legacy system rationalization decision

1425
00:51:53,480 --> 00:51:55,000
that two separate problems.

1426
00:51:55,000 --> 00:51:59,800
You can solve the agent problem without solving the legacy system problem simultaneously.

1427
00:51:59,800 --> 00:52:04,640
This is why organizations with complex heterogeneous IT environments can actually move faster on

1428
00:52:04,640 --> 00:52:07,080
agents than organizations with simplest stacks.

1429
00:52:07,080 --> 00:52:09,000
You've already built the integration discipline.

1430
00:52:09,000 --> 00:52:11,640
You already have API conversations happening across systems.

1431
00:52:11,640 --> 00:52:14,480
You're just adding agents to that orchestration layer.

1432
00:52:14,480 --> 00:52:15,480
It's not a stretch.

1433
00:52:15,480 --> 00:52:16,480
It's an evolution.

1434
00:52:16,480 --> 00:52:17,800
The data quality prerequisite.

1435
00:52:17,800 --> 00:52:19,520
You can't automate garbage.

1436
00:52:19,520 --> 00:52:22,960
Here is the conversation happening in too many offices right after the agents are built

1437
00:52:22,960 --> 00:52:24,320
and ready to go.

1438
00:52:24,320 --> 00:52:27,680
Someone in IT says the reconciliation agent is finished and they are ready to go live in

1439
00:52:27,680 --> 00:52:28,680
production.

1440
00:52:28,680 --> 00:52:30,120
Then someone from finance speaks up.

1441
00:52:30,120 --> 00:52:33,840
They point out that the supplier master is a mess because 30 different vendors have multiple

1442
00:52:33,840 --> 00:52:35,440
names in the system.

1443
00:52:35,440 --> 00:52:37,560
Sometimes the company is listed as ACME Corp.

1444
00:52:37,560 --> 00:52:41,120
Other times it is just ACME or maybe ACME Corporation.

1445
00:52:41,120 --> 00:52:44,760
The agent is going to see an invoice from one and think it is a totally different vendor

1446
00:52:44,760 --> 00:52:45,760
than the other.

1447
00:52:45,760 --> 00:52:49,040
It will end up flagging errors on invoices that should have been a perfect match.

1448
00:52:49,040 --> 00:52:50,880
That is the moment where things get real.

1449
00:52:50,880 --> 00:52:53,240
You can build the most advanced agent in the world.

1450
00:52:53,240 --> 00:52:57,040
But if the data underneath it is broken, the agent's decisions will be broken too.

1451
00:52:57,040 --> 00:52:59,040
You aren't actually automating excellence.

1452
00:52:59,040 --> 00:53:02,120
You are just automating mistakes at a much higher speed in scale.

1453
00:53:02,120 --> 00:53:07,080
This is the data quality problem and you have to solve it before a single agent is deployed.

1454
00:53:07,080 --> 00:53:11,600
Most companies live with dirty data because it is tolerable when humans are the ones doing

1455
00:53:11,600 --> 00:53:12,600
the work.

1456
00:53:12,600 --> 00:53:16,240
A person reconciling an invoice can see that ACME Corp and ACME are the same place because

1457
00:53:16,240 --> 00:53:18,120
they have context and common sense.

1458
00:53:18,120 --> 00:53:22,640
If a lead record is missing a company name, a human does a quick search to figure it out.

1459
00:53:22,640 --> 00:53:27,440
When customer info is incomplete, people dig into related files and infer what is missing.

1460
00:53:27,440 --> 00:53:30,680
Humans are flexible enough to work around these gaps, but agents are not.

1461
00:53:30,680 --> 00:53:32,000
An agent does not have judgment.

1462
00:53:32,000 --> 00:53:33,320
It only has data and logic.

1463
00:53:33,320 --> 00:53:36,800
If a customer record is missing a phone number, the agent won't try to guess it from other

1464
00:53:36,800 --> 00:53:37,800
files.

1465
00:53:37,800 --> 00:53:39,560
It just marks the record as incomplete.

1466
00:53:39,560 --> 00:53:43,000
If a lead is missing an industry tag, the agent cannot score it against your ideal patterns

1467
00:53:43,000 --> 00:53:45,400
because the information simply isn't there.

1468
00:53:45,400 --> 00:53:49,400
When a purchase order lacks a cost center, the agent cannot root it for approval because

1469
00:53:49,400 --> 00:53:52,040
it has no idea where the money is supposed to come from.

1470
00:53:52,040 --> 00:53:56,480
The messy data your team tolerates today will become a total blocker for your agents tomorrow.

1471
00:53:56,480 --> 00:54:00,800
You cannot deploy these tools at scale on a foundation of duplicates and inconsistencies.

1472
00:54:00,800 --> 00:54:05,320
The agent will either make a bad call or escalate every single task back to a human because

1473
00:54:05,320 --> 00:54:07,000
it can't find the context it needs.

1474
00:54:07,000 --> 00:54:09,760
Either way, you lose the automation benefit you are paying for.

1475
00:54:09,760 --> 00:54:12,240
So before you launch, you have to fix the foundation.

1476
00:54:12,240 --> 00:54:13,600
This starts with data cleansing.

1477
00:54:13,600 --> 00:54:16,640
You need to find the duplicates and merge them into one clean record.

1478
00:54:16,640 --> 00:54:21,120
You have to standardize your naming conventions if there are 30 ways to spell a vendor name.

1479
00:54:21,120 --> 00:54:22,920
You pick one and apply it everywhere.

1480
00:54:22,920 --> 00:54:26,960
If contact records are missing numbers, you either fill them in or mark them in valid,

1481
00:54:26,960 --> 00:54:28,440
so the agent knows to skip them.

1482
00:54:28,440 --> 00:54:29,920
This isn't high-level strategy.

1483
00:54:29,920 --> 00:54:33,440
It is tedious manual work where you go through your records and fix them one by one.

1484
00:54:33,440 --> 00:54:35,160
Next, you need master data governance.

1485
00:54:35,160 --> 00:54:38,040
You cannot just clean the data once and hope it stays that way.

1486
00:54:38,040 --> 00:54:41,200
You need strict rules for how new data gets created from now on.

1487
00:54:41,200 --> 00:54:44,800
When someone adds a new vendor, you have to decide which fields are mandatory and what

1488
00:54:44,800 --> 00:54:46,400
naming style they must use.

1489
00:54:46,400 --> 00:54:49,280
You also need to know who approves that vendor before it hits the system.

1490
00:54:49,280 --> 00:54:53,080
Without these rules, you will fix the mess today only to see it return in six months because

1491
00:54:53,080 --> 00:54:55,680
people are still entering data without a standard.

1492
00:54:55,680 --> 00:54:59,320
Finally, you need consistency in what your data actually means.

1493
00:54:59,320 --> 00:55:03,480
When you label a leader's qualified, that word needs to mean the exact same thing across

1494
00:55:03,480 --> 00:55:04,520
the whole company.

1495
00:55:04,520 --> 00:55:09,360
If urgent means something different to every field team, the agent's reasoning will be all

1496
00:55:09,360 --> 00:55:10,760
over the place.

1497
00:55:10,760 --> 00:55:14,600
Agents think based on these values, so the definitions have to be identical everywhere.

1498
00:55:14,600 --> 00:55:16,400
This work takes time.

1499
00:55:16,400 --> 00:55:19,760
Realistically, you are looking at four to eight weeks of prep before you can deploy at

1500
00:55:19,760 --> 00:55:20,760
scale.

1501
00:55:20,760 --> 00:55:23,200
You cannot skip this just because you want to move fast.

1502
00:55:23,200 --> 00:55:26,360
If you try, you will launch on garbage data and get bad results.

1503
00:55:26,360 --> 00:55:29,160
Then you will have to stop everything and fix the data anyway.

1504
00:55:29,160 --> 00:55:33,400
You will end up wasting more time backtracking than you ever saved by rushing the start.

1505
00:55:33,400 --> 00:55:35,960
But once the data is cleaned, the whole dynamic shifts.

1506
00:55:35,960 --> 00:55:39,560
Accurate information stops being a chore and starts being a real asset.

1507
00:55:39,560 --> 00:55:43,480
Your agents become much more valuable because they are finally working with data they can

1508
00:55:43,480 --> 00:55:44,880
actually trust.

1509
00:55:44,880 --> 00:55:46,760
The change management reality.

1510
00:55:46,760 --> 00:55:48,160
Agents disrupt workflows.

1511
00:55:48,160 --> 00:55:51,440
The most important conversation in your company happens before the agents go live, not

1512
00:55:51,440 --> 00:55:52,440
after.

1513
00:55:52,440 --> 00:55:55,720
It is the talk between an operations leader and their team and it almost always starts

1514
00:55:55,720 --> 00:55:57,200
with a lot of anxiety.

1515
00:55:57,200 --> 00:56:01,720
The leader announces they are deploying an agent to handle 40% of the reconciliation work.

1516
00:56:01,720 --> 00:56:03,600
The reaction is usually predictable.

1517
00:56:03,600 --> 00:56:06,200
People hear that and immediately think their job is going away.

1518
00:56:06,200 --> 00:56:08,200
They feel like they are being replaced.

1519
00:56:08,200 --> 00:56:10,760
And because nobody likes that feeling, they start to resist.

1520
00:56:10,760 --> 00:56:12,480
They aren't being difficult or irrational.

1521
00:56:12,480 --> 00:56:16,240
They are reacting to something that sounds like a direct threat to their livelihood because

1522
00:56:16,240 --> 00:56:18,760
nobody explained what the change actually means.

1523
00:56:18,760 --> 00:56:20,880
This resistance pattern is a very real hurdle.

1524
00:56:20,880 --> 00:56:25,040
Teams worry that if an agent can do their tasks, the company won't need them in six months.

1525
00:56:25,040 --> 00:56:27,800
They start looking at their options and updating their resumes.

1526
00:56:27,800 --> 00:56:31,000
You won't lose your best people because they hate the new technology.

1527
00:56:31,000 --> 00:56:34,720
You will lose them because you didn't explain how their roles are actually changing.

1528
00:56:34,720 --> 00:56:36,760
Here is the reality you have to communicate.

1529
00:56:36,760 --> 00:56:39,800
Agents eliminate the boring work, not the expertise.

1530
00:56:39,800 --> 00:56:42,000
An agent does not replace a finance analyst.

1531
00:56:42,000 --> 00:56:44,600
It replaces the data entry part of their afternoon.

1532
00:56:44,600 --> 00:56:48,320
It takes over the repetitive matching and the simple patterns that don't require any real

1533
00:56:48,320 --> 00:56:49,320
thought.

1534
00:56:49,320 --> 00:56:52,280
The agent handles the routine while your analyst handles the exceptions.

1535
00:56:52,280 --> 00:56:56,160
They step in when something doesn't fit the mold and a human needs to figure out why.

1536
00:56:56,160 --> 00:56:57,320
That isn't replacing someone.

1537
00:56:57,320 --> 00:56:58,560
It is amplifying them.

1538
00:56:58,560 --> 00:57:02,640
You are taking a smart person and moving them away from chores so they can do work that

1539
00:57:02,640 --> 00:57:03,640
actually matters.

1540
00:57:03,640 --> 00:57:05,640
That shift in framing changes the whole mood.

1541
00:57:05,640 --> 00:57:08,920
Instead of telling someone their job is disappearing, you are telling them their job is

1542
00:57:08,920 --> 00:57:09,920
getting better.

1543
00:57:09,920 --> 00:57:13,280
Your analyst no longer has to spend three days a week matching invoices.

1544
00:57:13,280 --> 00:57:17,040
Now they can spend that time looking at why certain suppliers have so many errors.

1545
00:57:17,040 --> 00:57:19,720
They can find ways to fix the process at the source.

1546
00:57:19,720 --> 00:57:23,920
That is better work that uses their brain instead of just their patience.

1547
00:57:23,920 --> 00:57:26,560
Your communication strategy has to start there.

1548
00:57:26,560 --> 00:57:29,880
Start making it about the software and start making it about the person.

1549
00:57:29,880 --> 00:57:33,640
Tell them they will have two fewer hours of data entry every day.

1550
00:57:33,640 --> 00:57:36,960
Explain that this gives them time for the high level analysis their manager has been asking

1551
00:57:36,960 --> 00:57:37,960
for.

1552
00:57:37,960 --> 00:57:42,200
When you frame it as less time on repetitive tasks and more time thinking about the business,

1553
00:57:42,200 --> 00:57:43,520
the message actually lands.

1554
00:57:43,520 --> 00:57:45,880
The conversation moves from a threat to an opportunity.

1555
00:57:45,880 --> 00:57:48,040
Once the message is clear, you run a pilot.

1556
00:57:48,040 --> 00:57:51,480
Do not try to change the entire company at the same time because that is a recipe for

1557
00:57:51,480 --> 00:57:52,480
chaos.

1558
00:57:52,480 --> 00:57:56,360
Pick one high volume process that is causing a lot of pain and run the agent there first.

1559
00:57:56,360 --> 00:57:57,840
Consider what actually happens on the ground.

1560
00:57:57,840 --> 00:58:02,600
If reconciliation time drops by 70% and the team handles the transition well, you finally

1561
00:58:02,600 --> 00:58:03,600
have proof.

1562
00:58:03,600 --> 00:58:06,360
You can show everyone exactly what changed and the impact it had.

1563
00:58:06,360 --> 00:58:08,520
That evidence makes the next rollout much easier.

1564
00:58:08,520 --> 00:58:10,800
The training you provide needs to be very specific.

1565
00:58:10,800 --> 00:58:14,040
You aren't teaching people how to do the work the agent is already doing.

1566
00:58:14,040 --> 00:58:16,200
You are teaching them how to manage the agent.

1567
00:58:16,200 --> 00:58:19,760
They need to know how to review its decisions and how to tell if the performance is starting

1568
00:58:19,760 --> 00:58:20,760
to slip.

1569
00:58:20,760 --> 00:58:23,400
They need to understand the new workflow for handling exceptions.

1570
00:58:23,400 --> 00:58:25,600
This is a different skill set than manual reconciliation.

1571
00:58:25,600 --> 00:58:27,280
It is about oversight.

1572
00:58:27,280 --> 00:58:30,640
People usually prefer this kind of training because it feels like they are gaining a new

1573
00:58:30,640 --> 00:58:33,080
valuable skill rather than being pushed out.

1574
00:58:33,080 --> 00:58:35,000
The timeline for this is longer than you think.

1575
00:58:35,000 --> 00:58:38,040
Expect it to take 6-12 weeks from the launch date to reach full adoption.

1576
00:58:38,040 --> 00:58:41,960
You can't just flip a switch and expect everyone to be comfortable supervising a machine

1577
00:58:41,960 --> 00:58:43,120
overnight.

1578
00:58:43,120 --> 00:58:46,360
They need time to see that the agent actually works and that it isn't just a passing

1579
00:58:46,360 --> 00:58:47,360
fare.

1580
00:58:47,360 --> 00:58:49,160
They need to build confidence in the new way of working.

1581
00:58:49,160 --> 00:58:51,120
During those weeks you have to keep talking.

1582
00:58:51,120 --> 00:58:54,560
Show the progress, celebrate the small wins and answer concerns directly instead of

1583
00:58:54,560 --> 00:58:55,560
ignoring them.

1584
00:58:55,560 --> 00:58:57,800
This is where these projects actually succeed or fail.

1585
00:58:57,800 --> 00:58:59,560
It isn't about the code or the servers.

1586
00:58:59,560 --> 00:59:03,960
It is about whether you actually helped your people move through the change.

1587
00:59:03,960 --> 00:59:05,800
The security and compliance story.

1588
00:59:05,800 --> 00:59:07,400
Agents don't bypass governance.

1589
00:59:07,400 --> 00:59:11,800
CFOs and compliance officers usually have one big question when they look at agents.

1590
00:59:11,800 --> 00:59:13,640
Are we creating a massive compliance risk?

1591
00:59:13,640 --> 00:59:14,640
The answer is no.

1592
00:59:14,640 --> 00:59:16,240
But only if you set them up correctly.

1593
00:59:16,240 --> 00:59:17,960
The core rule is simple.

1594
00:59:17,960 --> 00:59:21,800
Agents operate within the exact same security boundaries as your employees.

1595
00:59:21,800 --> 00:59:24,840
They aren't a separate system that gets to skip your governance.

1596
00:59:24,840 --> 00:59:29,400
If an agent is assigned to your finance team, it can only see and touch data that a finance

1597
00:59:29,400 --> 00:59:30,800
person can see and touch.

1598
00:59:30,800 --> 00:59:34,960
A sales agent can't suddenly look at customer payment histories because sales users don't

1599
00:59:34,960 --> 00:59:36,200
have that permission.

1600
00:59:36,200 --> 00:59:39,640
Row-level security applies to agents the same way it applies to your staff.

1601
00:59:39,640 --> 00:59:43,960
If a manager in Denver is restricted to seeing Denver opportunities, an agent running under

1602
00:59:43,960 --> 00:59:45,960
that role is restricted to that same view.

1603
00:59:45,960 --> 00:59:48,880
This isn't a special agent mode you have to configure.

1604
00:59:48,880 --> 00:59:52,040
It's the same security model your company is already running.

1605
00:59:52,040 --> 00:59:53,920
But here's the problem most people expect.

1606
00:59:53,920 --> 00:59:56,400
They think they'll have to build new security structures.

1607
00:59:56,400 --> 00:59:58,400
In reality, you're just using what you already have.

1608
00:59:58,400 --> 01:00:02,720
Your IT team has already spent years defining what different roles can see and do and agents

1609
01:00:02,720 --> 01:00:04,400
simply inherit that work.

1610
01:00:04,400 --> 01:00:08,280
This means adding agents doesn't create a new administrative nightmare for your security

1611
01:00:08,280 --> 01:00:09,280
team.

1612
01:00:09,280 --> 01:00:11,960
You aren't managing two different sets of rules for humans and machines.

1613
01:00:11,960 --> 01:00:14,080
There is one set of rules and everyone follows them.

1614
01:00:14,080 --> 01:00:17,520
The audit trail is where compliance actually becomes a strength.

1615
01:00:17,520 --> 01:00:19,640
Every single decision an agent makes is logged.

1616
01:00:19,640 --> 01:00:22,280
Not just the final result but the context behind it.

1617
01:00:22,280 --> 01:00:26,040
When a reconciliation agent suggests a journal entry, the log shows every step.

1618
01:00:26,040 --> 01:00:30,480
It shows the sub-ledger at checked, the g-l-it-compared, the tolerance rules it applied and the exact

1619
01:00:30,480 --> 01:00:31,880
discrepancy it found.

1620
01:00:31,880 --> 01:00:36,800
If an auditor asks why a specific entry was created, you have the entire decision chain ready.

1621
01:00:36,800 --> 01:00:38,600
You don't have to piece it together from memory.

1622
01:00:38,600 --> 01:00:42,440
You don't have to ask an analyst if they remember why they posted something six months ago.

1623
01:00:42,440 --> 01:00:43,640
The system has the answer.

1624
01:00:43,640 --> 01:00:45,600
This is actually a compliance improvement.

1625
01:00:45,600 --> 01:00:47,440
Human decisions are often a black box.

1626
01:00:47,440 --> 01:00:50,480
Someone matches an invoice, hits a proof and moves on.

1627
01:00:50,480 --> 01:00:54,120
When auditors ask why later that person has to try and remember their logic, maybe they

1628
01:00:54,120 --> 01:00:55,560
do and maybe they don't.

1629
01:00:55,560 --> 01:00:58,760
But an agent's decisions are always documented and always reproducible.

1630
01:00:58,760 --> 01:01:03,000
If you need to prove that a policy was followed, the audit trail is your proof.

1631
01:01:03,000 --> 01:01:06,040
Approval workflows are the guardrails that keep agents from going rogue.

1632
01:01:06,040 --> 01:01:09,640
A procurement agent can automatically update a routine delivery date because that's a low

1633
01:01:09,640 --> 01:01:10,840
risk task.

1634
01:01:10,840 --> 01:01:15,200
But that same agent can't authorize a new vendor or change a contract on its own.

1635
01:01:15,200 --> 01:01:17,040
Those actions require a human signature.

1636
01:01:17,040 --> 01:01:19,200
The workflow is where the human gate sits.

1637
01:01:19,200 --> 01:01:22,840
The agent does the heavy lifting by gathering info and proposing an action, but it doesn't

1638
01:01:22,840 --> 01:01:25,400
pull the trigger on anything that exceeds your thresholds.

1639
01:01:25,400 --> 01:01:28,920
It escalates the case to the right person with all the context they need.

1640
01:01:28,920 --> 01:01:31,520
That person makes the call and only then does the action happen.

1641
01:01:31,520 --> 01:01:34,840
The human stays in the loop where it matters.

1642
01:01:34,840 --> 01:01:38,800
Data loss prevention policies also apply to agents exactly like they do to users.

1643
01:01:38,800 --> 01:01:42,680
An agent cannot send sensitive data through an unapproved channel because your DLP rules

1644
01:01:42,680 --> 01:01:43,680
won't allow it.

1645
01:01:43,680 --> 01:01:47,680
If you have a policy that stops customer data from leaving your internal systems, that rule

1646
01:01:47,680 --> 01:01:50,640
holds whether a human or an agent tries to move it.

1647
01:01:50,640 --> 01:01:54,240
You aren't giving agents special keys to bypass your data protection.

1648
01:01:54,240 --> 01:01:58,480
In regulated industries like finance or healthcare, agents actually make things safer because

1649
01:01:58,480 --> 01:01:59,640
they are consistent.

1650
01:01:59,640 --> 01:02:03,400
A human compliance analyst might get it right 98% of the time.

1651
01:02:03,400 --> 01:02:06,880
They get tired, they make judgment calls, or they just miss a detail.

1652
01:02:06,880 --> 01:02:10,080
An agent applies the rule perfectly every single time.

1653
01:02:10,080 --> 01:02:13,680
However, that isn't just a neutral change, it's a massive upgrade.

1654
01:02:13,680 --> 01:02:17,480
So what's actually happening is that the risk isn't the agent, it's the deployment.

1655
01:02:17,480 --> 01:02:21,880
If you ignore the governance tools, skip the approval steps and just let agents run wild

1656
01:02:21,880 --> 01:02:23,840
then yes you've created a risk.

1657
01:02:23,840 --> 01:02:26,640
But that is a choice, not a flaw in the technology.

1658
01:02:26,640 --> 01:02:30,520
Agents with governance are significantly safer than humans without it.

1659
01:02:30,520 --> 01:02:32,160
The competitive angle.

1660
01:02:32,160 --> 01:02:34,160
Your competitors are already moving.

1661
01:02:34,160 --> 01:02:36,800
This isn't a conversation about what might happen in the future.

1662
01:02:36,800 --> 01:02:42,280
This is about right now. By the end of 2026, 40% of enterprise apps will have task-specific

1663
01:02:42,280 --> 01:02:43,760
agents built right into them.

1664
01:02:43,760 --> 01:02:45,400
That isn't a guess from a software salesman.

1665
01:02:45,400 --> 01:02:49,240
That is Godness assessment based on what companies are actually building and buying today.

1666
01:02:49,240 --> 01:02:51,800
We aren't talking about science projects in a lab.

1667
01:02:51,800 --> 01:02:55,720
We are talking about real agents running real business processes in the real world.

1668
01:02:55,720 --> 01:02:58,600
That number matters because it defines the world you're competing in.

1669
01:02:58,600 --> 01:03:00,360
Think about what that looks like on the ground.

1670
01:03:00,360 --> 01:03:02,600
Your competitors are splitting into two groups.

1671
01:03:02,600 --> 01:03:04,760
One group is using agents and the other isn't.

1672
01:03:04,760 --> 01:03:07,120
The ones using agents are simply moving faster.

1673
01:03:07,120 --> 01:03:11,000
Their sales teams qualify leads in half the time and their finance teams close the books

1674
01:03:11,000 --> 01:03:12,240
in a third of the time.

1675
01:03:12,240 --> 01:03:16,400
Their procurement teams respond to suppliers instantly instead of waiting 24 hours for an email

1676
01:03:16,400 --> 01:03:17,400
reply.

1677
01:03:17,400 --> 01:03:18,840
That speed advantage compounds.

1678
01:03:18,840 --> 01:03:22,880
A sales team that qualifies leads twice as fast can chase twice as many deals with the same

1679
01:03:22,880 --> 01:03:23,960
number of people.

1680
01:03:23,960 --> 01:03:28,320
A finance team that closes faster gives the CEO real numbers days earlier.

1681
01:03:28,320 --> 01:03:32,400
A procurement team that handles messages instantly can manage more vendors without getting

1682
01:03:32,400 --> 01:03:33,400
overwhelmed.

1683
01:03:33,400 --> 01:03:34,400
That's the gap.

1684
01:03:34,400 --> 01:03:36,840
It's about which software has more buttons.

1685
01:03:36,840 --> 01:03:41,680
It's about operational speed and speed turns into lower costs and better customer experiences.

1686
01:03:41,680 --> 01:03:43,840
For your customers, your speed feels like better service.

1687
01:03:43,840 --> 01:03:47,240
When a support agent routes a customer to the right person immediately, it feels like

1688
01:03:47,240 --> 01:03:49,200
you actually care about their time.

1689
01:03:49,200 --> 01:03:52,880
When a problem gets fixed on the first call instead of being escalated three times, that's

1690
01:03:52,880 --> 01:03:54,280
how you win loyalty.

1691
01:03:54,280 --> 01:03:57,880
Customers don't see your back office tech but they definitely feel the speed and accuracy

1692
01:03:57,880 --> 01:03:59,360
of how you treat them.

1693
01:03:59,360 --> 01:04:02,400
For your business, that efficiency shows up in your margins.

1694
01:04:02,400 --> 01:04:06,480
You get a lower cost per transaction and better working capital because your cycles are shorter.

1695
01:04:06,480 --> 01:04:08,960
You get more work done without having to hire more people.

1696
01:04:08,960 --> 01:04:13,280
A competitor using agents can serve more customers at a lower cost, which means they can

1697
01:04:13,280 --> 01:04:15,960
undercut your price or spend more on innovation.

1698
01:04:15,960 --> 01:04:17,080
These aren't small wins.

1699
01:04:17,080 --> 01:04:19,200
Over time, they determine who wins the market.

1700
01:04:19,200 --> 01:04:22,080
And there is a talent side to this that most people miss.

1701
01:04:22,080 --> 01:04:24,760
Companies that automate the boring stuff attract better people.

1702
01:04:24,760 --> 01:04:28,920
A standard procurement job is often just managing spreadsheets and emails, which isn't exactly

1703
01:04:28,920 --> 01:04:29,920
exciting work.

1704
01:04:29,920 --> 01:04:33,040
They don't have to work with people who just execute, but a job where an agent handles the routine

1705
01:04:33,040 --> 01:04:36,360
stuff so the human can focus on strategy and relationships.

1706
01:04:36,360 --> 01:04:37,840
That's a job people actually want.

1707
01:04:37,840 --> 01:04:39,880
It attracts thinkers, not just doers.

1708
01:04:39,880 --> 01:04:42,760
Your competitors who move fast will have this advantage.

1709
01:04:42,760 --> 01:04:46,800
They will hire the talent you can't get because your roles are still defined by drudgery

1710
01:04:46,800 --> 01:04:47,800
instead of judgment.

1711
01:04:47,800 --> 01:04:49,720
The shift is that this is becoming the baseline.

1712
01:04:49,720 --> 01:04:51,680
It's no longer a nice to have feature.

1713
01:04:51,680 --> 01:04:56,280
In five years, nobody will ask if you have agents, they'll ask how good your agent workflows

1714
01:04:56,280 --> 01:04:57,280
are.

1715
01:04:57,280 --> 01:05:00,120
The results start now will be three years ahead of the ones that wait.

1716
01:05:00,120 --> 01:05:02,440
And in your industry, three years is a lifetime.

1717
01:05:02,440 --> 01:05:03,600
The reality is simple.

1718
01:05:03,600 --> 01:05:05,720
Your company will eventually use agents.

1719
01:05:05,720 --> 01:05:10,360
The only real decision is whether you start now or spend the next two years playing catch-up

1720
01:05:10,360 --> 01:05:13,680
while your competitors move at a speed you can't match.

1721
01:05:13,680 --> 01:05:14,680
The road map.

1722
01:05:14,680 --> 01:05:15,960
How to start moving?

1723
01:05:15,960 --> 01:05:17,480
You understand the landscape now.

1724
01:05:17,480 --> 01:05:20,520
You know why agents matter, how they work and how to govern them.

1725
01:05:20,520 --> 01:05:23,640
But understanding a concept and actually starting are two different things.

1726
01:05:23,640 --> 01:05:26,960
Most organizations get stuck because they try to plan for everything at once.

1727
01:05:26,960 --> 01:05:27,960
Where it breaks.

1728
01:05:27,960 --> 01:05:29,160
You don't need a perfect plan.

1729
01:05:29,160 --> 01:05:33,320
You need a road map that moves you from where you are now to where you need to be.

1730
01:05:33,320 --> 01:05:35,240
This road map breaks into four phases.

1731
01:05:35,240 --> 01:05:36,960
These aren't strict dates on a calendar.

1732
01:05:36,960 --> 01:05:38,120
They are decision points.

1733
01:05:38,120 --> 01:05:39,400
And the model is simple.

1734
01:05:39,400 --> 01:05:42,880
Early phases pay for the later ones through the ROI you generate.

1735
01:05:42,880 --> 01:05:44,480
Phase one covers your first three months.

1736
01:05:44,480 --> 01:05:45,480
Pick one process.

1737
01:05:45,480 --> 01:05:48,560
Not five, not a massive pilot that touches every department, just one.

1738
01:05:48,560 --> 01:05:50,720
It needs to be high volume and high impact.

1739
01:05:50,720 --> 01:05:54,560
It should be a process that is causing real operational pain for your team right now.

1740
01:05:54,560 --> 01:05:57,520
If you are in finance, look at reconciliation if you are in sales.

1741
01:05:57,520 --> 01:05:58,640
Look at lead qualification.

1742
01:05:58,640 --> 01:06:01,200
If you are in procurement, look at supplier communications.

1743
01:06:01,200 --> 01:06:03,520
You need a process where you already know the numbers.

1744
01:06:03,520 --> 01:06:06,840
You know how much time it takes, how much it costs and how often errors happen.

1745
01:06:06,840 --> 01:06:07,840
Write those numbers down.

1746
01:06:07,840 --> 01:06:09,240
This is your before snapshot.

1747
01:06:09,240 --> 01:06:10,240
Then.

1748
01:06:10,240 --> 01:06:11,240
Deploy your agent.

1749
01:06:11,240 --> 01:06:12,240
Keep the scope tight.

1750
01:06:12,240 --> 01:06:15,280
You aren't trying to automate the entire workflow from start to finish.

1751
01:06:15,280 --> 01:06:18,040
You are only automating the parts that agents do best.

1752
01:06:18,040 --> 01:06:19,160
The patent matching.

1753
01:06:19,160 --> 01:06:20,360
The data enrichment.

1754
01:06:20,360 --> 01:06:21,800
The decision framework.

1755
01:06:21,800 --> 01:06:25,680
Let that agent run for eight weeks and document every single thing that happens.

1756
01:06:25,680 --> 01:06:27,920
Phase two happens between months, four and six.

1757
01:06:27,920 --> 01:06:29,440
This is where you prove the impact.

1758
01:06:29,440 --> 01:06:31,200
Go back to those numbers you wrote down.

1759
01:06:31,200 --> 01:06:32,640
Did the cycle time actually drop?

1760
01:06:32,640 --> 01:06:34,640
By how much?

1761
01:06:34,640 --> 01:06:36,000
What does the error rate look like now?

1762
01:06:36,000 --> 01:06:38,680
You need to know if the agent is making good decisions.

1763
01:06:38,680 --> 01:06:41,520
Or if it's just escalating every little thing to a human.

1764
01:06:41,520 --> 01:06:45,080
This is the moment you learn if your pilot worked or if you need to pivot.

1765
01:06:45,080 --> 01:06:47,760
Most organizations see a clear improvement here.

1766
01:06:47,760 --> 01:06:48,760
Reconciliation gets faster.

1767
01:06:48,760 --> 01:06:51,000
Lead qualification gets more accurate.

1768
01:06:51,000 --> 01:06:53,680
Reconciliation shifts from slow daily batches to real-time action.

1769
01:06:53,680 --> 01:06:55,800
Once you see that win, you expand.

1770
01:06:55,800 --> 01:06:57,960
But don't jump to another department yet.

1771
01:06:57,960 --> 01:07:01,040
Move to the adjacent processes that look similar to your first one.

1772
01:07:01,040 --> 01:07:05,400
If you started with reconciliation, move to accounts payable, reconciliation next.

1773
01:07:05,400 --> 01:07:08,480
You're using the same agent framework and the same governance.

1774
01:07:08,480 --> 01:07:09,840
Only the data changes.

1775
01:07:09,840 --> 01:07:11,480
You aren't learning a new system.

1776
01:07:11,480 --> 01:07:13,400
You're just applying what already works.

1777
01:07:13,400 --> 01:07:15,800
Phase three spans months seven through twelve.

1778
01:07:15,800 --> 01:07:17,480
This is the shift.

1779
01:07:17,480 --> 01:07:19,200
Governance is no longer an afterthought.

1780
01:07:19,200 --> 01:07:20,200
It becomes formal.

1781
01:07:20,200 --> 01:07:21,640
You aren't in pilot mode anymore.

1782
01:07:21,640 --> 01:07:22,960
You are in production at scale.

1783
01:07:22,960 --> 01:07:24,240
Agent 365 is live.

1784
01:07:24,240 --> 01:07:27,640
You've built an agent registry so you can see exactly what is running, who owns it, and

1785
01:07:27,640 --> 01:07:28,640
how it's performing.

1786
01:07:28,640 --> 01:07:31,400
You've set up the approval workflows for high-risk actions.

1787
01:07:31,400 --> 01:07:34,400
You've configured the audit logs and assigned security roles.

1788
01:07:34,400 --> 01:07:36,280
But more importantly, you've trained your people.

1789
01:07:36,280 --> 01:07:37,880
They aren't doing the manual work anymore.

1790
01:07:37,880 --> 01:07:39,400
They are supervising the agents.

1791
01:07:39,400 --> 01:07:41,560
This is where you measure ROI formally.

1792
01:07:41,560 --> 01:07:44,280
The financial case for the next step should be obvious by now.

1793
01:07:44,280 --> 01:07:47,920
If your reconciliation agent saved 200 hours a month, the math is done.

1794
01:07:47,920 --> 01:07:50,160
That ROI pays for everything that comes next.

1795
01:07:50,160 --> 01:07:51,840
Phase four is year two and beyond.

1796
01:07:51,840 --> 01:07:53,840
Now, you're thinking about orchestration.

1797
01:07:53,840 --> 01:07:57,720
Sales agents are talking to finance agents to handle the quote to cash cycle.

1798
01:07:57,720 --> 01:08:01,160
Procurement agents are coordinating with supply chain agents to manage orders.

1799
01:08:01,160 --> 01:08:03,400
You aren't just building individual agents anymore.

1800
01:08:03,400 --> 01:08:05,880
You're building an ecosystem.

1801
01:08:05,880 --> 01:08:09,880
Specialized agents are handing off tasks to each other and collaborating on complex work.

1802
01:08:09,880 --> 01:08:11,560
Full maturity takes about two years.

1803
01:08:11,560 --> 01:08:13,160
But that isn't two years of heavy lifting.

1804
01:08:13,160 --> 01:08:16,640
It's two years of slowly expanding your scope once you've proved the model works.

1805
01:08:16,640 --> 01:08:19,000
The timeline for ROI is straightforward.

1806
01:08:19,000 --> 01:08:21,280
It's a big project to see a return within your first 12 months.

1807
01:08:21,280 --> 01:08:22,280
Full payback.

1808
01:08:22,280 --> 01:08:25,440
Where the agents have saved enough to cover their entire setup cost.

1809
01:08:25,440 --> 01:08:26,720
Usually happens in year one.

1810
01:08:26,720 --> 01:08:29,200
The financial case is positive from month four onward.

1811
01:08:29,200 --> 01:08:30,560
That's why expanding makes sense.

1812
01:08:30,560 --> 01:08:31,720
But here's the problem.

1813
01:08:31,720 --> 01:08:33,680
Most people wait.

1814
01:08:33,680 --> 01:08:34,960
They wait for perfect data.

1815
01:08:34,960 --> 01:08:36,400
They wait for perfect governance.

1816
01:08:36,400 --> 01:08:37,400
Don't do that.

1817
01:08:37,400 --> 01:08:39,800
Start with the process that hurts the most right now.

1818
01:08:39,800 --> 01:08:41,160
Fix the data as you go.

1819
01:08:41,160 --> 01:08:42,400
Build the governance as you learn.

1820
01:08:42,400 --> 01:08:45,960
You will move faster than anyone trying to be perfect before they start.

1821
01:08:45,960 --> 01:08:49,520
The shift from systems of record to systems of action is real.

1822
01:08:49,520 --> 01:08:50,840
It's happening right now.

1823
01:08:50,840 --> 01:08:52,800
Your organization will make this transition.

1824
01:08:52,800 --> 01:08:55,440
The only variable is when the agents we've talked about.

1825
01:08:55,440 --> 01:08:58,200
Sales, reconciliation, customer intent.

1826
01:08:58,200 --> 01:09:00,480
They solve structural problems in how you operate.

1827
01:09:00,480 --> 01:09:02,480
They work because they are grounded in data.

1828
01:09:02,480 --> 01:09:03,680
Your actual business data.

1829
01:09:03,680 --> 01:09:07,280
They are safe because they live inside your existing security and order rules.

1830
01:09:07,280 --> 01:09:11,440
And they are valuable because they free your teams to do work that actually requires a human

1831
01:09:11,440 --> 01:09:12,440
brain.

1832
01:09:12,440 --> 01:09:13,440
The organization is moving now.

1833
01:09:13,440 --> 01:09:14,880
We'll outpace everyone else.

1834
01:09:14,880 --> 01:09:16,440
The advantage is speed.

1835
01:09:16,440 --> 01:09:17,960
Start with one process.

1836
01:09:17,960 --> 01:09:18,960
Measure it.

1837
01:09:18,960 --> 01:09:19,960
Scale it.