Dynamics 365 Autonomous Agents - Simply Explained

1
00:00:00,000 --> 00:00:03,540
Imagine you run a busy customer service desk where messages keep coming in,

2
00:00:03,540 --> 00:00:06,060
cases need updates, and customers are waiting.

3
00:00:06,060 --> 00:00:09,760
While the answers they need are buried in old notes nobody has time to dig through.

4
00:00:09,760 --> 00:00:12,920
A copilot handles direct questions when someone walks up and asks,

5
00:00:12,920 --> 00:00:15,040
but an autonomous agent is different AU.

6
00:00:15,040 --> 00:00:17,860
It has a specific job, watches for work that needs doing,

7
00:00:17,860 --> 00:00:20,640
and takes action on its own within the rules you give it.

8
00:00:20,640 --> 00:00:22,960
I'm Mirko Peters from M365, FM.

9
00:00:22,960 --> 00:00:25,280
This knowledge nugget explains the idea in plain English.

10
00:00:25,280 --> 00:00:28,840
By the end you'll know what Dynamics 365 Autonomous Agents actually do,

11
00:00:28,840 --> 00:00:32,440
where they work and why people still stay in charge of the important decisions.

12
00:00:32,440 --> 00:00:35,780
Think of a business like an office building, copilot works at the front desk,

13
00:00:35,780 --> 00:00:41,920
helping visitors who walk up and ask, while agents handle the jobs that happen behind the scenes.

14
00:00:41,920 --> 00:00:46,120
The simple definition, a digital team member, with a narrow job.

15
00:00:46,120 --> 00:00:48,960
Repeat work eats up a lot of time in most companies.

16
00:00:48,960 --> 00:00:53,080
A sales rep sorts incoming leads, a service rep fills in the same case fields

17
00:00:53,080 --> 00:00:55,920
and accountant matches records before the book's close,

18
00:00:55,920 --> 00:00:59,560
and a supply chain worker chases a supplier for a delivery date,

19
00:00:59,560 --> 00:01:03,040
even though the purchase order already has most of the details.

20
00:01:03,040 --> 00:01:06,080
Each task looks small on its own, but the pile grows fast.

21
00:01:06,080 --> 00:01:07,680
So what exactly is an autonomous agent?

22
00:01:07,680 --> 00:01:10,440
It's a specialized AI tool that reviews business data,

23
00:01:10,440 --> 00:01:13,960
picks from a loud next steps, and carries out work toward a defined goal.

24
00:01:13,960 --> 00:01:17,080
It doesn't get a vague instruction like run the company AU.

25
00:01:17,080 --> 00:01:21,000
It gets a narrow job with clear boundaries and that narrow focus is the whole point.

26
00:01:21,000 --> 00:01:24,280
Imagine the office building again, copilot sits at the reception desk.

27
00:01:24,280 --> 00:01:28,480
You walk up and ask for help, and copilot answers your question, finds information,

28
00:01:28,480 --> 00:01:29,840
or helps you write something.

29
00:01:29,840 --> 00:01:33,000
Autonomous agents work in the rooms behind reception AU.

30
00:01:33,000 --> 00:01:35,800
One might handle the mail room sorting incoming requests,

31
00:01:35,800 --> 00:01:38,560
another works the service desk keeping cases up to date,

32
00:01:38,560 --> 00:01:43,040
and a third sits in the accounts office helping staff match financial records.

33
00:01:43,040 --> 00:01:45,840
Many people hear agent and think of a chatbot.

34
00:01:45,840 --> 00:01:48,440
A chatbot waits for you to type a question and then replies,

35
00:01:48,440 --> 00:01:51,600
"Oh, useful for a conversation, but it's just talking."

36
00:01:51,600 --> 00:01:54,320
An autonomous agent can react when something changes.

37
00:01:54,320 --> 00:01:56,960
When a new case arrives, a customer sends another message,

38
00:01:56,960 --> 00:02:00,400
or an order needs confirmation, the agent reviews the situation

39
00:02:00,400 --> 00:02:03,760
and continues the process without someone prompting every single step.

40
00:02:03,760 --> 00:02:05,200
That's the real difference.

41
00:02:05,200 --> 00:02:09,440
Old-style automation handles repeat work too, but follows a fixed path.

42
00:02:09,440 --> 00:02:11,000
If this happens, do that.

43
00:02:11,000 --> 00:02:13,320
If a field has this value, send that email.

44
00:02:13,320 --> 00:02:17,160
Fixed flow still work when everything stays predictable, and agent adds context.

45
00:02:17,160 --> 00:02:19,080
It reads the words in a customer message,

46
00:02:19,080 --> 00:02:22,960
looks at the connected record and chooses between actions its rules allow.

47
00:02:22,960 --> 00:02:25,680
It isn't blindly following a rigid list AU,

48
00:02:25,680 --> 00:02:27,920
it's working with the information right in front of it.

49
00:02:27,920 --> 00:02:30,320
Generative AI helps with the language part.

50
00:02:30,320 --> 00:02:33,520
It can read a long message, identify what the person needs,

51
00:02:33,520 --> 00:02:37,240
write a draft response, or pull a useful answer from your knowledge base.

52
00:02:37,240 --> 00:02:40,240
But writing text alone doesn't complete a business task.

53
00:02:40,240 --> 00:02:43,080
Autonomous behavior connects that understanding to action.

54
00:02:43,080 --> 00:02:46,320
For example, the agent might identify a common support request,

55
00:02:46,320 --> 00:02:51,640
suggest the right knowledge article, update a case field, or ask a person to review something outside its rules.

56
00:02:51,640 --> 00:02:55,160
Your business decides which of those actions it can take on its own.

57
00:02:55,160 --> 00:02:56,640
And this part often gets missed,

58
00:02:56,640 --> 00:02:59,480
an autonomous agent doesn't have free access to everything.

59
00:02:59,480 --> 00:03:03,400
Your organization sets its instructions, data access permissions and approval points,

60
00:03:03,400 --> 00:03:06,560
and the agent works inside that defined space AU,

61
00:03:06,560 --> 00:03:10,280
like a new team member who can enter certain rooms, use certain files,

62
00:03:10,280 --> 00:03:12,120
and complete certain kinds of work.

63
00:03:12,120 --> 00:03:15,040
So it's not magic or a general purpose robot.

64
00:03:15,040 --> 00:03:20,160
It's an AI worker with a narrow job, a set of rules and real business data to work from.

65
00:03:20,160 --> 00:03:23,520
Next, let's look at the office where those agents find that information.

66
00:03:23,520 --> 00:03:29,440
That's Dynamics 365, where they work Dynamics 365 as the business office.

67
00:03:29,440 --> 00:03:33,400
So where do these agents get the specific details they need to turn a general answer

68
00:03:33,400 --> 00:03:35,280
into something useful for your business?

69
00:03:35,280 --> 00:03:37,960
That outtems Dynamics 365.

70
00:03:37,960 --> 00:03:41,960
Think of it as the business office itself with connected rooms for customer records,

71
00:03:41,960 --> 00:03:46,760
sales leads, support cases, orders, money, and supply work.

72
00:03:46,760 --> 00:03:50,760
A customer record tells you who someone is and what they outtem, the bought.

73
00:03:50,760 --> 00:03:54,720
A case shows past conversations, fixes, and who owns the next step.

74
00:03:54,720 --> 00:03:57,360
An order reveals what was promised, where it should go,

75
00:03:57,360 --> 00:03:59,160
and whether anyone still needs to confirm.

76
00:03:59,160 --> 00:04:01,960
That context is what changes the answer.

77
00:04:01,960 --> 00:04:06,600
A generic AI tool here's the sentence, "Omi Delivery" still has no time to arrive.

78
00:04:06,600 --> 00:04:10,000
But it has no clue which delivery you mean, whether the order shipped,

79
00:04:10,000 --> 00:04:12,120
or if there are terms in open case.

80
00:04:12,120 --> 00:04:16,040
Dynamics 365 holds all those business facts in one place.

81
00:04:16,040 --> 00:04:17,720
Her outtems are real scenario.

82
00:04:17,720 --> 00:04:20,400
A customer sends a message asking where their order is,

83
00:04:20,400 --> 00:04:24,000
and that message enters a support conversation or becomes part of a case.

84
00:04:24,000 --> 00:04:26,440
But the message alone tells only half the story.

85
00:04:26,440 --> 00:04:28,960
The agent looks at the customer record connected to that message

86
00:04:28,960 --> 00:04:31,680
and sees if this person contacted support yesterday,

87
00:04:31,680 --> 00:04:33,520
if a colleague already replied,

88
00:04:33,520 --> 00:04:34,880
what product they ordered,

89
00:04:34,880 --> 00:04:37,360
and which knowledge article covers late deliveries.

90
00:04:37,360 --> 00:04:41,160
Then it suggests the next step or carries out an action the organization allows.

91
00:04:41,160 --> 00:04:43,680
Maybe it points the service rep to the right answer,

92
00:04:43,680 --> 00:04:47,080
updates the case with a summary, or prepares a reply for review.

93
00:04:47,080 --> 00:04:50,400
The exact job depends on how the company configured the agent.

94
00:04:50,400 --> 00:04:53,960
But here, ointments the thing, the agent does not want to work in a vacuum.

95
00:04:53,960 --> 00:04:56,960
It works with the record in front of it, the knowledge it can access,

96
00:04:56,960 --> 00:04:58,840
and the permissions attached to its role.

97
00:04:58,840 --> 00:05:01,520
When it lacks information, reaches a sensitive decision,

98
00:05:01,520 --> 00:05:03,600
or needs approval, a person steps in.

99
00:05:03,600 --> 00:05:05,800
People still handle the difficult cases.

100
00:05:05,800 --> 00:05:09,480
Like a customer disputing a charge, a shipment involving a serious failure,

101
00:05:09,480 --> 00:05:11,800
or someone asking for an exception to policy.

102
00:05:11,800 --> 00:05:16,600
Those situations need judgment, authority, and a conversation no agent should complete alone.

103
00:05:16,600 --> 00:05:18,520
Behind the scenes, the path is fairly simple.

104
00:05:18,520 --> 00:05:21,800
Information comes in, the agent reviews the relevant business record,

105
00:05:21,800 --> 00:05:25,560
and approved knowledge takes, suggests, or roots the next action.

106
00:05:25,560 --> 00:05:28,040
And the result goes back into the same record.

107
00:05:28,040 --> 00:05:31,320
That way the team can see what happened without hunting across separate tools.

108
00:05:31,320 --> 00:05:33,160
That record matters more than it sounds.

109
00:05:33,160 --> 00:05:36,840
Without it, a service rep might read a customer email in one place,

110
00:05:36,840 --> 00:05:39,880
search order details in another, check old notes somewhere else,

111
00:05:39,880 --> 00:05:41,960
and then type the same update into a case.

112
00:05:41,960 --> 00:05:45,160
Dynamics 365 brings those work areas together,

113
00:05:45,160 --> 00:05:48,120
so the agent uses the same context as the people on the team.

114
00:05:48,120 --> 00:05:49,960
Generic AI knows how language works,

115
00:05:49,960 --> 00:05:54,440
but a Dynamics 365 agent uses language while looking at the actual business situation

116
00:05:54,440 --> 00:05:57,480
connected to that customer, order, lead, or case.

117
00:05:57,480 --> 00:06:00,120
It does not mean just read the question.

118
00:06:00,120 --> 00:06:02,920
It connects the question to the work already happening around it,

119
00:06:02,920 --> 00:06:05,560
within the access and rules the organization sets.

120
00:06:05,560 --> 00:06:08,360
That do TMEs where these agents become easier to picture,

121
00:06:08,360 --> 00:06:09,880
especially in customer service,

122
00:06:09,880 --> 00:06:11,960
where incoming questions, case updates,

123
00:06:11,960 --> 00:06:14,040
and shared knowledge all meet in one place.

124
00:06:14,040 --> 00:06:18,040
Service agents, finding intent, managing cases,

125
00:06:18,040 --> 00:06:19,160
keeping knowledge.

126
00:06:19,160 --> 00:06:24,360
Current customer service gives us the clearest view of what autonomous agents can do,

127
00:06:24,360 --> 00:06:26,680
support teams deal with the same pattern every day.

128
00:06:26,680 --> 00:06:29,800
Our lots of conversations, lots of small updates,

129
00:06:29,800 --> 00:06:32,600
and useful answers trapped inside work people finished weeks ago.

130
00:06:32,600 --> 00:06:35,160
Let's start with the customer intent agent.

131
00:06:35,160 --> 00:06:37,240
An intent is the reason the customer contacts you,

132
00:06:37,240 --> 00:06:40,600
like our e-con-automate sign-in, our "oh, where is my order?"

133
00:06:40,600 --> 00:06:42,680
"A or A or A" I need to return this item.

134
00:06:42,680 --> 00:06:45,400
A/O support teams need to know these reasons,

135
00:06:45,400 --> 00:06:47,320
because they shape the help articles,

136
00:06:47,320 --> 00:06:50,040
self-service options, and training their people need.

137
00:06:50,040 --> 00:06:53,240
But customer questions don't know to to match these days still.

138
00:06:53,240 --> 00:06:55,960
A company launches a new product, changes a delivery partner,

139
00:06:55,960 --> 00:06:57,400
or updates a billing process,

140
00:06:57,400 --> 00:07:00,440
and then customers start asking new questions in their own words.

141
00:07:00,440 --> 00:07:02,680
The old list of topics no longer fits.

142
00:07:02,680 --> 00:07:05,720
The customer intent agent looks through past and current cases

143
00:07:05,720 --> 00:07:08,200
and customer conversations to find those patterns.

144
00:07:08,200 --> 00:07:10,520
Instead of a manager reading hundreds of chats by hand,

145
00:07:10,520 --> 00:07:13,240
the agent spots that many people mention the same problem

146
00:07:13,240 --> 00:07:15,240
and suggests a new intent for that topic.

147
00:07:15,240 --> 00:07:17,080
That helps keep self-service from going stale

148
00:07:17,080 --> 00:07:20,840
by recommending the knowledge article most likely to help with a customer-automass issue.

149
00:07:20,840 --> 00:07:24,280
That recommendation can support a customer trying to solve a problem alone,

150
00:07:24,280 --> 00:07:28,280
or it can help a service rep find the right answer faster during a conversation.

151
00:07:28,280 --> 00:07:32,280
The agent does no-mud-invent company policy, or it works with the knowledge the company approves

152
00:07:32,280 --> 00:07:34,840
while helping the team see where that knowledge needs work.

153
00:07:34,840 --> 00:07:37,640
Now think about the case itself, A/O.

154
00:07:37,640 --> 00:07:41,240
The work record for a customer problem, that includes the customer A/O,

155
00:07:41,240 --> 00:07:42,680
TM.piles.missage.

156
00:07:42,680 --> 00:07:45,080
Notes from service reps, the current status,

157
00:07:45,080 --> 00:07:47,080
and the steps taken to solve it.

158
00:07:47,080 --> 00:07:49,480
Creating and updating those records takes time,

159
00:07:49,480 --> 00:07:52,120
especially when a team handles a high volume of requests.

160
00:07:52,120 --> 00:07:55,480
The case management agent in Dynamics 365 Customer Service

161
00:07:55,480 --> 00:07:59,160
helps with the case lifecycle, out creating cases, updating them,

162
00:07:59,160 --> 00:08:02,280
moving them toward resolution, following up and closing them.

163
00:08:02,280 --> 00:08:05,480
A service rep spends less time filling in routine details,

164
00:08:05,480 --> 00:08:08,520
and more time reading the customer-autom situation carefully.

165
00:08:08,520 --> 00:08:10,040
Every support team works differently.

166
00:08:10,040 --> 00:08:13,320
One company wants a case closed only after a customer confirms,

167
00:08:13,320 --> 00:08:16,360
another needs a manager to review cases above a certain level,

168
00:08:16,360 --> 00:08:19,400
and some cases need a specific follow-up after a set period.

169
00:08:19,400 --> 00:08:22,920
Administrators shape the case management agent around those local rules,

170
00:08:22,920 --> 00:08:27,160
so it follows the company U/TMA's process rather than a generic path.

171
00:08:27,160 --> 00:08:29,960
It handles repeat steps, but the organization decides

172
00:08:29,960 --> 00:08:32,520
when a person needs to check the work or take control.

173
00:08:32,520 --> 00:08:35,800
There are terms another problem in service that many teams know well.

174
00:08:35,800 --> 00:08:37,960
A skilled support rep solves a tricky issue,

175
00:08:37,960 --> 00:08:40,600
writes useful notes, and closes the case.

176
00:08:40,600 --> 00:08:42,840
Later, another customer reports the same problem,

177
00:08:42,840 --> 00:08:45,160
but the answer sits buried in old case notes,

178
00:08:45,160 --> 00:08:47,880
a chat transcript, or a summary that nobody thinks to search.

179
00:08:47,880 --> 00:08:51,080
That U/Temarts, where the customer-knowledge management agent comes in,

180
00:08:51,080 --> 00:08:54,920
it examines completed case details to pull useful knowledge, find gaps,

181
00:08:54,920 --> 00:08:59,160
and keep knowledge current while checking for the company-autom that his compliance needs.

182
00:08:59,160 --> 00:09:02,760
Think of a support team, or a team as old case notes,

183
00:09:02,760 --> 00:09:05,240
like loose papers spread across many drawers.

184
00:09:05,240 --> 00:09:07,800
Someone may have written the perfect answer six months ago,

185
00:09:07,800 --> 00:09:11,080
but it would now, to may, it help the next customer if nobody can find it.

186
00:09:11,080 --> 00:09:13,800
The knowledge agent helps turn repeat answers into material

187
00:09:13,800 --> 00:09:16,760
the team can review and place in a shared knowledge folder.

188
00:09:16,760 --> 00:09:19,000
People still review what becomes official guidance

189
00:09:19,000 --> 00:09:21,720
because a case note can describe a one-time exception,

190
00:09:21,720 --> 00:09:25,080
an outdated workaround, or advice that only fit one customer.

191
00:09:25,080 --> 00:09:26,760
The agent helps surface the material,

192
00:09:26,760 --> 00:09:29,320
but people decide what should become a trusted article.

193
00:09:29,320 --> 00:09:32,120
Here out, Merser, real example, a customer writes,

194
00:09:32,120 --> 00:09:33,720
"Omi deliveries late again.

195
00:09:33,720 --> 00:09:35,640
Can you tell me what is happening?"

196
00:09:35,640 --> 00:09:37,800
A.O. the customer intent agent recognizes

197
00:09:37,800 --> 00:09:39,880
that this belongs to a delivery-related topic

198
00:09:39,880 --> 00:09:42,120
and points toward the right knowledge article.

199
00:09:42,120 --> 00:09:43,560
If the request needs a case,

200
00:09:43,560 --> 00:09:46,200
the case management agent helps capture the details,

201
00:09:46,200 --> 00:09:50,360
update the record, and support the next steps under the team i.e. our team's rules.

202
00:09:50,360 --> 00:09:54,040
After the issue closes, the customer knowledge management agent

203
00:09:54,040 --> 00:09:57,560
reviews the case for information that may improve the help content later.

204
00:09:57,560 --> 00:09:59,720
If the late delivery came from a new problem,

205
00:09:59,720 --> 00:10:01,560
many customers now report,

206
00:10:01,560 --> 00:10:03,560
a person investigates, approves a new article

207
00:10:03,560 --> 00:10:05,400
or speaks directly with the customer.

208
00:10:05,400 --> 00:10:08,200
The agents reduce the repeat admin work around the case,

209
00:10:08,200 --> 00:10:10,600
while the service team handles the judgment calls.

210
00:10:10,600 --> 00:10:13,640
Service agents focus on customer questions and shared answers.

211
00:10:13,640 --> 00:10:16,040
In other parts of the business, agents take on different jobs,

212
00:10:16,040 --> 00:10:18,760
from sorting sales leads to checking financial records

213
00:10:18,760 --> 00:10:20,280
and following up with suppliers.

214
00:10:20,280 --> 00:10:24,680
Sales, finance, and supply chain agents with different desks.

215
00:10:24,680 --> 00:10:27,960
Customer service is only one part of the picture.

216
00:10:27,960 --> 00:10:29,960
Actually, you'll find the same pattern everywhere,

217
00:10:29,960 --> 00:10:32,680
O. sales teams, finance teams, purchasing teams,

218
00:10:32,680 --> 00:10:35,400
dispatchers all have work that follows a routine

219
00:10:35,400 --> 00:10:38,280
where the details change from one record to the next,

220
00:10:38,280 --> 00:10:39,960
but the shape stays the same.

221
00:10:39,960 --> 00:10:41,160
So let's start with sales.

222
00:10:41,160 --> 00:10:43,880
Imagine a salesperson receives a large number of inbound leads.

223
00:10:43,880 --> 00:10:46,360
Some people are ready to talk, others are just looking around,

224
00:10:46,360 --> 00:10:48,200
and a few may not fit the company at all.

225
00:10:48,200 --> 00:10:49,000
Here's the issue.

226
00:10:49,000 --> 00:10:51,080
Someone still needs to research each one

227
00:10:51,080 --> 00:10:52,840
and decide who deserves attention first.

228
00:10:52,840 --> 00:10:54,280
That's where the sales qualification agent

229
00:10:54,280 --> 00:10:56,280
for Dynamics 365 sales steps.

230
00:10:56,280 --> 00:10:58,520
In now, it helps with that early sorting work.

231
00:10:58,520 --> 00:11:01,160
This agent can research and prioritize inbound leads,

232
00:11:01,160 --> 00:11:04,120
then develop personalized sales emails to start a conversation.

233
00:11:04,120 --> 00:11:06,920
But here's the thing, it doesn't close a deal by itself,

234
00:11:06,920 --> 00:11:08,680
and it doesn't replace the salesperson

235
00:11:08,680 --> 00:11:10,200
who builds trust with the customer.

236
00:11:10,200 --> 00:11:13,160
Instead, it helps the salesperson begin with a clearer list.

237
00:11:13,160 --> 00:11:15,240
Picture a crowded sales inbox on Monday morning

238
00:11:15,240 --> 00:11:18,440
where without help, a rep might spend hours opening messages,

239
00:11:18,440 --> 00:11:19,800
searching for company details,

240
00:11:19,800 --> 00:11:21,880
and drafting similar first replies.

241
00:11:21,880 --> 00:11:23,640
The agent can prepare that early work,

242
00:11:23,640 --> 00:11:25,160
so the rep can focus on the leads

243
00:11:25,160 --> 00:11:27,560
that look most ready for a real conversation.

244
00:11:27,560 --> 00:11:30,920
Now, a person still decides what good sales contact looks like.

245
00:11:30,920 --> 00:11:32,600
They review the email before it goes out,

246
00:11:32,600 --> 00:11:34,600
add their own knowledge, and decide whether a lead

247
00:11:34,600 --> 00:11:37,320
deserves a call while the agent handles the repeat research

248
00:11:37,320 --> 00:11:39,000
and drafting around the edges.

249
00:11:39,000 --> 00:11:42,120
Small and medium-sized businesses often face a related job

250
00:11:42,120 --> 00:11:44,040
after a customer decides to buy.

251
00:11:44,040 --> 00:11:46,680
An order arrives through email or another channel,

252
00:11:46,680 --> 00:11:49,160
and someone needs to enter it, check the details,

253
00:11:49,160 --> 00:11:52,360
capture the customer's preferences, and send confirmation.

254
00:11:52,360 --> 00:11:55,000
That work can take time, especially when orders arrive

255
00:11:55,000 --> 00:11:56,520
in different formats.

256
00:11:56,520 --> 00:11:59,880
The sales order agent in Dynamics 365 Business Central

257
00:11:59,880 --> 00:12:01,960
handles the order intake process from entry

258
00:12:01,960 --> 00:12:03,560
all the way through confirmation.

259
00:12:03,560 --> 00:12:05,400
This agent can interact with customers

260
00:12:05,400 --> 00:12:07,560
while gathering the information needed for an order,

261
00:12:07,560 --> 00:12:10,040
so the team doesn't need to copy every detail by hand

262
00:12:10,040 --> 00:12:11,640
before moving the order forward.

263
00:12:11,640 --> 00:12:14,040
Now, Finance has its own pile of repeat work.

264
00:12:14,040 --> 00:12:15,960
Before a financial period closes,

265
00:12:15,960 --> 00:12:17,880
teams need to prepare and clean data sets

266
00:12:17,880 --> 00:12:19,880
where numbers sit in different records,

267
00:12:19,880 --> 00:12:21,480
formatting needs attention,

268
00:12:21,480 --> 00:12:24,040
and people spend long stretches finding information

269
00:12:24,040 --> 00:12:25,880
that doesn't match.

270
00:12:25,880 --> 00:12:27,400
The Financial Reconciliation Agent

271
00:12:27,400 --> 00:12:30,040
for Microsoft 365 Co-Pilot for Finance

272
00:12:30,040 --> 00:12:31,880
helps prepare and cleanse data sets

273
00:12:31,880 --> 00:12:33,160
for that period-close work,

274
00:12:33,160 --> 00:12:34,600
think of it as sorting paperwork

275
00:12:34,600 --> 00:12:37,160
before the accountant begins the serious review.

276
00:12:37,160 --> 00:12:39,320
It doesn't decide whether the company's financial reports

277
00:12:39,320 --> 00:12:41,880
are correct or finance professionals

278
00:12:41,880 --> 00:12:43,480
still own that responsibility.

279
00:12:43,480 --> 00:12:46,600
There's another agent inside Dynamics 365 Finance,

280
00:12:46,600 --> 00:12:48,360
the Account Reconciliation Agent.

281
00:12:48,360 --> 00:12:50,040
It helps match and clear transactions

282
00:12:50,040 --> 00:12:52,440
between sub-ledges and the General Ledger O,

283
00:12:52,440 --> 00:12:53,720
those terms sound technical.

284
00:12:53,720 --> 00:12:54,920
So let's keep this simple.

285
00:12:54,920 --> 00:12:57,320
A sub-ledger holds detailed records for one area,

286
00:12:57,320 --> 00:12:59,400
like customer payments or supplier bills,

287
00:12:59,400 --> 00:13:01,720
while the General Ledger holds the larger financial record

288
00:13:01,720 --> 00:13:02,760
for the business.

289
00:13:02,760 --> 00:13:04,680
Accountants need those records to agree,

290
00:13:04,680 --> 00:13:05,800
and when they don't,

291
00:13:05,800 --> 00:13:07,240
someone needs to find the gap.

292
00:13:07,240 --> 00:13:09,320
The agent helps with the matching and clearing work,

293
00:13:09,320 --> 00:13:11,240
so accountants can spend more of their time

294
00:13:11,240 --> 00:13:13,000
investigating records that don't fit

295
00:13:13,000 --> 00:13:15,880
rather than checking every obvious match one by one.

296
00:13:15,880 --> 00:13:18,200
Supply chain teams have another kind of follow-up problem.

297
00:13:18,200 --> 00:13:20,440
A business sends a purchase order to a supplier,

298
00:13:20,440 --> 00:13:22,120
but the team still needs confirmation

299
00:13:22,120 --> 00:13:24,200
that the supplier can deliver it when expected.

300
00:13:24,200 --> 00:13:25,320
If nobody follows up,

301
00:13:25,320 --> 00:13:26,920
a delay can sit unnoticed

302
00:13:26,920 --> 00:13:28,840
until it becomes a customer problem.

303
00:13:28,840 --> 00:13:31,320
The Supply Accomunications Agent for Dynamics,

304
00:13:31,320 --> 00:13:33,240
365 Supply Chain Management,

305
00:13:33,240 --> 00:13:34,840
handles collaboration with suppliers

306
00:13:34,840 --> 00:13:36,360
to confirm order delivery.

307
00:13:36,360 --> 00:13:38,920
It can handle work connected to confirming purchase orders

308
00:13:38,920 --> 00:13:41,160
and help spot possible delays earlier.

309
00:13:41,160 --> 00:13:43,880
That way, procurement specialists can spend more time

310
00:13:43,880 --> 00:13:46,520
on the supplier relationships and delivery problems

311
00:13:46,520 --> 00:13:48,760
that need a human conversation.

312
00:13:48,760 --> 00:13:50,520
Field service adds movement to the picture.

313
00:13:50,520 --> 00:13:53,080
A dispatcher may plan a technician's day carefully,

314
00:13:53,080 --> 00:13:54,760
but then traffic slows someone down,

315
00:13:54,760 --> 00:13:57,320
a customer cancels, or two bookings collide.

316
00:13:57,320 --> 00:13:58,920
The schedule that looked fine at 8 o'clock

317
00:13:58,920 --> 00:14:01,000
may create gaps and conflicts by lunchtime.

318
00:14:01,000 --> 00:14:04,440
The scheduling operations agent for Dynamics 365 Field Service

319
00:14:04,440 --> 00:14:07,480
helps dispatchers adjust schedules as those conditions change.

320
00:14:07,480 --> 00:14:09,880
It helps them work toward better schedules for technicians

321
00:14:09,880 --> 00:14:12,360
when the day refuses to follow the original plan.

322
00:14:12,360 --> 00:14:14,200
The dispatchers still decide what makes sense

323
00:14:14,200 --> 00:14:15,720
for their customers and their staff,

324
00:14:15,720 --> 00:14:17,880
especially when a change involves priority work

325
00:14:17,880 --> 00:14:19,640
or a difficult customer promise.

326
00:14:19,640 --> 00:14:21,400
So these agents work at different desks,

327
00:14:21,400 --> 00:14:23,240
but their jobs follow the same shape.

328
00:14:23,240 --> 00:14:25,720
The sales qualification agent works with leads,

329
00:14:25,720 --> 00:14:28,280
the reconciliation agents work with financial records,

330
00:14:28,280 --> 00:14:30,520
the supplier agent works with purchase orders

331
00:14:30,520 --> 00:14:33,400
and the scheduling agent works with technician appointments.

332
00:14:33,400 --> 00:14:35,560
Each one focuses on a defined business task,

333
00:14:35,560 --> 00:14:37,560
uses the data connected to that task,

334
00:14:37,560 --> 00:14:39,800
takes only the actions it has permission to take

335
00:14:39,800 --> 00:14:42,360
and leaves a record of the work for the people who follow.

336
00:14:42,360 --> 00:14:45,240
More action can save time,

337
00:14:45,240 --> 00:14:47,480
but it also raises a fair question

338
00:14:47,480 --> 00:14:50,280
who decides what an agent can read, change, and send.

339
00:14:50,280 --> 00:14:53,640
Control, trust, and the human team.

340
00:14:53,640 --> 00:14:57,320
Here's a myth I want to clear up right away.

341
00:14:57,320 --> 00:15:00,840
Don't ask whether an agent can act, ask who decided what it can do.

342
00:15:01,880 --> 00:15:04,280
Every action comes from choices the organization made

343
00:15:04,280 --> 00:15:06,120
before it ever touched a customer record

344
00:15:06,120 --> 00:15:08,040
and order or a financial transaction.

345
00:15:08,040 --> 00:15:09,240
People remain responsible.

346
00:15:09,240 --> 00:15:11,400
Your company chooses the job, sets the rules,

347
00:15:11,400 --> 00:15:13,240
and decides the data the agent can see,

348
00:15:13,240 --> 00:15:14,600
the actions it can take,

349
00:15:14,600 --> 00:15:16,840
and the moments where a human must approve the next step.

350
00:15:16,840 --> 00:15:19,480
An agent may handle the routine path,

351
00:15:19,480 --> 00:15:21,160
but it doesn't carry the responsibility

352
00:15:21,160 --> 00:15:22,760
for a bad customer promise,

353
00:15:22,760 --> 00:15:25,080
an incorrect payment or a sensitive decision.

354
00:15:25,080 --> 00:15:28,200
Let's bring back the reception desk analogy.

355
00:15:28,200 --> 00:15:29,560
Think of EntraID,

356
00:15:29,560 --> 00:15:31,320
Microsoft's identity service,

357
00:15:31,320 --> 00:15:33,720
as the reception desk for the digital workplace.

358
00:15:33,720 --> 00:15:36,120
When an employee signs in EntraID,

359
00:15:36,120 --> 00:15:38,120
helps confirm who that person is

360
00:15:38,120 --> 00:15:39,640
and what rooms they may enter.

361
00:15:39,640 --> 00:15:41,720
The same idea controls the agent O.

362
00:15:41,720 --> 00:15:44,520
It should only reach the records and tools needed

363
00:15:44,520 --> 00:15:45,720
for its assigned job,

364
00:15:45,720 --> 00:15:47,960
not wander through every part of the business

365
00:15:47,960 --> 00:15:49,320
that keeps the job narrow.

366
00:15:49,320 --> 00:15:50,760
Before an agent starts work,

367
00:15:50,760 --> 00:15:52,840
the organization needs clear guardrails.

368
00:15:52,840 --> 00:15:54,520
What may it read, what may it update,

369
00:15:54,520 --> 00:15:56,680
can it prepare an email or can it send one,

370
00:15:56,680 --> 00:15:58,600
can it suggest a case should close

371
00:15:58,600 --> 00:16:00,840
or does a service rep need to approve that closure?

372
00:16:01,240 --> 00:16:04,040
Those decisions are what turn a broad AI idea

373
00:16:04,040 --> 00:16:05,800
into a controlled business process.

374
00:16:05,800 --> 00:16:08,360
A sensible first project usually involves repeatable work

375
00:16:08,360 --> 00:16:10,200
where a mistake has a clear way back.

376
00:16:10,200 --> 00:16:12,120
Maybe the agent drafts follow-up messages

377
00:16:12,120 --> 00:16:13,960
for review, sorts incoming requests

378
00:16:13,960 --> 00:16:15,480
or fills in routine case details.

379
00:16:15,480 --> 00:16:16,760
The team can check the results,

380
00:16:16,760 --> 00:16:17,560
correct errors,

381
00:16:17,560 --> 00:16:18,440
and improve the setup

382
00:16:18,440 --> 00:16:20,040
before allowing more independent action.

383
00:16:20,040 --> 00:16:20,840
Start small,

384
00:16:20,840 --> 00:16:22,680
but that doesn't mean ignoring the risks.

385
00:16:22,680 --> 00:16:24,680
An agent can misunderstand a request,

386
00:16:24,680 --> 00:16:26,040
use poor information

387
00:16:26,040 --> 00:16:27,160
or produce an answer

388
00:16:27,160 --> 00:16:29,160
that doesn't fit the customer's situation.

389
00:16:29,160 --> 00:16:30,840
If it reaches data, it shouldn't see

390
00:16:30,840 --> 00:16:32,200
the problem grows quickly.

391
00:16:32,200 --> 00:16:34,600
Some industries also face strict rules about privacy,

392
00:16:34,600 --> 00:16:36,440
record keeping, and how decisions are checked.

393
00:16:36,440 --> 00:16:38,360
So the work continues after setup

394
00:16:38,360 --> 00:16:40,440
and teams need to review what the agent did,

395
00:16:40,440 --> 00:16:42,680
read feedback from employees and customers,

396
00:16:42,680 --> 00:16:44,600
and inspect the records it touched.

397
00:16:44,600 --> 00:16:46,280
If it repeatedly misses context,

398
00:16:46,280 --> 00:16:48,440
the organization can adjust its instructions,

399
00:16:48,440 --> 00:16:50,040
improve its approved knowledge,

400
00:16:50,040 --> 00:16:51,080
change permissions,

401
00:16:51,080 --> 00:16:52,840
or add a human approval point.

402
00:16:52,840 --> 00:16:54,280
The agent needs supervision

403
00:16:54,280 --> 00:16:57,000
just like any process that affects customers or money.

404
00:16:57,000 --> 00:16:58,360
This changes the human job,

405
00:16:58,360 --> 00:16:59,640
but it doesn't remove it.

406
00:16:59,640 --> 00:17:01,560
Instead of copying details between records

407
00:17:01,560 --> 00:17:03,080
or chasing routine updates,

408
00:17:03,080 --> 00:17:05,000
people can focus on the exceptions,

409
00:17:05,000 --> 00:17:06,440
the unclear cases,

410
00:17:06,440 --> 00:17:09,160
and the decisions that need experience and accountability.

411
00:17:09,160 --> 00:17:10,600
Someone still owns the outcome,

412
00:17:10,600 --> 00:17:12,200
and that's the simple rule to remember.

413
00:17:12,200 --> 00:17:14,600
An agent can carry out a defined piece of work,

414
00:17:14,600 --> 00:17:16,280
but people decide the boundaries

415
00:17:16,280 --> 00:17:18,760
and remain answerable for what happens inside them.

416
00:17:18,760 --> 00:17:19,560
Other.

417
00:17:19,560 --> 00:17:20,280
Conclusion.

418
00:17:20,280 --> 00:17:22,520
One job, clear rules, human oversight.

419
00:17:22,520 --> 00:17:24,360
So here's the bottom line.

420
00:17:24,360 --> 00:17:26,680
Dynamics 365 Autonomous Agents

421
00:17:26,680 --> 00:17:29,720
handle one specific job using context already in your work,

422
00:17:29,720 --> 00:17:30,680
but you set the rules,

423
00:17:30,680 --> 00:17:31,880
you approve the actions,

424
00:17:31,880 --> 00:17:33,160
and you own the outcome.

425
00:17:33,160 --> 00:17:34,280
That's the whole point.

426
00:17:34,280 --> 00:17:36,280
Subscribe on your favorite podcast platform

427
00:17:36,280 --> 00:17:37,880
for more plain English knowledge nuggets,

428
00:17:37,880 --> 00:17:40,520
and share this with someone who hears AI agents

429
00:17:40,520 --> 00:17:42,920
and still pictures a chatbot answering questions.