Copilot Studio, Dataverse MCP & The Future of Agentic AI in Microsoft 365 with Nathan Rose [MVP]

1
00:00:00,000 --> 00:00:07,040
Welcome to back to the M665, the podcast where we explore the people, technologies and ideas,

2
00:00:07,040 --> 00:00:11,400
shaping the future of Microsoft 365 AI and modern world places.

3
00:00:11,400 --> 00:00:16,160
Today I'm joined by Nature Rose, a Microsoft business application and we be from Auckland.

4
00:00:16,160 --> 00:00:20,640
Nathan is a power platform solution architect with experience,

5
00:00:20,640 --> 00:00:26,960
interesting from the CRM to 2011 days all the way to days,

6
00:00:26,960 --> 00:00:29,200
AI part, local ecosystem.

7
00:00:29,680 --> 00:00:32,720
He worked with organizations across Australia and New Zealand,

8
00:00:32,720 --> 00:00:38,640
helping both public and private sector, customers build enterprise-grads solutions

9
00:00:38,640 --> 00:00:40,800
using Microsoft business applications.

10
00:00:40,800 --> 00:00:44,800
Nathan is also a particular passionate about co-pilot studio data

11
00:00:44,800 --> 00:00:52,320
worse and MCP and helping organizations create scalable AI agents that actively solve business problems.

12
00:00:52,320 --> 00:00:56,080
You may know Nathan from his content platform, no code no mercy,

13
00:00:56,080 --> 00:01:00,640
his YouTube channel off from speaking at events, likes and New Zealand Business

14
00:01:00,640 --> 00:01:02,560
Application Summit and definitely.

15
00:01:02,560 --> 00:01:06,320
Yeah Nathan, welcome to the 365 podcast.

16
00:01:06,320 --> 00:01:08,880
Thanks so much, Mirko. Thanks for having me on.

17
00:01:08,880 --> 00:01:09,600
Great to be here.

18
00:01:09,600 --> 00:01:17,840
So tell us a little bit about your journey into the Microsoft technology.

19
00:01:17,840 --> 00:01:24,560
Sure. So I always used to say that I have a non-traditional background for

20
00:01:25,280 --> 00:01:30,400
technical people but that's not actually true anymore because in the low-code

21
00:01:30,400 --> 00:01:35,840
power platform world pretty much all of us come from doing something else.

22
00:01:35,840 --> 00:01:38,960
Mark Smith was a groundskeeper.

23
00:01:38,960 --> 00:01:46,880
Craig White was a bookkeeper, Natalie Leander's,

24
00:01:46,880 --> 00:01:53,600
was a hairdresser. We all did different things and then eventually got hands-on with

25
00:01:54,160 --> 00:01:57,040
power platform, fell in love with it and here we all are.

26
00:01:57,040 --> 00:02:03,280
So I worked in corporate sales for about a decade, was really bad at it.

27
00:02:03,280 --> 00:02:06,000
Don't ask me how to close a deal, I don't know.

28
00:02:06,000 --> 00:02:15,760
And segwayed into working as a business analyst and being a BA is not a bad role.

29
00:02:15,760 --> 00:02:21,280
And just for me, I got to, and I was always working with Dynamics, CRM and

30
00:02:21,280 --> 00:02:26,080
eventually power platform. But I found that I was much more interested in actually solving

31
00:02:26,080 --> 00:02:30,400
the problem than doing what a BA is supposed to do, which is to find the problem.

32
00:02:30,400 --> 00:02:35,840
Yeah. And so, yeah, I eventually found my way into the consulting world and

33
00:02:35,840 --> 00:02:40,880
yeah, that's what I've been doing for the last six years working as a consultant and architect

34
00:02:40,880 --> 00:02:43,360
with Dynamics and Power Platform.

35
00:02:43,360 --> 00:02:49,920
This is also, yeah, I think a lot of people have not the classic, I don't know,

36
00:02:49,920 --> 00:02:58,880
study IT or something background. And yeah, with the citizen development, Microsoft, do a lot of

37
00:02:58,880 --> 00:03:08,080
that all the, I say, normal people can join these areas. But your long time was in the CRM area,

38
00:03:08,080 --> 00:03:10,560
what changes to see you there?

39
00:03:10,560 --> 00:03:18,240
Yeah. So I started with version 2011 of CRM and just always worked with

40
00:03:19,440 --> 00:03:23,200
what today is called Dynamics 365 CE or customer engagement.

41
00:03:23,200 --> 00:03:31,440
In 2019, I was part of a private preview for what we now know today is co-pilot studio,

42
00:03:31,440 --> 00:03:39,760
what was Power Virtual Agent, what it was first introduced. And I absolutely fell in love

43
00:03:39,760 --> 00:03:44,240
with the Power Platform because for the first time, I got hands on the Power Automate

44
00:03:44,240 --> 00:03:50,560
and all these different tools because on its own Power Virtual Agent when it first came out was

45
00:03:50,560 --> 00:03:54,880
somewhat limited. It was really cool that we could build these deterministic chat experiences.

46
00:03:54,880 --> 00:04:00,080
But what was really amazing about it was we could hook it up to Dataverse by way of Power Automate,

47
00:04:00,080 --> 00:04:05,200
which is the first thing I did. And back then, we didn't have connectors. I literally had

48
00:04:05,200 --> 00:04:11,200
people in product group in Redmond writing JSONs for me because prior to this, you know, I thought

49
00:04:11,200 --> 00:04:16,400
JSON was a movie character. I didn't realize that it was something that we actually did professionally.

50
00:04:16,400 --> 00:04:22,720
But that was how we connected to Dataverse using HTTP connectors.

51
00:04:22,720 --> 00:04:31,440
So I just absolutely fell in love with it. And kind of started my tradition of

52
00:04:31,440 --> 00:04:36,720
the demos I build are things that you should not try at home. Don't do this in production kids.

53
00:04:37,760 --> 00:04:42,880
I literally hooked Dataverse up to the to the bottom was opening and closing cases and customer

54
00:04:42,880 --> 00:04:48,560
service and inserting rows and doing all kinds of things. And I showed it to a colleague whose first

55
00:04:48,560 --> 00:04:53,920
response was, why are you letting authenticated, unauthenticated users into Dataverse? And I was like,

56
00:04:53,920 --> 00:04:58,480
well, you know, the joy of preview features is we get to work out what's possible. And then we walk

57
00:04:58,480 --> 00:05:06,000
it back to what's responsible for GA. But that was how I got into Power Platform through what we

58
00:05:06,000 --> 00:05:10,000
now know today is Copilot Studio. So it is something very special to my heart.

59
00:05:10,000 --> 00:05:18,000
And what could you give advice for someone who entering the Power Platform, so they don't field,

60
00:05:18,000 --> 00:05:20,720
like meeting JSON?

61
00:05:20,720 --> 00:05:32,320
Yeah, I would say, get hands on as fast as you can. Don't be afraid. So just try stuff.

62
00:05:33,040 --> 00:05:39,520
The community is incredibly welcoming. That's probably one of the best things about Power Platform.

63
00:05:39,520 --> 00:05:47,600
Well, it's just everybody is so incredibly inclusive and welcoming. We're not fussed about

64
00:05:47,600 --> 00:05:54,800
what you study or what your background is. We just want to hear your ideas and we want to

65
00:05:54,800 --> 00:06:00,160
help you solve problems. And that's kind of what we all do for each other. And that's what makes it

66
00:06:00,160 --> 00:06:05,200
such an amazing thing to be a part of. So it's say there's never been a better time to get into

67
00:06:05,200 --> 00:06:12,480
Power Platform if you're thinking of doing it. The best way is just to get hands on. And you're

68
00:06:12,480 --> 00:06:19,600
probably going to find, you know, you'll just start doing solving problems in your workplace today.

69
00:06:19,600 --> 00:06:25,200
That's how most of us got onto it. You know, we discovered Power Apps, Power Automate, Dataverse,

70
00:06:25,200 --> 00:06:32,880
things like that. And we started making business problems go away after I had had my private preview

71
00:06:32,880 --> 00:06:40,560
experience with Power Virtual Agent and started to get good with Power Automate. I had a process

72
00:06:40,560 --> 00:06:45,200
as part of my job that took three days a week of my time. So there was a design flaw on our CRM.

73
00:06:45,200 --> 00:06:52,880
We had a custom table associated to cases that had payment details for warranty upgrades. So when

74
00:06:52,880 --> 00:06:57,840
people wanted to, you know, they had a warranty issue and it was like, hey, we could fix it or for

75
00:06:57,840 --> 00:07:04,320
another hundred bucks, you could upgrade, right? So who's not going to do that? The problem was,

76
00:07:04,320 --> 00:07:11,360
they would close the cases which meant all the child tables, you couldn't do anything on them. So

77
00:07:11,360 --> 00:07:18,000
by the time something got shipped from the factory, the case was closed and so the credit cards

78
00:07:18,000 --> 00:07:23,520
never got built. So we get these huge spreadsheets from finance saying we've got $30,000 worth of

79
00:07:23,520 --> 00:07:27,840
product that we've shipped and, you know, we haven't collected it in cash for it. And so I would

80
00:07:27,840 --> 00:07:34,880
literally have to go in, open the case, do the thing to build a credit card and close the case again.

81
00:07:34,880 --> 00:07:40,320
And that took three days a week in my time. And I eventually got to the point where I had it built

82
00:07:40,320 --> 00:07:44,720
in Power Automate. And this was a iterative thing. But by the end, literally the email would come in

83
00:07:44,720 --> 00:07:52,320
from the failed integration, snip out the case ID from the subject line. Go look it up,

84
00:07:52,320 --> 00:07:58,080
open the case, do the thing, close the case, and three days a week in my time went to zero. And

85
00:07:58,080 --> 00:08:04,480
so those are the kinds of things, you know, if you've got access to Power Platform to start looking

86
00:08:04,480 --> 00:08:09,920
for those things and work, it's even more exciting now is we're moving into this agentic world. So

87
00:08:09,920 --> 00:08:17,280
the example I just gave was in this, you know, deterministic world that most of us lived in prior to

88
00:08:17,280 --> 00:08:24,240
LLM's. Today it is, you know, there's so many more problems that you can solve because you are not

89
00:08:24,240 --> 00:08:30,400
limited by, you know, we have this deterministic thing that always must work this way, you know, we can

90
00:08:30,400 --> 00:08:37,680
start to, you know, when we need to look at an email and decide who it goes to and all that type of

91
00:08:37,680 --> 00:08:44,880
stuff, we can start to bring in that agentic layer. And in many ways, building agents is in some ways

92
00:08:44,880 --> 00:08:51,280
a lot easier than building the deterministic workflows that we did in the past. I mean, it's funny,

93
00:08:51,280 --> 00:08:55,360
I was thinking about this this morning. There are so many times when I'm doing something with

94
00:08:55,360 --> 00:08:59,920
generative AI, whether it's building a prompt, whether it's creating a skill, whether it's

95
00:08:59,920 --> 00:09:06,640
building an agent where it's literally as easy as just writing out what I wanted to do.

96
00:09:07,280 --> 00:09:12,240
In the low-code world, it was never that easy, you know, it was, you know, okay, I might describe what I

97
00:09:12,240 --> 00:09:16,560
wanted to do, but that's the document and then I have to go in and start configuring stuff. Now,

98
00:09:16,560 --> 00:09:21,600
like, I literally just say, I want you to go do these things and it goes and does those things. So

99
00:09:21,600 --> 00:09:24,800
the barrier to entry is getting lower and lower all the time.

100
00:09:24,800 --> 00:09:32,880
What did you think is, okay, we have a low-code, no-code, and I think,

101
00:09:32,880 --> 00:09:41,600
Goliik, Basbord is, is, is, is, why I quote, but how did think, is there, what is there any new

102
00:09:41,600 --> 00:09:51,520
challenges that's coming up with this kind of AI functions and, oh, yeah, tons. So I, I released a

103
00:09:51,520 --> 00:09:57,760
video on Friday showing some of the new dataverse plugins from the product team and doing some

104
00:09:57,760 --> 00:10:02,400
amazing things with, with GitHub Copilot, stuff that, you know, used to take us hours and hours.

105
00:10:02,800 --> 00:10:07,440
Creating tables and columns and adding stuff to forms and moving users between business units.

106
00:10:07,440 --> 00:10:12,080
And that's just the tip of the iceberg of what these things can do. And in the comments, you know,

107
00:10:12,080 --> 00:10:18,000
there's all the usual, well, you know, you've got to look out for hallucinations and, you know,

108
00:10:18,000 --> 00:10:24,880
this doesn't replace good design and all that and it's, and, like, yeah, of course, obviously. But,

109
00:10:27,440 --> 00:10:33,600
yeah, it's, so, you know, you do need, you know, those are all things that need to be taken into

110
00:10:33,600 --> 00:10:38,960
consideration. I don't, you know, even those of us who, you know, I would say, yeah, I'm a bit of a

111
00:10:38,960 --> 00:10:47,600
cheerleader and an apologist for, for AI and, and natural language. But at the same time, we've got to

112
00:10:47,600 --> 00:10:54,320
do these things responsibly. And so, yeah, you know, there, there are a lot of challenges that we have

113
00:10:54,320 --> 00:11:00,160
to look at in terms of, let's make sure that, you know, we are getting the intended effect that we're

114
00:11:00,160 --> 00:11:08,640
not creating, you know, just more issues for ourselves. I think a lot of company think also about

115
00:11:08,640 --> 00:11:17,200
the topic, governance, compliance, risk. How do you see these topics, especially with this AI,

116
00:11:17,200 --> 00:11:22,560
all-partful, local staff? Yep. I think they're becoming increasingly more important. So,

117
00:11:23,200 --> 00:11:30,160
in the low-code era of power platform, we didn't hear a lot about governance until kind of

118
00:11:30,160 --> 00:11:38,320
the late 2010s, 2020s. That's when governance started to kind of take off as a conversation.

119
00:11:38,320 --> 00:11:49,520
The power platform, sorry, I'm going to blank on the governance tool that you can install. I've

120
00:11:49,520 --> 00:11:55,360
never actually successfully installed it. It's a bit of a drama. We probably should edit that bit out.

121
00:11:55,360 --> 00:12:06,320
But the, you know, that was a bit of an afterthought. So, because it was like, hey, let's turn the

122
00:12:06,320 --> 00:12:13,520
citizen devs loose. And then it was like, oh, no, we've got all these apps and all these automations

123
00:12:13,520 --> 00:12:18,720
and so many, you know, same, same, but different. You know, we've got 17 different expensive approval

124
00:12:18,720 --> 00:12:23,040
apps because Nathan has his app and it's different from Mary's, which is different from Miracles,

125
00:12:23,040 --> 00:12:30,960
which is, and it was this idea of, oh, wow, we need to govern this. Microsoft have been very early

126
00:12:30,960 --> 00:12:39,680
in the piece with AI. There's agent 365, which came out earlier this year. It's been really

127
00:12:39,680 --> 00:12:44,800
encouraging to see them. So, early to the table saying, hey, here's an inventory of your agents. Here's

128
00:12:44,800 --> 00:12:50,640
how you manage and control this. Here's how you can block agents and push them to specific users

129
00:12:50,640 --> 00:12:57,680
and have some control over what's going on with AI in your organization. I had a look at it the

130
00:12:57,680 --> 00:13:03,600
other day and post-build. There's a new shadow AI feature in there. And so it's really starting to

131
00:13:03,600 --> 00:13:13,600
help with the governance of this. And I think our role as consultants is going to increasingly move

132
00:13:13,600 --> 00:13:20,720
from being hands-on tools, which is what we've done today, right? Our job has been to go away and

133
00:13:20,720 --> 00:13:28,640
build this stuff. And as we are increasingly seeing, AI can do a lot of this stuff for us. Now,

134
00:13:28,640 --> 00:13:37,360
it's not going to 100% replace what we do as consultants today, but I think we are going to be

135
00:13:37,360 --> 00:13:42,880
in, you know, it is going to push us in an exciting direction, I think, which is more towards

136
00:13:42,880 --> 00:13:47,600
looking at things like governance, looking at things like how does this align with strategy,

137
00:13:47,600 --> 00:13:56,640
how does this align with business value and business goals? Which to me, those are the exciting

138
00:13:56,640 --> 00:14:03,120
conversations to have as opposed to, you know, did I use the right solution and that type of thing.

139
00:14:03,120 --> 00:14:09,760
So, yeah, so that's, and let me just caveat, when I say solution, I mean, the solution in

140
00:14:09,760 --> 00:14:17,920
data verse to move stuff around. So, yeah, that's, so yeah, so that's, I do see governance is

141
00:14:17,920 --> 00:14:24,000
increasingly becoming a big part of what we will be doing as consultants. And yeah, you have

142
00:14:24,000 --> 00:14:31,200
become particularly known for your work with co-pilot studio. What check you do it?

143
00:14:31,200 --> 00:14:37,920
Yeah, so as I said, I have a kind of deep history with co-pilot studio because it was,

144
00:14:38,560 --> 00:14:43,120
I was involved with the private preview of its predecessor and that kind of started everything

145
00:14:43,120 --> 00:14:53,120
for me with, with Power Platform. I did, I was a bit resistant at first because I, you know, I kind of

146
00:14:53,120 --> 00:14:59,040
came up in the low code world, the low code world kind of gave me everything. You know, I, you know,

147
00:14:59,040 --> 00:15:04,640
what it's kind of what allowed me to break out from a more unfulfilling role as a business analyst

148
00:15:04,640 --> 00:15:11,520
into consulting. It's kind of where I made my name and how I became an MVP working with a lot

149
00:15:11,520 --> 00:15:18,960
of this low code technology. And the writing was on the wall pretty early with AI that it's not clear

150
00:15:18,960 --> 00:15:26,400
that low code is going to be the way forward. And, you know, like anyone, I felt so much threatened,

151
00:15:26,400 --> 00:15:30,240
you know, because I was like, oh, no, like, you know, this stuff that I love and I'm kind of known for

152
00:15:30,240 --> 00:15:39,360
as an earth threat. And then last year at Power Platform Bootcamp, Clive Oldridge, who's one of our

153
00:15:39,360 --> 00:15:44,560
other MVPs here in New Zealand was giving a demo showing just some amazing stuff in data

154
00:15:44,560 --> 00:15:49,840
verse with using some AI tools that he had built and the penny dropped and I was like, okay,

155
00:15:49,840 --> 00:15:57,600
low code has an expiry date. And so I, you know, very quickly shifted my focus in terms of the,

156
00:15:57,600 --> 00:16:02,880
the demos I build and the content that I produced to be heavily co-pilot studio and AI focused

157
00:16:02,880 --> 00:16:09,920
just because that is the direction of travel. And it was a good choice because at Power Platform

158
00:16:09,920 --> 00:16:14,720
Conference last year, Charles LeMonna dropped the mic and said that low code is dead as we know it.

159
00:16:14,720 --> 00:16:26,240
And so yeah, that's kind of how and why I shifted into being more focused on co-pilot studio.

160
00:16:27,360 --> 00:16:35,920
We have got an in power apps. We have these, I think, how other we have the Kanba app and the

161
00:16:35,920 --> 00:16:42,240
model driven app announced some new AI code apps. Yeah, and the other product.

162
00:16:42,240 --> 00:16:46,960
What's the difference between code apps and and co-pilot studio?

163
00:16:46,960 --> 00:16:55,440
Yep. So code apps are simply applications that you are that you can build with React code. So

164
00:16:56,480 --> 00:17:03,760
they, my understanding is that it's probably going to be the replacement for Kanba Saps at some point.

165
00:17:03,760 --> 00:17:09,200
That's just my personal opinion that's nothing official. But, you know, the idea of Kanba Saps was for

166
00:17:09,200 --> 00:17:17,360
the first time we could create a custom UI for for the user and do that in a low code way.

167
00:17:17,360 --> 00:17:24,320
With code apps and the challenge we have with Kanba Saps was used Power Effects, which

168
00:17:24,320 --> 00:17:32,000
was kind of what I got known for as that abstraction layer to create the application. It basically

169
00:17:32,000 --> 00:17:37,040
lets you use those Excel type formulas. So you could use Excel. You could create an app to go

170
00:17:37,040 --> 00:17:44,480
and customize the UI and do things with data. Now, because you had that layer between the JavaScript

171
00:17:44,480 --> 00:17:51,120
that was being compiled underneath and the Power Effects, you did have some performance issues and

172
00:17:51,120 --> 00:17:58,400
other types of things that would present themselves. There were delegation type of issues when you

173
00:17:58,400 --> 00:18:02,240
were dealing with larger amounts of data and some of it was in the app and some of it you'd have to

174
00:18:02,240 --> 00:18:10,880
go off to the dataverse to get that type of thing. With code apps, we can now control the experience

175
00:18:10,880 --> 00:18:18,560
far more because it is a full fledged pro code app as opposed to this low code app using an abstraction

176
00:18:18,560 --> 00:18:24,720
layer. So it is pro code that's being written. And so as a dev, you can go and write the React app

177
00:18:24,720 --> 00:18:32,160
today and deploy it to power apps. There is also the vibe.powerapps experience where those of us

178
00:18:32,160 --> 00:18:35,920
who don't write code can just put in natural language and it will go and create all of the

179
00:18:35,920 --> 00:18:43,360
dataverse tables and that React app for you. Now, at the moment, you can see the code files for

180
00:18:43,360 --> 00:18:49,680
vibe.powerapps, you can't edit them. Somebody might need to backcheck that. Last I checked it a

181
00:18:49,680 --> 00:18:55,680
few months ago, you couldn't do it. The plan is to bring them to parity. So whether you start with

182
00:18:55,680 --> 00:19:01,760
writing code or you start with vibe.powerapps, the plan is for them to be basically the same thing.

183
00:19:01,760 --> 00:19:08,960
But it does give you a, you know, that much more custom experience. Now, how that differs from

184
00:19:08,960 --> 00:19:16,400
co-pilot studio is co-pilot studio is really meant to be an agentic experience. So the idea that

185
00:19:16,400 --> 00:19:24,480
you are either conversing with an AI agent or an AI agent is autonomously acting on a trigger to

186
00:19:24,480 --> 00:19:31,600
go and do stuff. So it's, it's a different experience than an application, whereas a user I go

187
00:19:31,600 --> 00:19:37,920
and I start interacting with things that either trigger automations or possibly agents to do things.

188
00:19:38,240 --> 00:19:44,720
Now, where this starts to get really interesting is there is an MCP server for power apps.

189
00:19:44,720 --> 00:19:53,840
And what that will do is that will surface a form, a power up form in a conversational agent.

190
00:19:53,840 --> 00:19:58,640
So that's something it's all my list of things to play with. But what gets, what's really exciting

191
00:19:58,640 --> 00:20:04,240
about that is there are times where let's say I'm working in an agentic experience,

192
00:20:05,120 --> 00:20:09,600
but I all necessarily want to have that turned my turn backwards and forth. Let's say I need to

193
00:20:09,600 --> 00:20:16,160
log a case. And the way you would log a case previously with co-pilot studio was you'd have this turn

194
00:20:16,160 --> 00:20:21,200
by turn conversation. What's your first name? What's your surname? What's your email? What's the issue?

195
00:20:21,200 --> 00:20:27,200
What's the right? And instead of having to wait for that, we can just surface up a case form.

196
00:20:27,200 --> 00:20:30,560
You know, so if you're conversing with the agent, it's like, hey, it sounds like you need to

197
00:20:31,520 --> 00:20:36,080
log a case that'll just go, hey, here's the form, fill this in. And then the agent can then take those

198
00:20:36,080 --> 00:20:45,280
parameters and work with them. So at the moment, they are distinct experiences where I think things are

199
00:20:45,280 --> 00:20:54,080
going is you are we're getting to a place where whether you are in business applications or so,

200
00:20:54,080 --> 00:20:59,840
you know, an internal user in an organization using business applications or an external user,

201
00:21:00,480 --> 00:21:05,280
you're going to have an agent experience and you're not going to care. Am I in co-pilot studio and

202
00:21:05,280 --> 00:21:09,920
my power apps, it doesn't matter, right? Because I'm just going to start from a screen and,

203
00:21:09,920 --> 00:21:15,200
you know, I might start conversing with the agent and then it's going to surface up a form and then

204
00:21:15,200 --> 00:21:22,560
it's going to go into other things and it might, you know, give me a chart if I need or, you know,

205
00:21:22,560 --> 00:21:28,320
but yeah, this this idea of, you know, this stuff kind of coming together and

206
00:21:29,600 --> 00:21:36,560
that's where I see it going and that's that's very exciting. Yeah, data was, it's a really interesting

207
00:21:36,560 --> 00:21:42,560
topic for me because I see I can choose so many databases when I use Azure, uh, uh,

208
00:21:42,560 --> 00:21:49,760
Svalve, Foundry, what is the big different or what makes data was standing out from a normal

209
00:21:49,760 --> 00:21:58,560
database? Yep. Yeah, so dataverse, what makes it so exciting is it is this common data layer

210
00:21:58,560 --> 00:22:05,040
in an organization. So prior to dataverse, when you needed to create different applications,

211
00:22:05,040 --> 00:22:11,520
let's say you had your CRM and then let's say you had, you know, for your salespeople and then let's

212
00:22:11,520 --> 00:22:16,640
say you had a marketing system and then you had a system for customer service and then maybe you

213
00:22:16,640 --> 00:22:21,520
had a system to manage assets and other things, they would all have distinct databases and,

214
00:22:22,640 --> 00:22:29,520
you know, you generally have a common pool of customers and people that work at those customers

215
00:22:29,520 --> 00:22:35,840
and then you need to integrate things because let's let's say, um, you know, somebody

216
00:22:35,840 --> 00:22:41,680
puts an order into the sales system while that needs to go to other systems to, um,

217
00:22:41,680 --> 00:22:50,960
be actioned. Um, and so you were writing integrations or you were putting in, um, complex middleware

218
00:22:50,960 --> 00:22:57,040
and things like that. Dataverse makes all of that irrelevant because you, it's basically the shared

219
00:22:57,040 --> 00:23:04,000
pool of data, which means once it's there, um, let's say, you know, we've set it up for our sales

220
00:23:04,000 --> 00:23:08,640
department and now we need to get started with customer service, well, we don't have to go and reintegrate

221
00:23:08,640 --> 00:23:15,680
because our customer data is already there. Um, and so let, you know, if I need to surface up cases

222
00:23:15,680 --> 00:23:20,240
to the salespeople so they can see what service cases have been logged, I don't have to go writing

223
00:23:20,240 --> 00:23:25,600
an integration, I just have to add that to the form and make it available to the sales users. So

224
00:23:25,600 --> 00:23:33,280
it really is that magic behind the scenes. Um, when Microsoft first introduced it, when it first

225
00:23:33,280 --> 00:23:39,680
started to kind of become a thing in the mid 2010, so it was kind of 2016, 2017, when, when I started

226
00:23:39,680 --> 00:23:49,040
to hear about it, my first response was, I think Microsoft is about to replicate in Bizzaps, what they

227
00:23:49,040 --> 00:23:56,640
did in desktop computing in the 1980s and, you know, where they basically have created the, the thing

228
00:23:56,640 --> 00:24:03,600
that everybody is going to use because it's just so easy. Um, so yeah, so, so data versus is

229
00:24:03,600 --> 00:24:07,200
incredibly special in terms of what it does. It is not just data base.

230
00:24:07,200 --> 00:24:15,280
And there's something, uh, I think you are really, you familiar with it. It's data worth business skills.

231
00:24:15,280 --> 00:24:22,080
I think it is, as as Microsoft is for quote by the agents when I'm right. Yeah. Can you explain

232
00:24:22,080 --> 00:24:28,080
this concept a little bit? Yeah. So this is a pre new feature from the dataverse team. Um, and it is,

233
00:24:28,080 --> 00:24:33,280
you can think of it. Um, so you might have started to hear a lot about skills, particularly at

234
00:24:33,280 --> 00:24:39,600
build last week. This is starting to become a thing. I am, I am saying that we in 26 is the year of skills.

235
00:24:41,280 --> 00:24:47,440
So if you have used tools like cloud code, you might be familiar with adding skills to it. Um,

236
00:24:47,440 --> 00:24:57,840
but dataverse skills is basically a low code way for you to add logic to your, um, to your agents.

237
00:24:57,840 --> 00:25:02,240
And so when you, when you create an agent, you're going to, it's going to have instructions and

238
00:25:02,240 --> 00:25:06,320
it's going to have tools like an mcp server or something like that that hooks it up to dataverse

239
00:25:06,320 --> 00:25:15,840
and let's do things. Um, now, um, let's, let's say, um, our agent works in our customer service

240
00:25:15,840 --> 00:25:21,920
department. And we've got different procedures that it follows based on, um, what needs to happen.

241
00:25:21,920 --> 00:25:26,320
And let's say we add a new procedure or we modify a procedure. Um,

242
00:25:26,320 --> 00:25:34,640
if we do this at the agent level, this is going to require us rewiring the agent adding instructions

243
00:25:34,640 --> 00:25:40,560
to it, possibly modifying tools, um, testing it and then moving it between environments. And

244
00:25:40,560 --> 00:25:44,400
there's always some level of risk when you do these things. I mean, it's, you know, we do this

245
00:25:44,400 --> 00:25:50,560
every day as consultants, but there is always risk associated with, with modifications and deployments.

246
00:25:50,560 --> 00:25:57,520
So just, there just is, um, but with business skills, these are discoverable by that mcp

247
00:25:57,520 --> 00:26:02,960
connection. So as I add and modify instructions, firstly, this can, this doesn't need to be done

248
00:26:02,960 --> 00:26:09,040
by a technical person. This can be done by the business user. So I manage during the customer

249
00:26:09,040 --> 00:26:14,000
service department can say, this is our new procedure for gold level clients or something like that.

250
00:26:14,000 --> 00:26:21,200
And they can write out instructions in business language. Um, and then the mcp server, these are

251
00:26:21,200 --> 00:26:30,160
discoverable by the mcp server. So, um, you, uh, so as our, as our procedures for the agent change,

252
00:26:30,800 --> 00:26:36,000
as we need to add new ones, we simply add them in, in data verse and we can do this in the production

253
00:26:36,000 --> 00:26:40,960
system. Now I know the ALM nerds are going to be freaking out at this point. Um, but, you know,

254
00:26:40,960 --> 00:26:45,280
we're not fundamentally changing the agent. We're just giving it different capabilities that are

255
00:26:45,280 --> 00:26:52,320
discoverable by its mcp tool set, um, which means we don't have to go modify the agent and redeploy it

256
00:26:52,320 --> 00:26:57,200
or anything like that. And so I think of it like a microservice because it's the small little bit of

257
00:26:57,200 --> 00:27:03,360
functionality that we can add, um, at the business level that extends the capability of that agent.

258
00:27:03,360 --> 00:27:08,960
So hey, we've got a new procedure for gold tiered customers or something like that. And we,

259
00:27:08,960 --> 00:27:16,720
we can go and do that without all the rigmarole of, of, of ALM. Yeah, uh, the mcp top, like I

260
00:27:16,720 --> 00:27:24,400
found it really interesting. Um, I read, uh, I don't know where it's very about the sublice. Yeah,

261
00:27:24,400 --> 00:27:33,760
mcp is a new name for API. So how will you explain mcp to someone who doesn't family always work?

262
00:27:33,760 --> 00:27:40,000
Yeah. So, um, earlier this year at Canadian Power Platform Summit, um,

263
00:27:40,000 --> 00:27:47,200
Rami Moonla Clive Aldridge and I gave a presentation on this. And the way I like to frame it is,

264
00:27:48,880 --> 00:27:56,960
an API is like Spotify. And mcp is like a DJ. Now you've probably never thought about why do we pay

265
00:27:56,960 --> 00:28:01,360
DJs when we just have Spotify to play music? It's an, it's an interesting question. I'd never

266
00:28:01,360 --> 00:28:06,480
thought of it before. But what the, what the DJ is doing is they're actually curating the experience

267
00:28:06,480 --> 00:28:11,920
because they are participating in the event with you. They are there. They are taking in all the

268
00:28:11,920 --> 00:28:16,000
context. They can see how many people are there and are they on the dance floor and enjoying

269
00:28:16,000 --> 00:28:21,200
themselves. And so they can adjust the music that they're playing accordingly, whereas

270
00:28:21,200 --> 00:28:31,760
Spotify playlist just shuffles through the songs. Um, so, um, mcp is far more than an API. It uses

271
00:28:31,760 --> 00:28:38,320
APIs, but you're basically bringing the brains of an LLM into the process and it can start to

272
00:28:38,320 --> 00:28:43,680
read the context. And so when it's triggered, it receives the prompt and this is,

273
00:28:44,640 --> 00:28:50,160
you know, this is what I'm being asked to do. Okay. Um, I'm being asked to do this. What do I know how to do?

274
00:28:50,160 --> 00:28:56,080
Okay, know how to do these things? So I think I need tool, you know, let's say I've got five tools.

275
00:28:56,080 --> 00:29:02,480
I need tools one, three and five. And so I can go and, and do the thing. And then, you know, it gets

276
00:29:02,480 --> 00:29:09,520
the next request. And it's like, oh, I need to run tool five and then tool one and then tool four

277
00:29:09,520 --> 00:29:14,560
based on what I'm, you know, and so it can, it can dynamically make these changes. Whereas APIs

278
00:29:14,560 --> 00:29:22,880
are far more brittle. They, they have a, they are, you know, they have a expected input, um, and

279
00:29:22,880 --> 00:29:29,920
to give you a defined output. So if that API is expecting a string, you would better give it a

280
00:29:29,920 --> 00:29:36,560
string. If you give it anything else, it's going to fail. Whereas mcp can determine what it's getting.

281
00:29:36,560 --> 00:29:43,280
Oh, um, you know, um, and, and it's cool to watch it learn on the fly and figure this out the way

282
00:29:43,280 --> 00:29:48,720
human would. So you pass it some, you know, you pass it something. It's like, oh, I've got, I've got

283
00:29:48,720 --> 00:29:54,080
a string, but I actually need a number. We'll hang on, they sent me 15, but it's in string format.

284
00:29:54,080 --> 00:29:58,800
Let me convert that to a number. And now I can use it with my tool. And it, and you watch it reason

285
00:29:58,800 --> 00:30:04,400
over and do these things. It's really, really cool to watch. Um, so it is, it is far more dynamic in

286
00:30:04,400 --> 00:30:10,640
terms of its capabilities. Um, yeah. So that's, that's how I would explain the difference. Yeah. I think

287
00:30:10,640 --> 00:30:19,200
that's a really good, uh, quote, which was the, the Spotify and the DJ. Um, but, uh, what role is,

288
00:30:19,200 --> 00:30:27,200
is then data was in this picture? Yep. So data verse is just your store of data. Um, so your,

289
00:30:27,200 --> 00:30:34,400
so your data is sitting in data verse, whether it's opportunities or leads, customers, um, or,

290
00:30:34,400 --> 00:30:39,520
you know, the business skills that we talked about previously. And when you connect it to an agent

291
00:30:39,520 --> 00:30:48,880
by a model context protocol, the, the exciting thing is that, um, MC, the MCP server can reason over

292
00:30:48,880 --> 00:30:55,360
your data and figure out how to solve things. So you don't have to explicitly put in all of that

293
00:30:55,360 --> 00:31:01,680
plumbing. So for example, if I, if I was just using an API, so things like agent flows, connectors,

294
00:31:01,680 --> 00:31:09,520
the traditional stuff that we've worked with, um, if I connected up to accounts, and then I ask

295
00:31:09,520 --> 00:31:13,440
it about contacts, well, it's not going to be able to do that. It's, it's going to fail. Whereas with

296
00:31:13,440 --> 00:31:20,080
MCP, it's going to say, Oh, yeah, I, I can talk to account, you know, I'm connected to accounts,

297
00:31:20,080 --> 00:31:26,240
I'm connected to contacts, I'm connected to all of that. Um, now in co pilot studio, you are, um,

298
00:31:26,240 --> 00:31:31,440
you know, it, what you can see and what you can do is governed by who you are as a user,

299
00:31:31,440 --> 00:31:39,040
and that's all baked in. So, um, which is, which is, um, um, you know, one of the kind of

300
00:31:39,040 --> 00:31:47,840
row selling points of, of the Microsoft solution, because, um, if, let's say, I am not entitled to

301
00:31:47,840 --> 00:31:52,960
look at customer records, it's not going to let me do that or it's not going to let me look at records

302
00:31:52,960 --> 00:31:58,240
that I don't have permissions to see. Maybe I'm only allowed to see my records and if I has to see yours,

303
00:31:58,240 --> 00:32:05,200
it's not going to let me do that. Um, so, but yeah, um, but yeah, MCP just has, you know, it has access to

304
00:32:05,200 --> 00:32:12,000
all the tables in Dataverse. And so it is, it, it, you know, can give you a far more dynamic, um,

305
00:32:13,040 --> 00:32:20,720
experience. Have you, uh, yeah, say a real world architecture from your, uh, from your work,

306
00:32:20,720 --> 00:32:28,240
you can walk out through without exploding with the client? Yeah. Um, so at this point, it's still

307
00:32:28,240 --> 00:32:38,160
early days. So, um, Dataverse MCP went GA, I want to say late last year. Um, so I've, I don't have

308
00:32:38,160 --> 00:32:45,200
anything in production that I've built for customers with, with Dataverse MCP. Um, some of the stuff that I

309
00:32:45,200 --> 00:32:51,280
have been building in terms of demos. Um, so one of the things that I've explored recently using

310
00:32:51,280 --> 00:32:58,640
business skills in Dataverse MCP is this idea of a CPQ or configure price course. So, um, it's something

311
00:32:58,640 --> 00:33:07,600
dynamics does not have natively, um, but it is fairly common in certain industries like telecommunications,

312
00:33:07,600 --> 00:33:14,960
specialty manufacturing where we need to configure, um, very complex quotes, um, quotes where there's

313
00:33:14,960 --> 00:33:21,120
a lot of governance. Or just to give a simple example, let's say we are building cabinets and we know

314
00:33:21,120 --> 00:33:30,240
that if we choose cabinet type X, it uses wood type Y, wood type Y requires screw type Z and screw

315
00:33:30,240 --> 00:33:38,080
type and we need paint type one. Um, and so when I use a CPQ, that's all taken care of for me. I add

316
00:33:38,080 --> 00:33:43,760
the cabinet and all of that other stuff is added, um, as opposed to having to trust the salesperson to

317
00:33:43,760 --> 00:33:48,720
make the right choice about what they add to the quote, because as we all know, salespeople tend to be

318
00:33:48,720 --> 00:33:54,880
allergic to process. Um, now CPQ systems are pretty expensive. Um, they're not cheap, the license,

319
00:33:54,880 --> 00:34:01,680
they have a lot of rules, um, um, rules that allow you to do this, this level of automation.

320
00:34:01,680 --> 00:34:06,640
So what I started playing with was, well, what if I just created instructions in data versus business

321
00:34:06,640 --> 00:34:13,680
skills and used an agent with data versus some CP to start assembling these types of things? And, um,

322
00:34:13,680 --> 00:34:20,400
it can do it. It can go and say, oh, for this, you know, you want this type of product that needs

323
00:34:20,400 --> 00:34:24,800
three of these and four of those and five of those and six of that and just goes and adds all that to

324
00:34:24,800 --> 00:34:28,800
the opportunity for you, which is really, really cool. The adds the opportunity product to the

325
00:34:28,800 --> 00:34:37,680
opportunity. So, um, yeah, we can, we can start to do some really cool stuff with, um, with data versus

326
00:34:37,680 --> 00:34:45,680
some CP, um, with our agents in terms of some of the business processes that we either needed to have

327
00:34:45,680 --> 00:34:53,920
very expensive automation for or things that, um, you know, we're just kind of one in the too hard basket

328
00:34:53,920 --> 00:35:01,760
to, um, to automate. So, for example, um, there was a government customer that we worked with here

329
00:35:01,760 --> 00:35:07,680
a number of years ago and this came out of a request that they have in case for it. So they used

330
00:35:07,680 --> 00:35:14,240
customer service to manage cases and they said, hey, we'd like to be able to mask profanity in cases.

331
00:35:14,240 --> 00:35:19,840
Sometimes people are just having a bad day, you know, they needed to complain to someone, they chose us

332
00:35:19,840 --> 00:35:26,480
and so if we could just, you know, mask profanity in the cases is, yeah, you know, who wants to read that.

333
00:35:26,480 --> 00:35:32,800
Um, and it ended up becoming a demo that I created using Copilot Studio to say, well,

334
00:35:32,800 --> 00:35:39,840
let's say 80% of your cases you could answer from knowledge on your website, right?

335
00:35:40,400 --> 00:35:46,640
on your government, you've got pre-detailed knowledge. So let's just answer that, right? Then there's a

336
00:35:46,640 --> 00:35:55,520
subset, say 15% of your cases that need to actually become a case that goes to a human. Well,

337
00:35:55,520 --> 00:36:01,280
the agent can make that determination and then there's that final 5% where somebody is just,

338
00:36:01,280 --> 00:36:05,680
they're having a bad day and they needed to complain to somebody, well, they just need to get an

339
00:36:05,680 --> 00:36:10,560
empathetic response. So why does a human need to read through that? You know, I mean, you know,

340
00:36:10,560 --> 00:36:17,200
be, you know, nobody wants to read through that all day, right? And so using, using an agent in

341
00:36:17,200 --> 00:36:26,000
Copilot Studio and MCP, you could do that. You can literally have it read incoming, in this case,

342
00:36:26,000 --> 00:36:30,400
I use emails, but you could use a portal or really anything you want it to, you know, because it's

343
00:36:30,400 --> 00:36:35,600
just language. You get language and say, what does this person want to do? Hey, do I know

344
00:36:35,600 --> 00:36:41,920
how to solve this from my, from my knowledge? No, okay. Do I think they need to create a case? Yeah, I do.

345
00:36:41,920 --> 00:36:47,120
Or no, I can't solve this from my knowledge. It's really nothing I can do anything about. So I just

346
00:36:47,120 --> 00:36:53,520
need to craft an empathetic reply and send it to them. Now, when we demoed this with the product

347
00:36:53,520 --> 00:36:59,200
group in Vegas at Power Platform conference last year, somebody in the front row hand goes up and

348
00:36:59,200 --> 00:37:03,440
they're like, hey, shouldn't a human review this first before the email gets sent? And I was like,

349
00:37:03,440 --> 00:37:09,360
yes, great point. I'm glad you asked. Don't try this at home. Absolutely human in the loop.

350
00:37:09,360 --> 00:37:20,000
But it gives you an idea of some of the stuff that we can start to automate. And again, we can have

351
00:37:20,000 --> 00:37:24,400
a human just reviewing. This is what was sent in and this is the response. Yep, that's good to go.

352
00:37:24,400 --> 00:37:29,680
So yeah, this is some of the stuff that's possible with with model context protocol.

353
00:37:30,800 --> 00:37:38,000
I really say, I think data was NCP and business skills are really new products. Are they enterprise

354
00:37:38,000 --> 00:37:47,600
ready yet? Or still beta? So business skills are still in preview. So it is, so I would say

355
00:37:47,600 --> 00:37:54,960
at the enterprise level definitely try them out in a sandbox environment. I'd be careful about

356
00:37:54,960 --> 00:38:02,480
using it in production just because as we always say with with things that aren't GA, you just,

357
00:38:02,480 --> 00:38:07,680
you want to be careful about using it in production because the feature could drastically change or

358
00:38:07,680 --> 00:38:14,880
disappear altogether without prior notice. That's that's the deal with preview. But they are definitely

359
00:38:14,880 --> 00:38:23,120
capable of handling enterprise grade tasks. So you think about, you know, assembling a bill of materials

360
00:38:23,120 --> 00:38:30,800
for a product or working out the next best action for a salesperson. It's in a lot of ways it's

361
00:38:30,800 --> 00:38:35,920
kind of like a prompt. You know, you're just giving instructions to your agent to say, okay,

362
00:38:35,920 --> 00:38:42,720
I'm working on this, you know, let's say I'm working with the salespeople and

363
00:38:42,720 --> 00:38:50,640
they've got an opportunity that has gone to this, you know, these things have happened on the

364
00:38:50,640 --> 00:38:54,640
opportunity. So I need to figure out the next best action and serve that up to the salesperson.

365
00:38:54,640 --> 00:39:00,720
So those are all things that you can do in business skills and you provide those kinds of

366
00:39:00,720 --> 00:39:08,080
instructions to the agent. So, so yeah, I would say enterprises should definitely be evaluating it

367
00:39:08,080 --> 00:39:14,960
because as it, as they progress towards GA, I think these are going to make agents just

368
00:39:14,960 --> 00:39:22,560
increasingly more scalable in the organization because again, if I don't have to go through a whole SDLC

369
00:39:22,560 --> 00:39:29,280
cycle to make changes every time, you know, because let's face in, you know, business processes change

370
00:39:29,280 --> 00:39:35,280
very quickly, which is why we often end up with all of the Excel spreadsheets and other things because

371
00:39:35,280 --> 00:39:40,880
we just IT doesn't have time to keep up. And so with things like business skills, we've got that

372
00:39:40,880 --> 00:39:47,120
microservice layer that can be edited at the business level to ensure that our agents can keep up

373
00:39:47,120 --> 00:39:56,400
with with our business. And in the case, Microsoft doesn't rename it. Where do you see the NCP in the

374
00:39:56,400 --> 00:40:10,240
next 12 or 24 months? So, MCP is, I see it, it's really exciting. So I was just chatting with a

375
00:40:10,240 --> 00:40:16,160
friend of mine the other day because I, I was very behind, I was one of the kind of the big, you know,

376
00:40:16,160 --> 00:40:23,440
cheerleaders for power FX functions, what was called low code plugins, which kind of got usurped by AI.

377
00:40:23,440 --> 00:40:30,960
So I was wrong about that, but I don't think I was wrong about MCP because you see it increasingly

378
00:40:30,960 --> 00:40:39,040
being used behind the scenes and things like work IQ and things like, um, co-work and in skills,

379
00:40:40,000 --> 00:40:47,360
all of these tools are using model context protocol behind the scenes. So it is increasingly becoming

380
00:40:47,360 --> 00:40:54,320
a big deal and a big part of how these AI tools work. So it is, it is that kind of secret sauce

381
00:40:54,320 --> 00:41:03,040
that's letting us do all kinds of stuff with, um, with our AI agents. And the best part about it is,

382
00:41:03,040 --> 00:41:07,520
we don't have to do a lot of plumbing. We literally just give it natural language and say,

383
00:41:08,320 --> 00:41:15,040
go do this thing and it can process that and go and start creating stuff in data verse and

384
00:41:15,040 --> 00:41:20,160
updating stuff in data verse for us. Whereas previously we'd have to go and configure

385
00:41:20,160 --> 00:41:26,400
something in power automate to our connector in our agent. So there'd be quite a bit of conflict,

386
00:41:26,400 --> 00:41:34,800
which we don't really have to do as much of now. That a lot of people, they, they ride something like,

387
00:41:34,800 --> 00:41:41,840
I don't know, run my entry company with agents. Yep. Okay, we have this human in the loop topic,

388
00:41:41,840 --> 00:41:49,520
but did you think, uh, organizations will eventually have dozens or even hundreds or thousands of

389
00:41:49,520 --> 00:41:55,840
specialized agents in this future? Oh, I think it's inevitable. Um, at some point, that's, that's what we

390
00:41:55,840 --> 00:42:03,920
will, we will see. Now, I am an optimist. I, you know, the doom, doom scenario is, you know, the agents

391
00:42:03,920 --> 00:42:10,400
are going to do all the jobs. We're not going to need humans anymore. We saw that from, um,

392
00:42:10,400 --> 00:42:19,120
um, Microsoft's AI CEO, um, sorry, his name is, I'm drawing a blank on his name. Um,

393
00:42:19,120 --> 00:42:25,760
Mustafa Sullivan, um, just a few weeks ago saying, you know, in 18 months, you know,

394
00:42:25,760 --> 00:42:33,520
we're not going to need humans anymore. Um, I, I don't agree with that. So now, do I think that

395
00:42:33,520 --> 00:42:41,840
AI is able to do a lot of the tasks that people do today? Yes. Um, but I believe it is going to create

396
00:42:41,840 --> 00:42:48,640
far more opportunities for us as humans to more fully express our humanity because we're not going to

397
00:42:48,640 --> 00:42:56,240
have to do the boring stuff that, um, automation has been able to do that AI will increasingly be able

398
00:42:56,240 --> 00:43:02,640
to do. Read this and summarize it. Read this and figure out what team it routes to that type of

399
00:43:02,640 --> 00:43:11,200
stuff. And that's going to free us up to spend time, um, really developing relationships,

400
00:43:11,200 --> 00:43:17,280
understanding. Um, the example that I love to give is five years ago when I bought my house. Um,

401
00:43:17,280 --> 00:43:23,360
as a first time home buyer, I had absolutely no idea what was going on, right? It's terrifying

402
00:43:23,360 --> 00:43:27,440
because you are spending the most amount of money you have ever spent in your life. Everybody

403
00:43:27,440 --> 00:43:34,560
involved in the process is rushing you along. Um, you know, your, your attorney is speaking a language

404
00:43:34,560 --> 00:43:38,800
that sounds like English, but clearly isn't because you have no idea what they said, right? They're

405
00:43:38,800 --> 00:43:44,560
getting paid a lot of money to do something that an AI agent could do. Now imagine if the AI agent

406
00:43:44,560 --> 00:43:52,480
is handling the paperwork, right? So now that lawyer, that mortgage broker, that real estate agent,

407
00:43:52,480 --> 00:43:59,680
is now freed up to spend time engaging with me, understanding me, understanding what I want,

408
00:43:59,680 --> 00:44:05,520
understanding my fears, guiding me through the process. It's, I guarantee that people would be

409
00:44:05,520 --> 00:44:10,240
willing to pay a lot more money for that than they are for, I mean, literally the other week,

410
00:44:10,240 --> 00:44:15,440
um, got a bill from my lawyer for looking at a contract for a house that we were looking to buy,

411
00:44:15,440 --> 00:44:19,520
for 300 bucks for an hour of their time. And I'm like, you got to be kidding me, right? Like,

412
00:44:20,160 --> 00:44:23,840
and, yeah, I feel like an idiot because I should have put it into co-pilot instead of having the,

413
00:44:23,840 --> 00:44:29,600
the lawyer look at it, but it's like, yeah, like, you know, you are clipping a ticket for a very low

414
00:44:29,600 --> 00:44:34,880
value task. And so I have no problem with that stuff going away, right? Because it's not creating a lot

415
00:44:34,880 --> 00:44:41,280
of value. Um, and yeah, there's going to be, you know, for people that can adapt, like, yeah, that's,

416
00:44:41,280 --> 00:44:49,280
that's not great, but look, you're not generating a lot of value. But if you are able to pivot and say,

417
00:44:49,280 --> 00:44:54,320
all right, I'm not going to focus so much on the nuts and bolts of shuffling paperwork around,

418
00:44:54,320 --> 00:45:01,200
but I am going to focus on understanding you, Nathan, as a customer, as a person, um, then you

419
00:45:01,200 --> 00:45:05,920
probably be able to charge a lot more money for that. So, so yeah, I'm incredibly optimistic about

420
00:45:05,920 --> 00:45:16,240
where it's going. Awesome. Yeah. Uh, yeah, congrats to the house. Um, and also what, what did you think,

421
00:45:16,240 --> 00:45:25,360
do's, um, yeah, I'd say leadership or, or company culture or changement and play in in these modern

422
00:45:25,360 --> 00:45:32,080
AI books? Yeah. That's a great question. Um, yeah, change management is the eternal question,

423
00:45:32,080 --> 00:45:38,800
right? We, we never seem to get it right. Um, I think, um, you know, we, you know, we try, um, but it,

424
00:45:38,800 --> 00:45:43,440
it just, I guess rule number one of life is it always could have been handled better. Um,

425
00:45:44,480 --> 00:45:53,760
I think for, I think in a lot of organizations, the challenge is really going to be,

426
00:45:53,760 --> 00:45:58,240
how do we bring everybody along on this journey? Because you're going to have people that are

427
00:45:58,240 --> 00:46:05,440
going to be very resistant and fearful, right? Take my solicitor, right? No, I don't want an AI agent

428
00:46:05,440 --> 00:46:09,440
doing that. You know, I've been doing this since longer than you've been a lot of Nathan and the agent

429
00:46:09,440 --> 00:46:14,080
can't do this, this and this. And it's like, sure. Okay. But, um, it still offers a lot more

430
00:46:14,080 --> 00:46:19,040
value. And so it's going to be, how do we bring people on this journey? How do we upskill them?

431
00:46:19,040 --> 00:46:28,160
How do we help them see, you know, the, the potential, um, and where, and how, you know, their jobs can

432
00:46:28,160 --> 00:46:34,000
open up and how they can do stuff that is far more fulfilling and creates a lot more value.

433
00:46:34,000 --> 00:46:42,480
There's also the, um, issue of how do we tackle shadow AI in the organization? Because,

434
00:46:43,280 --> 00:46:48,320
you know, we've got official tools like them, 365 co-pilot that, you know, that's what we want you to

435
00:46:48,320 --> 00:46:54,640
use. But it is, it is, you know, it's like whack them all trying to say, okay, well, how do we,

436
00:46:54,640 --> 00:47:00,720
how do we prevent somebody from going to clot or chat GPT or perplexity or one of these other tools

437
00:47:00,720 --> 00:47:07,520
in a browser or, you know, even on a mobile device? I mean, you know, it just, it is, it is

438
00:47:07,520 --> 00:47:13,760
increasingly difficult to try and lock the stuff down. And so it, it's really, you know, there,

439
00:47:13,760 --> 00:47:19,440
there needs to be this education piece around, here's how you do this stuff responsibly. And, and this is

440
00:47:19,440 --> 00:47:23,760
what it's going to enable you to do and how it's going to open your horizons as a result of doing it.

441
00:47:23,760 --> 00:47:32,000
So it's incredibly important that we get it right. Um, exactly how we do that, um, is, yeah, that's,

442
00:47:32,000 --> 00:47:39,760
that's the million dollar question. Yeah, I have also a client, uh, the first time I have talked in

443
00:47:39,760 --> 00:47:46,000
they, they will start with only, yeah, they think about starting with power platform and they have also

444
00:47:46,000 --> 00:47:52,560
the shadow, I, I, T topic. And I say, hey, we don't activate it. And I say, okay, the people use other tools.

445
00:47:52,560 --> 00:47:58,480
So it's better have the shadow, I, T, you can control in your company tenant or you have it

446
00:47:58,480 --> 00:48:03,920
as an external and the same company have also the questions about co pilot, they're saying the same thing.

447
00:48:03,920 --> 00:48:11,520
Um, yeah, I think that's, that's the topic, but there are, I think in Microsoft, there are really

448
00:48:11,520 --> 00:48:18,560
great tools like PueView and when you use it right, I think you can find a lot of shadow, I, T stuff. So

449
00:48:18,560 --> 00:48:25,440
that's, it's something you, yeah, you have to think about, but I think the risk is more high,

450
00:48:26,080 --> 00:48:34,800
if the people do we get the addict there? So it, it's, uh, I think, yeah, that's, it's a topic, um,

451
00:48:34,800 --> 00:48:43,680
yeah, companies must, must think about, um, yeah. Uh, so what, when you have, uh, one thing on

452
00:48:43,680 --> 00:48:50,720
data worse or on, on, on co pilot studio, you can say to Microsoft, they should develop a feature,

453
00:48:51,360 --> 00:48:57,760
what should it be? I'm sorry. Uh, when Microsoft, when you get got Microsoft's, it's

454
00:48:57,760 --> 00:49:05,280
nation, uh, you get, uh, you, you get all the money, the time and, uh, the resources you need,

455
00:49:05,280 --> 00:49:10,000
what feature will you develop for data worse or for, uh, co pilot studio?

456
00:49:10,000 --> 00:49:13,680
Hmm, that's a very good question. Um,

457
00:49:16,160 --> 00:49:23,760
yeah, I don't have a great, like, I don't have an amazing answer, but I would, probably

458
00:49:23,760 --> 00:49:30,000
the biggest one would be, I would love when I copy environments, I would love to be able to specify

459
00:49:30,000 --> 00:49:37,200
which tables and data verse are reference data versus which ones are transactional or operational

460
00:49:37,200 --> 00:49:42,400
data, because there are times when you're copying an environment and I don't want to bring all the

461
00:49:42,400 --> 00:49:46,560
data across, but there's some data that I need to bring across in order to use it, you know,

462
00:49:46,560 --> 00:49:52,080
master data, reference data, stuff like that. To be able to classify that when you're copying

463
00:49:52,080 --> 00:49:59,520
environments, that would be amazing. Um, that, that, so it's not a super sexy feature, but it would,

464
00:49:59,520 --> 00:50:04,880
that would make life so much easier. Yeah, the, the, the most normal, the most sexy thing.

465
00:50:06,160 --> 00:50:17,280
Um, so, um, let's jump in in the quick fire round. Uh, yeah, coffee, uh, tea or energy drain

466
00:50:17,280 --> 00:50:22,160
during development. Coffee. Power apps or power automate?

467
00:50:22,160 --> 00:50:30,560
Oh, it's a tough one. Um, can I say agent? Okay, yeah, yeah.

468
00:50:33,600 --> 00:50:38,800
Dejaunversa, GL, dataverse, makers or develop us.

469
00:50:38,800 --> 00:50:48,480
What's the difference? Oh, so, so a good answer. From engineer or traditional coding.

470
00:50:48,480 --> 00:50:54,000
From engineer. Uh, keeps our outlook.

471
00:50:56,960 --> 00:51:03,840
Neither. And the piece I'm not well-ignite. MVP summit all the way.

472
00:51:03,840 --> 00:51:12,480
Biggest productivity hack. Biggest productivity hack, um, when I said neither for teams in

473
00:51:12,480 --> 00:51:18,960
outlook, if, if we could uninvent distractions all day, that would be amazing. Um, so yeah,

474
00:51:18,960 --> 00:51:23,520
shutting down teams, shutting down outlook, um, let's be focused.

475
00:51:24,560 --> 00:51:29,280
And one product that deserves more attention. Who?

476
00:51:29,280 --> 00:51:37,360
That's a tough one. Um,

477
00:51:37,360 --> 00:51:46,240
yeah, I'm completely drawing a blank. Um,

478
00:51:49,760 --> 00:52:00,080
yeah, um, yeah, I would say, um, dataverse and dataverse um, CP because that is,

479
00:52:00,080 --> 00:52:06,480
that is powering so much of what we, what we do today is powered by that. So yeah,

480
00:52:06,480 --> 00:52:15,200
I think a lot of people are a little bit overwhelming with all these new Microsoft releases.

481
00:52:15,200 --> 00:52:22,000
How do you learn about it and what tips can you get? Yeah, that's a great question. I was actually

482
00:52:22,000 --> 00:52:29,360
talking with my team about this last week. And one of the things that I'm keen to come up with

483
00:52:29,360 --> 00:52:36,080
personally is just a framework for what do you focus on? Um, because it does, you know, it does

484
00:52:36,080 --> 00:52:41,040
change quite rapidly and, you know, we've all got limited time for learning and development.

485
00:52:41,040 --> 00:52:47,120
You want to make sure you focus on the right thing. Um, I would say, um, based on everything

486
00:52:47,120 --> 00:52:53,680
that I'm seeing, you can't go wrong with M365 Copilot at this point. That is increasingly becoming

487
00:52:53,680 --> 00:53:01,520
um, our kind of anchor during the day. It used to be outlook, then it was teams. It is increasingly

488
00:53:01,520 --> 00:53:09,360
becoming M365 Copilot. Um, and we'll probably even more so be co-work very soon. Um, that's,

489
00:53:09,360 --> 00:53:17,760
that's still a new new feature. Um, but, um, yeah, I would say you can't go wrong with M365 Copilot.

490
00:53:17,760 --> 00:53:28,880
Um, and I would say, yeah, um, otherwise, you know, do you learn MCP, do you learn skills? You know,

491
00:53:28,880 --> 00:53:34,640
it's not always a bad idea to kind of give it a bit of time to see what kind of shakes out.

492
00:53:34,640 --> 00:53:40,880
Um, generally in the MVP community, we tend to be very far out on the bleeding edge. Um, but,

493
00:53:40,880 --> 00:53:47,520
you know, um, the stuff does change a lot. So for the average, um, you know, kind of, um,

494
00:53:47,520 --> 00:53:54,720
listener at home, you know, I've got limited time, you know, just give it a few months. See,

495
00:53:54,720 --> 00:54:00,640
you know, you don't necessarily have to be the first on, on every new thing that's announced, um,

496
00:54:01,520 --> 00:54:07,680
in the AI sphere, or you can kind of wait to see, you know, what, what starts to get a bit of traction. Um,

497
00:54:07,680 --> 00:54:13,520
yeah, yeah, that's, and that's the only thing you're actually learning.

498
00:54:13,520 --> 00:54:22,960
So I, I am really focused on kind of what came out of build last week. So starting to get hands on

499
00:54:22,960 --> 00:54:30,560
with skills and putting plugins into co-work and, um, yeah, I'm pretty excited of this idea of,

500
00:54:31,520 --> 00:54:37,280
you know, um, this, really this idea of what I call headless CRM or headless CRP where, you

501
00:54:37,280 --> 00:54:44,960
know, I can start to use dataverse skills and plugins to, um, in tools like co-work. And suddenly,

502
00:54:44,960 --> 00:54:50,800
the idea of the CRM as an application starts to go away because I can start chatting with it and

503
00:54:50,800 --> 00:54:58,560
it knows to update opportunities or work with cases or whatnot based on who I am or, um,

504
00:54:58,560 --> 00:55:08,480
what context I give it. And so I think our, our days of building these, um, you know, very comprehensive

505
00:55:08,480 --> 00:55:13,200
monodriven apps may, may be coming to a close, which, which is interesting and exciting to see how

506
00:55:13,200 --> 00:55:20,080
we can do this agentically. Yeah, thank you. So then my last question is for the listeners.

507
00:55:20,080 --> 00:55:24,560
What, what shall they take from, from the conversion to, uh,

508
00:55:25,280 --> 00:55:32,320
conversion? Yeah. From the talk today, uh, quote about co-pilot studio dataverse, CPNs and the agent.

509
00:55:32,320 --> 00:55:37,600
What, what should it be when they take one thing from this session? I would say if you take one

510
00:55:37,600 --> 00:55:47,520
thing away is to be open-minded about change and getting hands on with it, um, don't, don't be

511
00:55:47,520 --> 00:55:54,160
resistant. I was a bit resistant at first, um, but this is, this is the where industry is going.

512
00:55:54,160 --> 00:56:01,600
Um, there will, you know, I think it will be the, it's like that quote from the Sun also rises by

513
00:56:01,600 --> 00:56:06,640
having way, how did you go broke, gradually, then suddenly, um, our world is going to follow a

514
00:56:06,640 --> 00:56:11,440
similar trajectory. How did you become an unemployed low-coder, gradually, then suddenly, right?

515
00:56:11,440 --> 00:56:16,480
There will be less and less of the traditional stuff that we have done, um, but there is going

516
00:56:16,480 --> 00:56:23,840
to be an exponential growth in agents and skills and understanding tools like MCP and understanding

517
00:56:23,840 --> 00:56:29,440
scout that was just announced last week and, um, Windows claw and all these different, different

518
00:56:29,440 --> 00:56:36,000
things and being able to, um, um, work with them, understand when to use them. So yeah, I would say,

519
00:56:36,000 --> 00:56:43,840
have an open mind, understand it, um, and yeah, you will, you will find that you are in demand for

520
00:56:43,840 --> 00:56:49,440
years to come. Yeah, then, uh, yeah, which thank you, Nathan Rose. This for joining us today and

521
00:56:49,440 --> 00:56:54,640
sharing all your insights about the power platform, co-pilot studio data, RIS, MCP business skills,

522
00:56:54,640 --> 00:57:02,320
and the future of the Eugenic AI with sparkles of 65. Yeah, um, and yeah, uh, yeah, this was awesome.

523
00:57:02,320 --> 00:57:09,680
So thank you very much. And yeah, one thing I can say, I hope the people, uh, see, uh, these AI

524
00:57:09,680 --> 00:57:15,600
still knows about it Jason more than, I don't know, uh, and any nice guy on, uh, the movies we have.

525
00:57:15,600 --> 00:57:22,080
And yeah, thank you so much for this. Yeah, thanks for having me, Mirko.