Extending Microsoft 365 Copilot: Agents, MCP & Production-Grade AI with Yves Habersaat [MVP]

1
00:00:00,000 --> 00:00:04,680
Welcome everybody to the MC65 and podcast.

2
00:00:04,680 --> 00:00:09,840
Today we are going beyond one of the most common conversations around Microsoft

3
00:00:09,840 --> 00:00:15,360
C-5 co-pilot prompt engineering because one of the organizations start using

4
00:00:15,360 --> 00:00:20,880
co-pilot seriously, the question quickly changed from how do we write better

5
00:00:20,880 --> 00:00:27,240
prompts to, how do we connect co-pilot to the applications, data, APIs, business

6
00:00:27,240 --> 00:00:31,080
processes and knowledge that actually run in our company.

7
00:00:31,080 --> 00:00:34,920
And that's where co-pilot extensibility becomes interesting.

8
00:00:34,920 --> 00:00:40,360
My guess today is Evis Hava State, Microsoft MVP Business Application Consultant,

9
00:00:40,360 --> 00:00:44,720
AI Ashley, its workgroup in Switzerland.

10
00:00:44,720 --> 00:00:50,960
Evis works across Microsoft C-65 Azure Dynamics 365 and the Power Platform.

11
00:00:50,960 --> 00:00:55,600
This particular focus on extending Microsoft C-65 co-pilot through Agents and Customs.

12
00:00:55,600 --> 00:00:56,600
The development.

13
00:00:56,600 --> 00:01:03,880
We are going to explore Microsoft 365 co-pilot extensibility, co-pilot studio, Microsoft

14
00:01:03,880 --> 00:01:04,880
Fowlery, Azure Foundry.

15
00:01:04,880 --> 00:01:10,800
I don't know, they changed the name, Microsoft Graph, Customs Development, MCPs and so on.

16
00:01:10,800 --> 00:01:14,000
So yeah, it was welcome to the MC65 and podcast.

17
00:01:14,000 --> 00:01:17,960
Yeah, hello everybody, thanks for having me, Mirko.

18
00:01:17,960 --> 00:01:24,240
It's a good opportunity for me to be here, so thanks for having me.

19
00:01:24,240 --> 00:01:26,120
Yeah, thank you.

20
00:01:26,120 --> 00:01:34,640
Before we get technical, can you tell a little bit about your journey, how you get into

21
00:01:34,640 --> 00:01:36,040
the Microsoft ecosystem?

22
00:01:36,040 --> 00:01:38,080
Yeah, yeah, yeah, of course.

23
00:01:38,080 --> 00:01:43,000
Because I started my career, not in the Microsoft world, to be honest, so I started my

24
00:01:43,000 --> 00:01:44,680
career on the Apple world.

25
00:01:44,680 --> 00:01:51,960
I was a software engineer on the Mac platform and after I moved to iOS, it's kind of

26
00:01:51,960 --> 00:02:00,320
like, and it was like a long time ago and maybe like, I don't know, 15 or 20 years ago,

27
00:02:00,320 --> 00:02:02,760
I moved to Microsoft world, to be honest.

28
00:02:02,760 --> 00:02:07,360
I started my career on the shop on the shop, on the shop, all the world.

29
00:02:07,360 --> 00:02:14,680
And after I moved to cloud Azure, M365, it was sharpened on premises, I started my career

30
00:02:14,680 --> 00:02:15,680
like this.

31
00:02:15,680 --> 00:02:21,760
And after I moved to all the cloud services and at the end, platform and more, most

32
00:02:21,760 --> 00:02:27,680
recently, all the AI stuff, to be honest, because it's now a big part of M365 now, because

33
00:02:27,680 --> 00:02:37,720
you have an adjunct teach transition now and basically, this is in an ad shell, my journey.

34
00:02:37,720 --> 00:02:44,200
But I started my career in the software engineering world, which is today, you may be a world

35
00:02:44,200 --> 00:02:50,480
that a lot of people are forgetting because you have all the AI stuff now.

36
00:02:50,480 --> 00:02:56,720
And basically, is this kind of activities or jobs have moved to AI?

37
00:02:56,720 --> 00:03:00,960
So it's quite different than 20 years ago to then as well.

38
00:03:00,960 --> 00:03:07,600
Yeah, you work, you say you work with Chapppoint and I see also you work with Graph, SPFX,

39
00:03:07,600 --> 00:03:17,400
Azure Dynamic, how did the background prepare you for the current generation of the AI development?

40
00:03:17,400 --> 00:03:26,560
And basically, I have my profile is really, it's really transversal, basically, because in

41
00:03:26,560 --> 00:03:30,640
my career, I use to work with many, many products.

42
00:03:30,640 --> 00:03:33,840
It could be Microsoft and non-Microsoft products.

43
00:03:33,840 --> 00:03:40,080
So basically, I have a large scope of skills on a lot of different technologies and tools.

44
00:03:40,080 --> 00:03:47,280
So all my jobs give me a good background because now with the AI, it's a lot of things

45
00:03:47,280 --> 00:03:54,280
it's a white topic because you have different kind of approaches, different kind of technologies

46
00:03:54,280 --> 00:03:56,120
to do AI now.

47
00:03:56,120 --> 00:04:00,240
It could be agentic, it could be more workflows, these kind of things.

48
00:04:00,240 --> 00:04:07,720
So basically, all my previous jobs give me a lot of skills to be ready for this AI transition

49
00:04:07,720 --> 00:04:09,120
now to be honest.

50
00:04:09,120 --> 00:04:13,760
So I move to local, to know, to to procre tools.

51
00:04:13,760 --> 00:04:26,480
So yes, all this opportunities and activities in my jobs give me a good good skills to take

52
00:04:26,480 --> 00:04:29,160
advantage of in the AI work to be honest.

53
00:04:29,160 --> 00:04:39,800
Yeah, and now you are and I, I actually, what do I, I actually, it's a really good question

54
00:04:39,800 --> 00:04:41,280
because a lot of people ask me.

55
00:04:41,280 --> 00:04:50,400
So my job is just to match functional specification to technical, to technical stack.

56
00:04:50,400 --> 00:04:56,120
So I'm taking the lead when the business analysis has been made.

57
00:04:56,120 --> 00:05:00,400
We have defined use case in the AI, for the AI staff.

58
00:05:00,400 --> 00:05:02,680
We have defined scenarios use case.

59
00:05:02,680 --> 00:05:08,440
We have defined, we have made the full business analysis and my job is, okay, I'm taking

60
00:05:08,440 --> 00:05:14,600
the, this needs and I'm matching the leads to the technical stack behind the scene.

61
00:05:14,600 --> 00:05:21,200
And after I'm leading all the implementation of the, of the tools and the, and the solution

62
00:05:21,200 --> 00:05:24,240
when, when we already to start developing.

63
00:05:24,240 --> 00:05:27,200
Yeah, interesting.

64
00:05:27,200 --> 00:05:38,160
I actually, I see a lot of companies by their employees trainings, especially, yeah, for

65
00:05:38,160 --> 00:05:47,360
prompt engineering and, and why will you say that not enough?

66
00:05:47,360 --> 00:05:51,240
Only prompt engineering, you mean, nowadays?

67
00:05:51,240 --> 00:06:00,720
Yeah, yeah, why, why, why, why, why it's only, you know, why is for, for companies, not

68
00:06:00,720 --> 00:06:04,360
enough, only teach the people that the prompt engineering.

69
00:06:04,360 --> 00:06:07,720
Ah, yeah, okay, I see, I see, okay, yeah, yeah.

70
00:06:07,720 --> 00:06:14,920
And basically, to be honest, nowadays you have, as I said before, you have a lot of different

71
00:06:14,920 --> 00:06:16,400
tools and technologies.

72
00:06:16,400 --> 00:06:23,680
So, and basically, at some point, yeah, I can, can understand what to do and maybe a little

73
00:06:23,680 --> 00:06:24,680
bit guide you.

74
00:06:24,680 --> 00:06:33,560
But for me, the role of, for example, architect or software designer, you need to have a

75
00:06:33,560 --> 00:06:37,440
human behind the scene because you need to understand the customer.

76
00:06:37,440 --> 00:06:42,000
Basically, AI sometimes cannot understand the context of the customer.

77
00:06:42,000 --> 00:06:43,920
You see, there's a full need.

78
00:06:43,920 --> 00:06:50,240
But behind the scene, to have just a focus on the prompt engineering is, it could be good

79
00:06:50,240 --> 00:06:56,480
at some point, but you need to have more because you need to have a behind the scene, some,

80
00:06:56,480 --> 00:07:01,400
some people that are architect software engineering behind the scene.

81
00:07:01,400 --> 00:07:07,480
And basically, a lot of people who have skills in technologies, information technologies,

82
00:07:07,480 --> 00:07:12,320
like SharePoint, like OneDrive, like, I don't know, Salesforce, this kind of stuff.

83
00:07:12,320 --> 00:07:18,320
You need to have specialists in those technologies because AI can understand a lot of things.

84
00:07:18,320 --> 00:07:25,400
You can guide you, you can generate some great technical or functional documents, but

85
00:07:25,400 --> 00:07:28,600
you need to have human behind the scene who take care of it.

86
00:07:28,600 --> 00:07:29,600
Okay.

87
00:07:29,600 --> 00:07:30,600
Is it good?

88
00:07:30,600 --> 00:07:35,480
Is it answering the customer needs, this kind of thing, to validate everything?

89
00:07:35,480 --> 00:07:42,440
And focusing on only the prompt engineering for me now is, yeah, it could be an option.

90
00:07:42,440 --> 00:07:51,760
If you don't have as a time to train or to invest in accruing some skills, but for me,

91
00:07:51,760 --> 00:07:58,280
you need to have good skills in foundation technologies where you want to build your solutions

92
00:07:58,280 --> 00:07:59,280
to be honest.

93
00:07:59,280 --> 00:08:01,320
Yeah, yeah.

94
00:08:01,320 --> 00:08:07,120
And you're an expert in co-pilot extensibility.

95
00:08:07,120 --> 00:08:16,920
Where do standard co-pilot, Microsoft, the 65 co-pilot stop and where did co-pilot extensibility

96
00:08:16,920 --> 00:08:17,920
begin?

97
00:08:17,920 --> 00:08:25,120
Yeah, it's a good question because to be honest, two years ago, when one co-pilot was released

98
00:08:25,120 --> 00:08:33,960
and maybe the first session, basically the chat was, to be honest, really minimalist.

99
00:08:33,960 --> 00:08:41,840
Basically, you can, okay, it was a chat GPT-like, but Microsoft has evolved Microsoft,

100
00:08:41,840 --> 00:08:43,400
crucified co-pilot a lot.

101
00:08:43,400 --> 00:08:47,080
Basically, now you have co-work, you have a different kind of feature.

102
00:08:47,080 --> 00:08:54,560
So now, the extensibility is less than before because in the past, you developed some

103
00:08:54,560 --> 00:08:56,280
plugin, these kind of things.

104
00:08:56,280 --> 00:09:05,200
But nowadays, to be honest, at this time, the extensibility is when you want to build something

105
00:09:05,200 --> 00:09:12,160
and you want to customize, basically, the orchestrator, or you want to personalize a little

106
00:09:12,160 --> 00:09:18,880
bit more, for example, you use your interface, if you want to integrate your agent, not only

107
00:09:18,880 --> 00:09:22,920
in Microsoft 365 chat, but for example, you know, website, these kind of things.

108
00:09:22,920 --> 00:09:29,880
So the extensibility begins when you want, basically, to customize more, behind the scene,

109
00:09:29,880 --> 00:09:31,920
you want to customize a little bit more.

110
00:09:31,920 --> 00:09:37,200
It could be orchestrator, if you want to customize your orchestrator, if you want to customize

111
00:09:37,200 --> 00:09:45,680
your interface, if you want to customize different aspects, where basically, co-pilot chat

112
00:09:45,680 --> 00:09:52,320
cannot do, cannot answer the slide.

113
00:09:52,320 --> 00:10:02,320
And I think a lot of people talk about agents.

114
00:10:02,320 --> 00:10:15,440
Why do we think so many now in co-pilot agents and why we don't build, I don't know, what

115
00:10:15,440 --> 00:10:25,440
one dec co-pilot for everything?

116
00:10:25,440 --> 00:10:26,280
Yeah, it's interesting because now, to be honest, Microsoft, as I said before, has evolved a lot.

117
00:10:26,280 --> 00:10:35,480
And maybe if we were six months before, okay, agents was a topic.

118
00:10:35,480 --> 00:10:41,680
Now Microsoft has introduced the co-pilot co-rocks and to be honest, nowadays, co-pilot co-rocks

119
00:10:41,680 --> 00:10:44,200
can replace a lot of agents.

120
00:10:44,200 --> 00:10:49,920
And while we think about agents, it's just to separate the things.

121
00:10:49,920 --> 00:10:56,760
Because now, we are more in a multi-agent model, where each agent has a scope and is

122
00:10:56,760 --> 00:10:58,000
responsibilities.

123
00:10:58,000 --> 00:11:03,960
And we don't do a big co-pilot in, basically, it's simple because of performances, first,

124
00:11:03,960 --> 00:11:10,400
because when you have a bit set of instruction and tooling in terms of performances, it will

125
00:11:10,400 --> 00:11:11,400
be a mess.

126
00:11:11,400 --> 00:11:15,880
And the architecture now is more multi-agent.

127
00:11:15,880 --> 00:11:21,800
It could be in co-pilot studio, it could be in custom agent in terms of procode or custom

128
00:11:21,800 --> 00:11:22,800
development.

129
00:11:22,800 --> 00:11:29,160
But nowadays, it's better to have a multi-agent approach where each agent has a scope and

130
00:11:29,160 --> 00:11:36,520
is responsibility in terms of just in terms of performance and to have a clear architecture

131
00:11:36,520 --> 00:11:39,480
in terms of agents.

132
00:11:39,480 --> 00:11:51,280
And when we look at this, I think, basic, I agent architecture, then we have some passwords

133
00:11:51,280 --> 00:12:02,440
that I would pass, but we have to think about reasoning, tools, knowledge, and instructions.

134
00:12:02,440 --> 00:12:09,400
How did this fit into an I agent architecture?

135
00:12:09,400 --> 00:12:18,120
The first thing to do is one of the questions you need to answer is, OK, what is the goal?

136
00:12:18,120 --> 00:12:20,920
The first objective of this is the agent.

137
00:12:20,920 --> 00:12:27,920
Basically, after agent architecture, it really depends on if you are in a local approach

138
00:12:27,920 --> 00:12:33,720
or if you are in a custom dev approach, because it's totally the approach, totally different,

139
00:12:33,720 --> 00:12:40,800
because in a custom dev approach, you manage everything, so you manage security, you manage

140
00:12:40,800 --> 00:12:45,160
the hosting, you manage a lot of different things.

141
00:12:45,160 --> 00:12:51,280
In the local approach, it's really different because you think about, OK, you give the instruction,

142
00:12:51,280 --> 00:12:52,680
you select some tools.

143
00:12:52,680 --> 00:12:59,960
So basically, the architecture in the local approach is really, it's not as simple as I

144
00:12:59,960 --> 00:13:03,680
said now, but it could be really fast because you can determine, OK, I'm going to have

145
00:13:03,680 --> 00:13:12,560
the instruction, you can select your tools, you can select a different other agent if you

146
00:13:12,560 --> 00:13:15,160
want to connect, but it's really simple.

147
00:13:15,160 --> 00:13:21,200
When you are dealing with custom dev or when you develop your own agent, it could be,

148
00:13:21,200 --> 00:13:26,600
for example, semantic kernel, agent from all these kind of things, an architect could be

149
00:13:26,600 --> 00:13:31,240
really more complex because you need to deal with concepts, really advanced concepts,

150
00:13:31,240 --> 00:13:38,320
as I said, security in terms of authentication, if you have multi, if you have for example,

151
00:13:38,320 --> 00:13:44,880
multi tools and you want to integrate some NCPs, you want to connect to internal tools

152
00:13:44,880 --> 00:13:47,960
protected by some reverse proxies, these kind of things.

153
00:13:47,960 --> 00:13:54,880
So basically, you need to reflect more in terms of architecture when you are dealing with

154
00:13:54,880 --> 00:13:55,880
custom agents.

155
00:13:55,880 --> 00:14:02,520
For example, in Copilot Studio, in terms of architecture, you have, it's really fast to

156
00:14:02,520 --> 00:14:04,160
define the architecture.

157
00:14:04,160 --> 00:14:10,840
In a custom dev approach, you have to deal with advanced concepts and for a lot of people,

158
00:14:10,840 --> 00:14:17,840
it could be difficult because it's really from that channel aspects.

159
00:14:17,840 --> 00:14:24,320
In my job, to be honest, I deal with a lot of different needs.

160
00:14:24,320 --> 00:14:29,080
It could be custom dev, it could be Copilot Studio, it could be simple agents in agent

161
00:14:29,080 --> 00:14:31,920
builder, with really simple needs.

162
00:14:31,920 --> 00:14:39,720
So basically, defining an architecture is really based on which approach you want to take

163
00:14:39,720 --> 00:14:40,720
basically.

164
00:14:40,720 --> 00:14:46,800
You have a general architecture you can make, but after it really depends on the approach

165
00:14:46,800 --> 00:14:49,200
you want to local or for code.

166
00:14:49,200 --> 00:14:50,200
You see?

167
00:14:50,200 --> 00:14:51,960
Yeah.

168
00:14:51,960 --> 00:15:02,200
And can we, can, could you describe the anatomy of a typical enterprise, AI agent?

169
00:15:02,200 --> 00:15:05,320
Oh, yes, I can.

170
00:15:05,320 --> 00:15:13,360
Basically, if you want to talk as simple as possible, when you have the needs, for example,

171
00:15:13,360 --> 00:15:18,160
it could be an IT support agent, for example.

172
00:15:18,160 --> 00:15:24,120
And basically, it could be, it could be, you have a set of things, basically, but after

173
00:15:24,120 --> 00:15:27,800
nowadays, as I said, you have this multi agent approach.

174
00:15:27,800 --> 00:15:33,760
And basically, I put, I put the first agent as the front, like a front end.

175
00:15:33,760 --> 00:15:35,440
So this is the front agent.

176
00:15:35,440 --> 00:15:43,040
And this front agent will route all the, all the requests across, for example, different

177
00:15:43,040 --> 00:15:44,040
agents.

178
00:15:44,040 --> 00:15:49,560
So, for example, if the user is asking about, I don't know, Salesforce question, for example,

179
00:15:49,560 --> 00:15:54,400
you will route all the requests to the Salesforce IT support agent.

180
00:15:54,400 --> 00:15:55,400
You see?

181
00:15:55,400 --> 00:16:00,880
So basically, for some customer, we have made this kind of architecture because each agent

182
00:16:00,880 --> 00:16:02,800
take care about the scope.

183
00:16:02,800 --> 00:16:09,520
It could be for products like Salesforce, it could be for m 365, these 10 of steps.

184
00:16:09,520 --> 00:16:15,960
And behind the scene, each agent has a set of instructions and a set of two.

185
00:16:15,960 --> 00:16:21,320
And nowadays, it's better to use MCP servers because the MCP server allows with universal plug

186
00:16:21,320 --> 00:16:27,920
for, to integrate multi services with AI.

187
00:16:27,920 --> 00:16:31,360
So I have a set of MCP server.

188
00:16:31,360 --> 00:16:35,480
And I have also some internal knowledge sources.

189
00:16:35,480 --> 00:16:40,800
It could be like GERA, it could be like service knowledge base, it kind of fix.

190
00:16:40,800 --> 00:16:44,800
So you have a set of instruction per agent.

191
00:16:44,800 --> 00:16:51,720
And this, those instruction define the scope of the agent and what is the role of each agent

192
00:16:51,720 --> 00:17:00,360
and after, you have a set of MCP server and maybe knowledge sources could be, PDF values.

193
00:17:00,360 --> 00:17:07,320
It could be multi kind of, multi kind of data sources.

194
00:17:07,320 --> 00:17:16,480
And after, when a user is requesting, for example, to the main agent after, it will define,

195
00:17:16,480 --> 00:17:23,120
okay, I need to delegate this task to the respective agent.

196
00:17:23,120 --> 00:17:31,200
So let's a little bit look into the possibilities or the co-pilot still you were, do you set fit

197
00:17:31,200 --> 00:17:35,840
into the overall architecture?

198
00:17:35,840 --> 00:17:41,880
Co-pilot Studio, to be honest, it's a local tool from Microsoft first.

199
00:17:41,880 --> 00:17:50,040
And in terms of architecture, co-pilot Studio is really, the, the, the file for, for maker,

200
00:17:50,040 --> 00:17:54,120
basically for, for people who are not developers.

201
00:17:54,120 --> 00:18:02,040
And in terms of the architecture, basically, for example, if we, if it takes the example,

202
00:18:02,040 --> 00:18:08,880
like to before the IT support, for example, co-pilot Studio will fit because in terms of architecture,

203
00:18:08,880 --> 00:18:13,680
co-pilot Studio will do the job because you can define, in some time, you can define

204
00:18:13,680 --> 00:18:19,320
MCP server, you can integrate ECD some, some external knowledge sources or internal knowledge

205
00:18:19,320 --> 00:18:21,320
sources if you want.

206
00:18:21,320 --> 00:18:26,920
After, as I said before, if for example, you have use case where you want to customize a

207
00:18:26,920 --> 00:18:32,800
little bit orchestrator because behind the scene of co-pilot Studio, you cannot customize

208
00:18:32,800 --> 00:18:33,800
the orchestrator.

209
00:18:33,800 --> 00:18:39,360
So you activate the AI orchestrator, that's all you cannot, for example, customize the

210
00:18:39,360 --> 00:18:44,800
priority of, for example, the task you want to achieve because behind the scene co-pilot

211
00:18:44,800 --> 00:18:48,400
Studio will define, okay, your plan, and after it will execute it.

212
00:18:48,400 --> 00:18:53,120
That's issue, if you want to have a little bit more customization.

213
00:18:53,120 --> 00:18:59,480
In this case, you need to use custom dev with, for example, semantic terminal, long chain,

214
00:18:59,480 --> 00:19:03,640
age of the framework most recently because Microsoft has released this framework.

215
00:19:03,640 --> 00:19:09,240
And in this case, you carry, easily customize orchestration.

216
00:19:09,240 --> 00:19:13,800
You can define some more flows, you can, you can customize what you want.

217
00:19:13,800 --> 00:19:21,680
And co-pilot Studio has an architect that is really defined for build roles and makers because

218
00:19:21,680 --> 00:19:24,200
it's really simple to have an agent working.

219
00:19:24,200 --> 00:19:29,200
The only problem with co-pilot Studio and you need to take care of this because when you

220
00:19:29,200 --> 00:19:35,040
define, when you choose your path and maybe you define your architect, you need to take care

221
00:19:35,040 --> 00:19:41,520
about the license because it's a big part of every solution now.

222
00:19:41,520 --> 00:19:48,160
Nowadays, co-pilot Studio has evolved a little bit in a pay as you will model and you need

223
00:19:48,160 --> 00:19:54,000
to take care about cost estimation because based on what you want to achieve, sometimes it's

224
00:19:54,000 --> 00:20:00,760
better to make a custom dev because after you roast where you want and you take care about

225
00:20:00,760 --> 00:20:06,040
all the costs and you can estimate more easily than now because with the co-pilot credits

226
00:20:06,040 --> 00:20:13,320
approach is sometimes difficult to predict how much, how much it costs for one month, two

227
00:20:13,320 --> 00:20:17,840
months and this kind of thing.

228
00:20:17,840 --> 00:20:29,000
And what are the limitations from co-pilot Studio when I shall choose then I think Microsoft

229
00:20:29,000 --> 00:20:38,800
or a phone, I don't know what's that really name, but when I have to say, I must do completely

230
00:20:38,800 --> 00:20:42,840
custom development.

231
00:20:42,840 --> 00:20:50,760
The first question to answer is, do I need to customize or question?

232
00:20:50,760 --> 00:20:57,640
That's the first for me, because this question will tell you, if you don't want to customize

233
00:20:57,640 --> 00:21:06,640
or question to go deep in all the or question is made or the plan is generated, you can

234
00:21:06,640 --> 00:21:09,720
use the approach with co-pilot Studio.

235
00:21:09,720 --> 00:21:15,000
For example, H&B there also, sometimes you have simple use case, you can use these kinds

236
00:21:15,000 --> 00:21:24,120
of tools, but if you want to customize the or question, you need to go to specific development

237
00:21:24,120 --> 00:21:31,360
and in this case you need to, you need to, you foundry for the model or if you have your own

238
00:21:31,360 --> 00:21:38,120
platform, because the advantage of custom dev is you want, you can integrate models that

239
00:21:38,120 --> 00:21:41,040
are not or state in Microsoft for me, if you want.

240
00:21:41,040 --> 00:21:50,520
You can use foundry models, but also if you have AWS or Google platform or Google platform

241
00:21:50,520 --> 00:21:58,420
for me, and you have a set jab now, for example, or you have deployed a model externally,

242
00:21:58,420 --> 00:22:02,880
you have the capability to integrate those models directly in your agent when you use

243
00:22:02,880 --> 00:22:04,280
custom dev approach.

244
00:22:04,280 --> 00:22:06,840
This is also an advantage.

245
00:22:06,840 --> 00:22:13,680
And after you can customize or question, all the plan is generated, all the things will

246
00:22:13,680 --> 00:22:20,320
go in the agent, foundry, the problem of foundry nowadays is at the beginning, foundry

247
00:22:20,320 --> 00:22:26,960
was just a model or state platform, when you see back a few years ago, it was like that,

248
00:22:26,960 --> 00:22:33,000
just simple model catalog and you deployed nowadays, Microsoft does evil this product to

249
00:22:33,000 --> 00:22:41,440
have like, okay, you can create agents with co-pilot Studio like user interface, basically

250
00:22:41,440 --> 00:22:49,280
without any code, you can manage governance, you can manage a lot of things, it became a

251
00:22:49,280 --> 00:22:54,720
really agente platform, like co-pilot Studio for record and this kind of things.

252
00:22:54,720 --> 00:23:03,360
And nowadays, foundry is really difficult to place in the ecosystem because foundry is

253
00:23:03,360 --> 00:23:12,920
ever fast and basically Microsoft has introduced some features that are really, I don't know,

254
00:23:12,920 --> 00:23:18,360
not different, but there is similarities with co-pilot Studio in terms of MCP integration,

255
00:23:18,360 --> 00:23:19,960
these kind of things.

256
00:23:19,960 --> 00:23:25,920
But after co-pilot Studio limitation is about the models.

257
00:23:25,920 --> 00:23:29,920
So you have, okay, you have the capability to integrate or to crawl the mistrile kind of

258
00:23:29,920 --> 00:23:30,920
things.

259
00:23:30,920 --> 00:23:36,600
So if you want to integrate, for example, your own model, because you have a specific

260
00:23:36,600 --> 00:23:39,400
graph here, your model, you cannot do.

261
00:23:39,400 --> 00:23:45,000
And basically, you need to use a custom desk to integrate your own, for example, LLM platform

262
00:23:45,000 --> 00:23:46,000
if you have.

263
00:23:46,000 --> 00:23:50,440
I have some customers who have their own AWS platform.

264
00:23:50,440 --> 00:23:59,200
In this case, you can integrate as this model with a pricing manager externally.

265
00:23:59,200 --> 00:24:09,600
Okay. And I think when we build an agent, one topic, it's also all, most people talk is

266
00:24:09,600 --> 00:24:13,960
it's MCP, the model context protocol.

267
00:24:13,960 --> 00:24:22,040
Why does it become one of the biggest topics in a gently development and for someone who

268
00:24:22,040 --> 00:24:24,800
has not worked with it yet?

269
00:24:24,800 --> 00:24:27,840
What is a model context protocol?

270
00:24:27,840 --> 00:24:30,560
Yeah, it's a big world.

271
00:24:30,560 --> 00:24:36,840
And basically, a lot of people are talking about that because it solves a major problem.

272
00:24:36,840 --> 00:24:47,280
It's all to unify the integration of tools, skills, maybe, I don't know, external tooling to

273
00:24:47,280 --> 00:24:48,280
the AI.

274
00:24:48,280 --> 00:24:53,720
And it solves this problem because in the past, without MCP, it was like a mess because a lot

275
00:24:53,720 --> 00:24:56,480
of services are talking differently.

276
00:24:56,480 --> 00:24:59,600
You have different technology protocols.

277
00:24:59,600 --> 00:25:03,320
You have different architecture approaches.

278
00:25:03,320 --> 00:25:11,000
And basically, MCP unify or you integrate services to your, it could be AI agent or AI solution.

279
00:25:11,000 --> 00:25:21,840
If I'm more generic, but MCP now is really the way to integrate external tooling or internal

280
00:25:21,840 --> 00:25:24,120
tooling, it could be the boss.

281
00:25:24,120 --> 00:25:32,640
But MCP now is a big part of AI architecture because you have this unity in terms of technology.

282
00:25:32,640 --> 00:25:39,320
And nowadays, it's a big point because in the past, it was API integration and now it's

283
00:25:39,320 --> 00:25:40,720
MCP.

284
00:25:40,720 --> 00:25:48,840
This is not the evolution, but in the past, you integrate your agent with API, you need

285
00:25:48,840 --> 00:25:54,120
to manage, okay, auto-optic, auto-manage, auto-tickle tool.

286
00:25:54,120 --> 00:25:58,040
Now the AI can, okay, I have this tool available.

287
00:25:58,040 --> 00:26:04,280
I can select what I need to achieve the goal.

288
00:26:04,280 --> 00:26:13,000
What has really changed from MCP to API connection?

289
00:26:13,000 --> 00:26:21,760
Basically, in the past, in terms of integration, basically APIs are not intelligent.

290
00:26:21,760 --> 00:26:24,360
I can technically tell that.

291
00:26:24,360 --> 00:26:31,400
APIs are just, okay, you call the tool and it gives you data.

292
00:26:31,400 --> 00:26:38,480
So the difference with the MCP is more intelligent in terms of architecture because behind

293
00:26:38,480 --> 00:26:44,880
the scene you have not only tools, who give you data or information, but you have also

294
00:26:44,880 --> 00:26:52,840
skills in instruction and different kinds of things that make MCP really intelligent.

295
00:26:52,840 --> 00:26:58,560
And basically, MCP can give the AI, okay, you want to achieve that.

296
00:26:58,560 --> 00:27:07,200
I can give you the exact tools that match your request and you can call not only one tool,

297
00:27:07,200 --> 00:27:12,520
but you need to call these tools in order to achieve what you want.

298
00:27:12,520 --> 00:27:14,280
This is a world of MCP.

299
00:27:14,280 --> 00:27:21,080
It gives the AI, okay, I have these tools and after you can call this one or this one

300
00:27:21,080 --> 00:27:24,920
or multiple tools to achieve your goals.

301
00:27:24,920 --> 00:27:30,120
If your gift AI is a context, the real context of everything.

302
00:27:30,120 --> 00:27:36,000
And API is just, okay, you have tools, you have description and that's pretty old.

303
00:27:36,000 --> 00:27:48,080
So it's the model context protocol, I think, similar to Meta data or how does it work.

304
00:27:48,080 --> 00:27:56,920
And no MCP behind the scene, it's just each protocol, basically it's a technology protocol.

305
00:27:56,920 --> 00:28:02,320
And behind the scene, you have, yes, you have metadata, you have description, you have

306
00:28:02,320 --> 00:28:07,800
AI skills that describe more, it's really focused on AI.

307
00:28:07,800 --> 00:28:12,880
Basically, you have skills, you have instructions and basically AI can understand better.

308
00:28:12,880 --> 00:28:20,360
Okay, these tools, do that, these tools do that, okay, I can call these skills, I can request

309
00:28:20,360 --> 00:28:25,560
this information to the MCP server in order to achieve my task.

310
00:28:25,560 --> 00:28:30,520
But you have, yes, you have a part of metadata because you need to define, okay, the name

311
00:28:30,520 --> 00:28:33,720
of the tool, what is doing behind the scene.

312
00:28:33,720 --> 00:28:41,400
So can you say the future is MCP, so we have to start to think about, I don't know, a company

313
00:28:41,400 --> 00:28:48,240
is a fund of rest API is inside the organization, show they start building MCP servers around them.

314
00:28:48,240 --> 00:28:57,240
Yes, yes, basically, basically, MCP server, we work with API behind the scene.

315
00:28:57,240 --> 00:29:04,600
This is not a replacement, basically, if I summarize this, because the MCP server is just,

316
00:29:04,600 --> 00:29:10,180
okay, it gives the AI the context and behind the scene, the MCP can call APIs behind the

317
00:29:10,180 --> 00:29:14,240
scene, but MCP server is a front-end.

318
00:29:14,240 --> 00:29:21,400
The front-end, the front tool, you need to call because behind the scene, you will dispatch

319
00:29:21,400 --> 00:29:26,600
your request based on, okay, behind the scene, it could be API, it could be database,

320
00:29:26,600 --> 00:29:31,920
it could be multiple tools, but you need, if you want to have your own MCP server, for example,

321
00:29:31,920 --> 00:29:40,520
if you have internally in your company, you have APIs and you have a different kind of services,

322
00:29:40,520 --> 00:29:44,080
it could be, I don't know, financial service, it could be HR service.

323
00:29:44,080 --> 00:29:49,400
If you want to build your own MCP, you need to deal behind this field with all these APIs

324
00:29:49,400 --> 00:29:58,640
and the MCP server, we will need to expose all these APIs, what are the services, what are

325
00:29:58,640 --> 00:30:05,040
the tools available, what are the skills and these kind of things.

326
00:30:05,040 --> 00:30:10,560
After you need to deal with authentication behind the scene because APIs are not working

327
00:30:10,560 --> 00:30:17,680
the same because you can have authentication by key, by entry, these kind of things, so you

328
00:30:17,680 --> 00:30:23,240
need to deal with that. But the MCP will expose the tools available and behind the scene,

329
00:30:23,240 --> 00:30:28,760
all these tools can ask you to pull APIs, one API, it could be, you have different kind

330
00:30:28,760 --> 00:30:32,640
of architecture behind the scene.

331
00:30:32,640 --> 00:30:43,880
Okay. And how do the MCP choose the right tool when I have, I don't know, 100, 500,

332
00:30:43,880 --> 00:30:52,280
I have 8 things. It will be based on some things, basically when you have a request in an

333
00:30:52,280 --> 00:31:00,280
agent, you will discover all the tools available, MCP server, you will check, okay, I have all

334
00:31:00,280 --> 00:31:08,960
these tools available and basically the orchestrator of the agent will understand, okay, I have

335
00:31:08,960 --> 00:31:15,360
this, I have an LLM available, after you will create the plan, if I summarize a little bit,

336
00:31:15,360 --> 00:31:21,720
you will create the plan with some action to do. And after in this section, in the section

337
00:31:21,720 --> 00:31:30,440
you want to execute, you can call one, two, three or more tools in order to achieve the task,

338
00:31:30,440 --> 00:31:35,840
the user and the question. So basically the orchestrator of the agent will understand

339
00:31:35,840 --> 00:31:43,320
the MCP tools and select the best for the request, you see. And this is really the orchestrator

340
00:31:43,320 --> 00:31:48,800
role of selecting tools and this kind of, there is a help of the LLM of course, because

341
00:31:48,800 --> 00:31:57,520
the LLM is as part of the brain and orchestrator will communicate with LLM and also the tools

342
00:31:57,520 --> 00:32:04,120
in order to make the best choice and the best plan in order to achieve the task.

343
00:32:04,120 --> 00:32:12,040
Yeah, I think for a lot of companies have started with PowerShell and now they use Microsoft

344
00:32:12,040 --> 00:32:23,680
Grave API, did a Grave and MCP comes or how I store the, yeah, I think the, actually, actually

345
00:32:23,680 --> 00:32:36,160
Grave API, to be honest, no, no, there is no MCP for Grave, you know, you have now in the

346
00:32:36,160 --> 00:32:43,720
Microsoft 365 world, Microsoft has introduced a lot of MCP servers and those MCP servers are

347
00:32:43,720 --> 00:32:50,440
based behind the scene on Microsoft Grave. So for example, if you want to use, I don't

348
00:32:50,440 --> 00:32:57,600
know, the team, the MCP server, behind the scene, you have the Grave API being called when

349
00:32:57,600 --> 00:33:04,520
you use the MCP server. So Grave API in the Microsoft 365 world is central because behind

350
00:33:04,520 --> 00:33:12,360
the scene, the MCP will communicate with Grave behind the scene to make the request, to make

351
00:33:12,360 --> 00:33:19,760
the request working well. But yes, nowadays Grave is central also in corporate in general

352
00:33:19,760 --> 00:33:26,520
because the semantic index of corporate is based on raw data. But actually, you know, you

353
00:33:26,520 --> 00:33:34,320
consume graphs through MCP sounds. Nowadays, it's like that.

354
00:33:34,320 --> 00:33:46,000
Microsoft 365 contains incredible value, organizational contexts, documents, meetings,

355
00:33:46,000 --> 00:33:54,720
emails, people, teams, conversations and, I say, shaple content. How do we expose that context

356
00:33:54,720 --> 00:34:01,880
without a grating, is a security nightmare?

357
00:34:01,880 --> 00:34:08,320
You mean in terms of Grave and on these kind of things? Yeah. Yeah. It's always a problem

358
00:34:08,320 --> 00:34:13,960
because a lot of people are building agents without making a proper Grave and on the

359
00:34:13,960 --> 00:34:20,240
scene. So I want to build an agent. Okay, let's go. We build and we deploy. And after that,

360
00:34:20,240 --> 00:34:27,360
okay, this agent will deal with a lot of data. How do I do my own? The first question is,

361
00:34:27,360 --> 00:34:35,120
okay, how do I put in place the agent governments before it could be like you said, the mess

362
00:34:35,120 --> 00:34:40,200
because a lot of people want to, okay, I want to build agents, agents, agents, or I want

363
00:34:40,200 --> 00:34:46,680
to build something, I want to build AI solution. But behind the scene, one of the, I sure,

364
00:34:46,680 --> 00:34:52,680
in a Microsoft 365 context, for example, one of the first things to do is, okay, to understand

365
00:34:52,680 --> 00:34:58,440
your data because without understanding your data, you are lost. You need to understand

366
00:34:58,440 --> 00:35:03,680
your data because it's important to understand, okay, am I dealing with confidential data or

367
00:35:03,680 --> 00:35:11,080
not? Am I dealing with highly confidential data? Okay, you need to understand your data.

368
00:35:11,080 --> 00:35:18,680
After, if you are in a Microsoft 365 world, you need to put in place, for example, DLP

369
00:35:18,680 --> 00:35:24,080
features, a sensitivity level in these kind of things because without this, it could be

370
00:35:24,080 --> 00:35:32,200
a problem. And after you need also to clean a little bit your data because AI, with poor,

371
00:35:32,200 --> 00:35:40,480
data, your agent won't work like you won't because if you have a mess in your, I don't

372
00:35:40,480 --> 00:35:47,120
know, a sharp point style team, team champion, this kind of thing, if it's a mess, basically

373
00:35:47,120 --> 00:35:53,920
it will look difficult to work correctly in the agentic approach, but yes, you need to

374
00:35:53,920 --> 00:36:01,720
understand your data, you need to put in place some security basics like authentication,

375
00:36:01,720 --> 00:36:10,760
like I said, for you features and for you became a central solution to put in place all

376
00:36:10,760 --> 00:36:17,960
this security and governance stuff in place. And after, when you have clean your data, when

377
00:36:17,960 --> 00:36:27,600
you have put in place some security basics, you need also to understand licensing because

378
00:36:27,600 --> 00:36:34,080
before starting to build the agent, you need to understand, okay, what are the developers?

379
00:36:34,080 --> 00:36:42,200
Are they trained for that? Are they do we have the correct licensing to start getting?

380
00:36:42,200 --> 00:36:51,000
Do we need either some suggestions or things? After, one also of the big problem is the

381
00:36:51,000 --> 00:37:00,840
twig, how do I integrate properly all the tools in my development, my agents, do I need

382
00:37:00,840 --> 00:37:08,000
to restrict some things, so it's a bit of a silly question. You have different activities

383
00:37:08,000 --> 00:37:16,640
to take in consideration, but for me the most important is to understand your data and

384
00:37:16,640 --> 00:37:24,480
if the company don't understand the data, it will be your problem to be honest.

385
00:37:24,480 --> 00:37:48,240
And when we, how did we ask the security topic, I think there are two kinds of people,

386
00:37:48,240 --> 00:37:57,480
they need user permissions and the other one says they need application permissions.

387
00:37:57,480 --> 00:38:04,480
So, general, how do you general operate?

388
00:38:04,480 --> 00:38:11,840
Generally, and my best recommendation is always basically, when you need to deal with

389
00:38:11,840 --> 00:38:19,080
the agent solution, always do in user permissions, delegated permissions, because in terms of

390
00:38:19,080 --> 00:38:26,400
security is to be honest, is better because the user will deal only with data and tools

391
00:38:26,400 --> 00:38:34,480
it has access to. That's the main principle. After, if you need application permission

392
00:38:34,480 --> 00:38:44,880
for specific, I don't know, specific API, specific context, you can, but don't use directly

393
00:38:44,880 --> 00:38:52,200
application permission, make it in a context of a wrapper, so you create, I don't know,

394
00:38:52,200 --> 00:38:59,000
it could be, it could be an MCP server, for example, and you integrate this MCP with application

395
00:38:59,000 --> 00:39:03,680
permission behind the scene, so the MCP will call, for example, I don't know you at the

396
00:39:03,680 --> 00:39:10,400
site behind the scene with application permission, and after, he gives you the information or

397
00:39:10,400 --> 00:39:18,400
the information you need, but the user don't deal directly with application permission.

398
00:39:18,400 --> 00:39:24,320
He calls, with user permission, the MCP server, for example, and the MCP server will call

399
00:39:24,320 --> 00:39:28,800
behind the scene application or services with application, with a dedicated application

400
00:39:28,800 --> 00:39:32,840
permission. You see the difference?

401
00:39:32,840 --> 00:39:42,600
Okay, interesting. I think when we think about architecture, we also have, we look a little

402
00:39:42,600 --> 00:39:50,160
bit deeper, I think there are three big topics in architecture. We have the retrieval,

403
00:39:50,160 --> 00:39:57,000
augmented generation, we have the knowledge topic, and we have the, I say, vector enterprise

404
00:39:57,000 --> 00:40:07,480
search, where does RAC fit into the architecture?

405
00:40:07,480 --> 00:40:19,800
RAC, to be honest, RAC is actually the most, to be honest, in terms of performance, because

406
00:40:19,800 --> 00:40:28,320
AI need to understand different kind of data, and basically, basically the RAC is here to

407
00:40:28,320 --> 00:40:36,920
answer a question, because in the past, when you don't have RAC, basically, you need to

408
00:40:36,920 --> 00:40:44,280
deal with different approach, we're not all school, but it was complicated, but RAC is

409
00:40:44,280 --> 00:40:53,000
central, because RAC will understand different kind of data, and you will retrieve basically

410
00:40:53,000 --> 00:41:00,880
the information that I need, documents, it could be multiple types of data. So yes, RAC

411
00:41:00,880 --> 00:41:10,600
is central nowadays in AI architecture, but it's true that for a lot of users, when you

412
00:41:10,600 --> 00:41:17,320
are talking about RAC, behind the scenes, they don't know what is it, because Microsoft nowadays

413
00:41:17,320 --> 00:41:27,120
are a lot of different software companies don't talk about this, because behind the scene,

414
00:41:27,120 --> 00:41:32,800
you have RAC, but for example, in Copa Studio, the user don't understand behind the scene,

415
00:41:32,800 --> 00:41:37,920
it's RAC, you see, and nowadays, when you want to go deep dive, okay, you have this kind

416
00:41:37,920 --> 00:41:44,760
of concept, but for a lot of users, RAC is totally, they don't know what is it, because they

417
00:41:44,760 --> 00:41:49,160
just use tools, and very behind the scene, you have those concepts.

418
00:41:49,160 --> 00:41:58,240
And then we have the next topic, it's knowledge, how we get these, where did the AI systems

419
00:41:58,240 --> 00:42:04,560
get their knowledge, is it a collect jump-oid, or how this works?

420
00:42:04,560 --> 00:42:16,320
Actually, you have to well ask you, the knowledge is, it could be roasted, a lot of different

421
00:42:16,320 --> 00:42:23,600
systems, it could be plain text file, it could be marked on file, it could be a knowledge

422
00:42:23,600 --> 00:42:29,840
based, it could be an API, but actually, if you have an MCP server, you can access those

423
00:42:29,840 --> 00:42:34,920
knowledge, or if you have, for example, I don't know, you have connectors to shoppers, you

424
00:42:34,920 --> 00:42:40,960
have connectors to many services nowadays, so accessing the knowledge is more easy than

425
00:42:40,960 --> 00:42:46,600
before, to be honest, because in the past, you need to be your plugin, and to be everything,

426
00:42:46,600 --> 00:42:52,040
nowadays, you have direct connectors to many systems, you can integrate your knowledge

427
00:42:52,040 --> 00:42:57,040
without dealing with big development stuff to be honest.

428
00:42:57,040 --> 00:43:04,280
I think, I think often we have, I don't know, I connect to my, I say to my, I don't know,

429
00:43:04,280 --> 00:43:06,560
one drive.

430
00:43:06,560 --> 00:43:17,000
And how do AI understand what, what are not already documents, or outdated, or I don't

431
00:43:17,000 --> 00:43:25,480
know, duplicated, how can, how understand this, an AI system, what it should use, but not

432
00:43:25,480 --> 00:43:26,480
it.

433
00:43:26,480 --> 00:43:37,240
Basically, you are using Ragn, because behind the scene, you need to understand, okay, what

434
00:43:37,240 --> 00:43:46,080
is your data, but to be honest, if I summarize a little bit, you start from the prompt,

435
00:43:46,080 --> 00:43:52,000
okay, the user wants to, okay, he wants to, for example, to search for us, I don't know,

436
00:43:52,000 --> 00:43:59,520
knowledge base, for example, one drive, behind the scene, when, when the, what the orchestrator

437
00:43:59,520 --> 00:44:08,880
with the help of LLM understands the prompt and the, the user intents, behind the scene,

438
00:44:08,880 --> 00:44:13,720
you have the index, basically, because for example, in Copilot, you have the semantic index,

439
00:44:13,720 --> 00:44:21,920
and this semantic index give you, basically, okay, I have all this knowledge, and

440
00:44:21,920 --> 00:44:28,960
you can, with the semantic index query, the AI can query directly, the semantic index to

441
00:44:28,960 --> 00:44:35,880
perform, to perform a Ragn, for example, and you can directly, okay, select the right

442
00:44:35,880 --> 00:44:41,840
documents based on the needs, but in Copilot, you have the semantic index, and the AI can

443
00:44:41,840 --> 00:44:48,120
query directly, this index, with natural language behind the scene, so, yeah, we will ask,

444
00:44:48,120 --> 00:44:55,320
okay, I want to understand, I don't know, how, how to set up my company VPN, and behind

445
00:44:55,320 --> 00:45:02,560
the scene, you will perform Ragn on the semantic index, and basically retrieve the documents

446
00:45:02,560 --> 00:45:08,160
that contain, for example, all the setup of the company VPN, for example.

447
00:45:08,160 --> 00:45:14,400
In Microsoft, to see if that Copilot chat, behind the scene, it's graph API, because as

448
00:45:14,400 --> 00:45:19,160
I said before, you have the semantic index built on graph API, it's graph feature behind

449
00:45:19,160 --> 00:45:26,760
the scene, but when you have custom development, you have, for example, a FoundryAQ behind

450
00:45:26,760 --> 00:45:31,880
the scene, it was an Azure services, with a different name, but now you can have your

451
00:45:31,880 --> 00:45:37,160
own RAC system, if you want, and basically, you can, when you have a custom agent, you

452
00:45:37,160 --> 00:45:46,880
can directly ask this FoundryAQ instance, okay, I want to use, I want to get the setup of

453
00:45:46,880 --> 00:45:54,280
my VPN company, and the AI can retrieve what in-in terms of documents, I could be here,

454
00:45:54,280 --> 00:46:09,600
so this kind of fix. And that's how the search works, right?

455
00:46:09,600 --> 00:46:13,840
Or behind the scene, you mean?

456
00:46:13,840 --> 00:46:14,840
Yeah.

457
00:46:14,840 --> 00:46:23,240
Ah, behind the scene, yes, you have the search, but basically, nowadays, you have behind

458
00:46:23,240 --> 00:46:30,360
the scene, behind the scene of Copilot chat, you have three capabilities, you have the

459
00:46:30,360 --> 00:46:37,080
standard search, which is basically a search, that's okay, you simply put some keywords,

460
00:46:37,080 --> 00:46:42,080
and you will search across your documents, and this kind of thing, this is a simple search,

461
00:46:42,080 --> 00:46:50,360
or we all know, a keyword search, like Google, after you have two different capabilities,

462
00:46:50,360 --> 00:46:58,120
you have first the RAC behind the scene, you have the capability of RAC in graph, and you

463
00:46:58,120 --> 00:47:05,320
have also the Copilot semantic index search, which is a little bit different, because RAC

464
00:47:05,320 --> 00:47:11,920
is really more advanced than the Copilot semantic in that search, but you have all these

465
00:47:11,920 --> 00:47:17,840
three capabilities behind the scene in order to make your, it's not true, but behind the

466
00:47:17,840 --> 00:47:25,720
scene, I can call these three different capabilities to retrieve the data across one drive teams,

467
00:47:25,720 --> 00:47:27,520
M365 in general.

468
00:47:27,520 --> 00:47:33,120
After, you can, the semantic index of Copilot include also the Copilot connectors.

469
00:47:33,120 --> 00:47:40,120
So, for example, if you have connected Salesforce with a Copilot connector, you can also search

470
00:47:40,120 --> 00:47:47,240
across all your data in Salesforce in service now in multiple tools, because it's everything

471
00:47:47,240 --> 00:47:52,840
with the Copilot connector, you have integrated all the data in the graph index, semantic index,

472
00:47:52,840 --> 00:47:57,480
so you can query directly your external systems.

473
00:47:57,480 --> 00:48:09,880
Okay, and I think when we build agents, we have the governance part, and one of the part

474
00:48:09,880 --> 00:48:15,400
is also for me, I say, the agent lifecycle management.

475
00:48:15,400 --> 00:48:18,520
How does this look actually?

476
00:48:18,520 --> 00:48:26,200
It looks actually, it's to be honest, in the Microsoft ecosystem, it's pretty simple,

477
00:48:26,200 --> 00:48:32,640
if I can say that, because when you are dealing with Lockout, for example, Copilot Studio,

478
00:48:32,640 --> 00:48:39,360
you are dealing with poor platform content, I-L-M, and nothing as change, so you have solution,

479
00:48:39,360 --> 00:48:44,480
you have A-L-M in place, you have pipeline, so when you are dealing with Copilot Studio,

480
00:48:44,480 --> 00:48:49,480
to be honest, it's like before, because you have all the stuff you invested to in terms

481
00:48:49,480 --> 00:48:55,560
of corporate from governance, lifecycle, A-L-M, it's strictly the same.

482
00:48:55,560 --> 00:49:00,800
This is a good thing, because a lot of people are asking me, "Oh, Copilot Studio, all do

483
00:49:00,800 --> 00:49:02,280
manage, A-L-M, is it different?"

484
00:49:02,280 --> 00:49:08,520
No, it's really simple, like, perhaps, or co-automate, application lifecycle management,

485
00:49:08,520 --> 00:49:14,080
so in this case, it's agent lifecycle management, but behind the scene, same concept,

486
00:49:14,080 --> 00:49:22,400
like, corporate form, and when you are dealing with, for example, Custom Dev, basically,

487
00:49:22,400 --> 00:49:28,080
it's like when you are building applications, but not applications kind of stuff, because

488
00:49:28,080 --> 00:49:35,080
you manage everything like code, so you have frameworks, you have security aspects, this

489
00:49:35,080 --> 00:49:41,240
kind of thing, so everything to be honest, in terms of agent lifecycle management, it's

490
00:49:41,240 --> 00:49:46,720
really the same concept than before, and this is a good point for me, because you don't

491
00:49:46,720 --> 00:49:55,320
have to invest a lot in terms of lifecycle process, this kind of things, but only you have

492
00:49:55,320 --> 00:50:01,640
some change, but one of the big changes, "Okay, you need to take care of that, A-L-M, this

493
00:50:01,640 --> 00:50:06,800
kind of stuff, the deployment, this kind of things," but behind the scene, it looks like,

494
00:50:06,800 --> 00:50:10,840
in the past, DevOps approach with pipeline, this kind of thing, so...

495
00:50:10,840 --> 00:50:18,880
In my opinion, nothing has changed because we manage agents, it could be Custom Dev, or

496
00:50:18,880 --> 00:50:25,400
capacity, with the same approach than before, to be honest, nowadays.

497
00:50:25,400 --> 00:50:36,360
And this means, when a company, they don't use before AI, but they have the perfect setup

498
00:50:36,360 --> 00:50:42,000
for PerView, for Defender, and all the tools.

499
00:50:42,000 --> 00:50:50,200
Is there any change that it would do, or is it this means AI readiness?

500
00:50:50,200 --> 00:51:01,800
Ah, it's a big question, because AI readiness for Microsoft is okay, are you a sharpening

501
00:51:01,800 --> 00:51:09,280
commission, manage, this kind of things, but... okay, AI readiness for me is not only a technical

502
00:51:09,280 --> 00:51:15,160
approach, because, okay, technical, we know what to do, okay, we need to manage a commission,

503
00:51:15,160 --> 00:51:22,840
a process, not a service, you know, for me, AI readiness must include also the data, the

504
00:51:22,840 --> 00:51:30,920
data, understand you, because when you don't understand your data, as I said before, you cannot

505
00:51:30,920 --> 00:51:36,680
manage the technical part, because when you don't know what is confidential or not, for

506
00:51:36,680 --> 00:51:46,240
example, it's difficult to put in place some security stuff, but if we follow Microsoft,

507
00:51:46,240 --> 00:51:51,640
AI readiness is okay, you want to manage commission, you want to put sensitivity, the level

508
00:51:51,640 --> 00:51:57,560
in place, you want to make GLP, this kind of thing.

509
00:51:57,560 --> 00:52:07,960
But you have also a more high level aspect, and for me, AI readiness is not only technical,

510
00:52:07,960 --> 00:52:14,120
you have also a functional or high level aspect to take your incarceration to be honest,

511
00:52:14,120 --> 00:52:15,120
nowadays.

512
00:52:15,120 --> 00:52:16,120
Hmm.

513
00:52:16,120 --> 00:52:21,560
Yeah, that's really interesting.

514
00:52:21,560 --> 00:52:27,480
I think a lot of companies, as well start at the AI, but now it was agents, a lot of companies

515
00:52:27,480 --> 00:52:39,080
say, "Oh, it's deleted all my, I don't know, emails or doing something evil stuff."

516
00:52:39,080 --> 00:52:40,560
How did you think?

517
00:52:40,560 --> 00:52:42,680
What will the company say?

518
00:52:42,680 --> 00:52:44,680
They fear AI?

519
00:52:44,680 --> 00:52:51,480
It's a good question, because I'm living in Switzerland, and basically we have a lot of private

520
00:52:51,480 --> 00:52:59,560
banks, and we have a lot of, to be honest, really sensitive customers for, yeah, to be honest.

521
00:52:59,560 --> 00:53:07,360
And basically, the most critical part is, okay, where the data is processed from, basically.

522
00:53:07,360 --> 00:53:16,080
And yes, also, okay, is AI searching across my emails, this kind of thing, so behind the scene,

523
00:53:16,080 --> 00:53:23,520
they always ask this, but the most critical part for them is, "Okay, I need to understand

524
00:53:23,520 --> 00:53:34,640
where the data is processed, and also, okay, I am the owner of the AI."

525
00:53:34,640 --> 00:53:41,640
So, I put the instruction to the AI, and the AI behind the scene will do the job.

526
00:53:41,640 --> 00:53:49,000
And they are afraid of, for example, the autonomous aspect, because nowadays, okay, you can do

527
00:53:49,000 --> 00:53:55,880
that, and behind the scene, they will search for something, query something behind the scene,

528
00:53:55,880 --> 00:54:01,800
and they are afraid of, "Okay, when I ask the AI to do something, what is behind the scene,

529
00:54:01,800 --> 00:54:02,800
apart from?"

530
00:54:02,800 --> 00:54:10,800
And yes, there is a lot of questions like this, "Oh, do I take the control of the food process

531
00:54:10,800 --> 00:54:15,000
behind the scene of, it could be an AI agent of an AI service?"

532
00:54:15,000 --> 00:54:24,880
Yes, there is some questions like this, but it's really a question of adoption, also, training

533
00:54:24,880 --> 00:54:30,520
to put in place some training, but, "Okay, all to understand AI, all to understand behind

534
00:54:30,520 --> 00:54:36,520
the scene, what is performed exactly, and transparency is also a bit important, because

535
00:54:36,520 --> 00:54:43,600
transparency needs to be put in place in every AI solution, because transparency is for

536
00:54:43,600 --> 00:54:50,600
me a big topic, because you need to put in place all everything, because your AI solution

537
00:54:50,600 --> 00:54:51,800
needs to be transparent.

538
00:54:51,800 --> 00:54:57,520
You need to understand what is performed behind the scene, what data is query, you need

539
00:54:57,520 --> 00:55:01,680
to trace, you need to log everything, so everything needs to be transparent, when you have

540
00:55:01,680 --> 00:55:09,120
a solution nowadays, because you need to understand exactly what is performed behind, and

541
00:55:09,120 --> 00:55:13,480
what data is processed, and what is the reasoning behind the scene.

542
00:55:13,480 --> 00:55:19,960
"Okay, how do you have generated this problem based on which facts and these kind of things?"

543
00:55:19,960 --> 00:55:20,960
Awesome.

544
00:55:20,960 --> 00:55:24,440
I see, we run a little bit out of time, so...

545
00:55:24,440 --> 00:55:25,440
Yeah, so...

546
00:55:25,440 --> 00:55:31,600
No, this was really amazing, but I have an every podcast, I have a rapid fire rounds,

547
00:55:31,600 --> 00:55:32,600
so...

548
00:55:32,600 --> 00:55:33,600
Yeah.

549
00:55:33,600 --> 00:55:40,480
So, I give a short sentence that you give a short answer, so are you ready?

550
00:55:40,480 --> 00:55:41,480
Yes.

551
00:55:41,480 --> 00:55:44,320
"Copilot Studio" or "Custom Code"?

552
00:55:44,320 --> 00:55:46,320
"Copilot Studio".

553
00:55:46,320 --> 00:55:49,880
"Microsoft Fowlery" or "Copilot Studio"?

554
00:55:49,880 --> 00:55:51,880
"Copilot Studio".

555
00:55:51,880 --> 00:55:53,880
"Typescript" or "C#".

556
00:55:53,880 --> 00:55:54,880
"Typescript".

557
00:55:54,880 --> 00:55:58,880
"Whip-A-Woe-Age-O-Deff-Ops".

558
00:55:58,880 --> 00:56:00,880
"Git-A-M".

559
00:56:00,880 --> 00:56:03,320
"Rack-O-Fine-Tuning".

560
00:56:03,320 --> 00:56:06,200
"Rack-O-Fine-Tuning".

561
00:56:06,200 --> 00:56:15,040
What's the most under-rapeable Microsoft 365 API?

562
00:56:15,040 --> 00:56:18,040
"Rack-O-F-E-I".

563
00:56:18,040 --> 00:56:23,400
One technology every Microsoft developer should learn in 2006.

564
00:56:23,400 --> 00:56:27,520
"Git-A-P-Copilot".

565
00:56:27,520 --> 00:56:33,560
Then, Sajan Idalya comes to you and says, "Give us, you get all the money and resources you

566
00:56:33,560 --> 00:56:37,120
want to make our AI platform better.

567
00:56:37,120 --> 00:56:38,320
You can develop one feature.

568
00:56:38,320 --> 00:56:40,560
What will it be?"

569
00:56:40,560 --> 00:56:45,200
All the governance stuff we need.

570
00:56:45,200 --> 00:56:46,200
And...

571
00:56:46,200 --> 00:56:52,800
Who shall I invite next to the podcast and what questions shall I ask?

572
00:56:52,800 --> 00:56:59,680
Oh, good question.

573
00:56:59,680 --> 00:57:02,040
I have a lot of friends.

574
00:57:02,040 --> 00:57:05,400
Yes, maybe you can ask to...

575
00:57:05,400 --> 00:57:07,360
I have a course speaker in many confronts.

576
00:57:07,360 --> 00:57:10,520
You can ask for example, Chloe Morro.

577
00:57:10,520 --> 00:57:16,440
She was a functional consultant before dealing with the technical side now.

578
00:57:16,440 --> 00:57:22,360
And she has a really good story about functional, technical stuff.

579
00:57:22,360 --> 00:57:30,360
And the Dynamics 365 world, so maybe you can invite her and ask, "Okay, tell me nowadays

580
00:57:30,360 --> 00:57:36,360
how to deal between functional and technical world because you have some big challenges

581
00:57:36,360 --> 00:57:37,360
there."

582
00:57:37,360 --> 00:57:44,240
Yeah, and my closing question is, what should people, when they can take one running from

583
00:57:44,240 --> 00:57:48,760
this session, what should they take?

584
00:57:48,760 --> 00:57:53,960
From this session, you mean?

585
00:57:53,960 --> 00:57:56,560
Basically, AI is evolving fast.

586
00:57:56,560 --> 00:58:04,400
And if you want to be up to date nowadays, you need to follow everything because nowadays,

587
00:58:04,400 --> 00:58:09,080
AI is really across all the Microsoft touch to the nest.

588
00:58:09,080 --> 00:58:16,560
And nowadays, one of the things you need to understand is the evolution class.

589
00:58:16,560 --> 00:58:23,120
And you need to be up to date really, really at any time because you will miss something

590
00:58:23,120 --> 00:58:24,120
at some point.

591
00:58:24,120 --> 00:58:26,120
You don't learn.

592
00:58:26,120 --> 00:58:30,400
You don't understand the evolutionary challenges.

593
00:58:30,400 --> 00:58:33,600
Yeah, so, yeah.

594
00:58:33,600 --> 00:58:35,200
But thank you for joining me.

595
00:58:35,200 --> 00:58:42,200
I think the important takeaway from this conversation is that Microsoft 365, co-pilot extensibility,

596
00:58:42,200 --> 00:58:47,800
isn't simply about writing, increasingly means of decayed prompts.

597
00:58:47,800 --> 00:58:55,440
I think we are moving towards architecture where co-pilot and agents can connect, organization,

598
00:58:55,440 --> 00:59:00,240
knowledge, Microsoft Graph, business application APIs, and so on.

599
00:59:00,240 --> 00:59:05,240
Yeah, I think co-pilot studio Microsoft finally custom developed and technically such as

600
00:59:05,240 --> 00:59:13,040
MCP, give developers an amazing, powerful toolbox.

601
00:59:13,040 --> 00:59:21,440
And yeah, but more capable agents become the more important from the mental spectrum as

602
00:59:21,440 --> 00:59:27,960
well, identity, permission, security, governance, testing, observability, application lifecycle

603
00:59:27,960 --> 00:59:29,560
management, and so on.

604
00:59:29,560 --> 00:59:35,000
So yeah, I think you have really, really good overview of all these topics.

605
00:59:35,000 --> 00:59:42,100
So thank you for joining me and for the listeners, all the information about if you find in

606
00:59:42,100 --> 00:59:45,800
the show notes on the MC65FM podcast patch.

607
00:59:45,800 --> 00:59:49,160
And yeah, well, you can connect, see the work of it.

608
00:59:49,160 --> 00:59:50,800
And yeah, thank you for being here.

609
00:59:50,800 --> 00:59:52,640
Thanks for having me.

610
00:59:52,640 --> 00:59:54,360
It was a really good time with you.

611
00:59:54,360 --> 01:00:04,360
[BLANK_AUDIO]