1
00:00:00,000 --> 00:00:03,640
Today's topic is one that almost everyone has heard of, but almost no one can actually explain.

2
00:00:03,640 --> 00:00:05,120
I'm talking about AI agents.

3
00:00:05,120 --> 00:00:08,800
You've probably heard the term in a meeting, read it in a blog post or seen it in a headline.

4
00:00:08,800 --> 00:00:10,280
Everyone's talking about them.

5
00:00:10,280 --> 00:00:13,760
But ask someone to define one, and the answers get fuzzy real fast.

6
00:00:13,760 --> 00:00:17,280
The hype is loud, the definitions are vague, and the phrase gets thrown around so loosely,

7
00:00:17,280 --> 00:00:19,000
it's lost almost all its meaning.

8
00:00:19,000 --> 00:00:20,760
Here's the most common mistake people make.

9
00:00:20,760 --> 00:00:23,400
They think an agent is just a chatbot with a fancier name.

10
00:00:23,400 --> 00:00:26,560
You ask it something, it replies, that's not an agent, that's a chatbot.

11
00:00:26,560 --> 00:00:28,720
And there is a real difference between the two.

12
00:00:28,720 --> 00:00:29,720
So what changed?

13
00:00:29,720 --> 00:00:31,880
Why is everyone suddenly obsessed with agents?

14
00:00:31,880 --> 00:00:35,120
And more importantly, why should you care if you're not a developer?

15
00:00:35,120 --> 00:00:38,960
By the end of this episode, you'll understand what an AI agent actually is,

16
00:00:38,960 --> 00:00:44,000
why Microsoft built a whole framework around them, and why this matters for the way you work.

17
00:00:44,000 --> 00:00:47,200
What exactly is an AI agent?

18
00:00:47,200 --> 00:00:50,200
You already know how a chatbot works, you type something, it responds.

19
00:00:50,200 --> 00:00:54,920
That's a conversation, it's reactive, it waits for you to say something, then it replies.

20
00:00:54,920 --> 00:00:57,240
An agent is different, it doesn't just answer.

21
00:00:57,240 --> 00:01:00,520
It acts, think, decide, act, that's the simplest definition I can give you.

22
00:01:00,520 --> 00:01:01,320
Here's how it works.

23
00:01:01,320 --> 00:01:02,800
First, the agent thinks.

24
00:01:02,800 --> 00:01:06,080
It uses a large language model to reason about what you actually need,

25
00:01:06,080 --> 00:01:07,800
not just what you typed, but what you meant.

26
00:01:07,800 --> 00:01:10,120
If you say book a meeting with Sarah next Tuesday,

27
00:01:10,120 --> 00:01:12,160
the agent doesn't just repeat that back to you,

28
00:01:12,160 --> 00:01:15,680
it thinks about what that request really means, then it decides.

29
00:01:15,680 --> 00:01:17,560
It breaks your request into steps.

30
00:01:17,560 --> 00:01:21,280
Book a meeting becomes a plan, check your calendar, find a free slot,

31
00:01:21,280 --> 00:01:24,280
check Sarah's availability, send an invite, the agent doesn't guess,

32
00:01:24,280 --> 00:01:25,760
it creates a sequence of actions.

33
00:01:25,760 --> 00:01:29,040
Finally, it acts, it calls tools and services to make it happen.

34
00:01:29,040 --> 00:01:31,360
It doesn't tell you how to book a meeting, it books the meeting,

35
00:01:31,360 --> 00:01:34,960
it doesn't explain how to check the weather, it checks the weather and gives you the answer.

36
00:01:34,960 --> 00:01:38,640
So think about the difference, a chatbot is reactive, it waits for your next question.

37
00:01:38,640 --> 00:01:42,320
An agent is proactive, you give it a goal and it goes off and works on it.

38
00:01:42,320 --> 00:01:45,000
It might come back with questions if it needs more information,

39
00:01:45,000 --> 00:01:47,560
but it doesn't need you to hold its hand through every step.

40
00:01:47,560 --> 00:01:49,920
A chatbot is like a reference desk at a library,

41
00:01:49,920 --> 00:01:52,720
you ask where a book is and they point you to the right aisle.

42
00:01:52,720 --> 00:01:56,120
An agent is like a personal assistant who has keys to the building.

43
00:01:56,120 --> 00:01:59,840
You say, "I need information about our top clients for tomorrow's meeting"

44
00:01:59,840 --> 00:02:02,920
and they go find it, organize it and hand you a briefing document.

45
00:02:02,920 --> 00:02:06,040
You didn't tell them where to look or what format to use, they figured it out.

46
00:02:06,040 --> 00:02:08,840
But the secret isn't just the LLM, the language model is smart,

47
00:02:08,840 --> 00:02:10,760
but by itself it's just a brain with no hands.

48
00:02:10,760 --> 00:02:13,680
What makes something an agent is the combination of three things.

49
00:02:13,680 --> 00:02:16,200
Reasoning, the ability to think through a problem.

50
00:02:16,200 --> 00:02:18,840
Tools, the ability to act on the world.

51
00:02:18,840 --> 00:02:22,040
And memory, the ability to remember what happened before,

52
00:02:22,040 --> 00:02:25,360
put those three together and you have something that doesn't just answer questions,

53
00:02:25,360 --> 00:02:26,760
it gets things done.

54
00:02:26,760 --> 00:02:29,080
Before the framework, the fragmented landscape.

55
00:02:29,080 --> 00:02:32,600
So if agents are so useful, why isn't everyone already building them?

56
00:02:32,600 --> 00:02:35,880
The short answer is that until recently, it was a messy process.

57
00:02:35,880 --> 00:02:38,200
You had to stitch together multiple tools yourself,

58
00:02:38,200 --> 00:02:41,280
kind of like trying to build a car by buying parts from different manufacturers

59
00:02:41,280 --> 00:02:42,880
and hoping they fit together.

60
00:02:42,880 --> 00:02:46,040
Microsoft's first attempt at fixing this was semantic kernel,

61
00:02:46,040 --> 00:02:48,640
an SDK for integrating AI into applications.

62
00:02:48,640 --> 00:02:50,840
Powerful stuff, but it was a toolbox, not a blueprint.

63
00:02:50,840 --> 00:02:54,000
It gave you the pieces you still had to figure out how to assemble them.

64
00:02:54,000 --> 00:02:56,160
Then came Autogen from Microsoft Research.

65
00:02:56,160 --> 00:02:58,800
This framework explored multi-agent conversations

66
00:02:58,800 --> 00:03:01,400
where AI agents talk, debate and collaborate.

67
00:03:01,400 --> 00:03:05,400
Cutting edge, yes, but it was experimental, not built for production systems.

68
00:03:05,400 --> 00:03:09,600
So developers had a tough choice, semantic kernel was stable and enterprise ready,

69
00:03:09,600 --> 00:03:12,000
but lacked the advanced orchestration patterns.

70
00:03:12,000 --> 00:03:14,840
Autogen had the cool multi-agent capabilities,

71
00:03:14,840 --> 00:03:17,640
but wasn't something you'd run in a live environment.

72
00:03:17,640 --> 00:03:21,120
Each had its own learning curve, its own way of doing things, its own community.

73
00:03:21,120 --> 00:03:22,560
And it wasn't just Microsoft.

74
00:03:22,560 --> 00:03:23,960
Other frameworks popped up.

75
00:03:23,960 --> 00:03:26,080
Langchain, crew AI, doesn't others.

76
00:03:26,080 --> 00:03:27,760
The ecosystem was all over the place.

77
00:03:27,760 --> 00:03:29,400
If you wanted to build agents at work,

78
00:03:29,400 --> 00:03:32,320
you had to become an expert in a rapidly changing world.

79
00:03:32,320 --> 00:03:33,560
Which framework do you learn?

80
00:03:33,560 --> 00:03:34,720
Which one will survive?

81
00:03:34,720 --> 00:03:36,560
Which one is Microsoft actually backing?

82
00:03:36,560 --> 00:03:38,160
For a beginner, it was overwhelming.

83
00:03:38,160 --> 00:03:41,440
Imagine you're a developer at a company exploring AI agents.

84
00:03:41,440 --> 00:03:43,600
You research and find five different frameworks.

85
00:03:43,600 --> 00:03:46,680
Each with its own documentation, its own patterns, its own community,

86
00:03:46,680 --> 00:03:48,040
you don't know where to start.

87
00:03:48,040 --> 00:03:49,800
You don't know which one to bet your time on.

88
00:03:49,800 --> 00:03:51,880
That fragmentation created a real barrier.

89
00:03:51,880 --> 00:03:53,360
It wasn't just about learning a tool.

90
00:03:53,360 --> 00:03:55,880
It was about betting on an entire ecosystem.

91
00:03:55,880 --> 00:03:59,760
And as agent use cases moved from experiments into real business processes,

92
00:03:59,760 --> 00:04:01,880
Microsoft realized this wasn't sustainable.

93
00:04:01,880 --> 00:04:05,640
They needed one framework, one unified approach, one answer to the question.

94
00:04:05,640 --> 00:04:07,320
How do I build an agent?

95
00:04:07,320 --> 00:04:09,600
What is Microsoft agent framework?

96
00:04:09,600 --> 00:04:11,440
So Microsoft took the two frameworks they had,

97
00:04:11,440 --> 00:04:14,440
semantic kernel and autogen, and combined them into one.

98
00:04:14,440 --> 00:04:18,920
The result is the Microsoft agent framework, a single unified SDK

99
00:04:18,920 --> 00:04:21,960
that brings together the enterprise stability of semantic kernel

100
00:04:21,960 --> 00:04:24,880
with the cutting edge multi agent patterns from autogen.

101
00:04:24,880 --> 00:04:28,200
The teams that built those two frameworks now work together on this one.

102
00:04:28,200 --> 00:04:30,920
This framework launched in preview in late 2025

103
00:04:30,920 --> 00:04:34,000
and reached version 1.0 in July, 2026.

104
00:04:34,000 --> 00:04:37,400
That timeline is important because it tells you this isn't a research project

105
00:04:37,400 --> 00:04:38,200
or an experiment.

106
00:04:38,200 --> 00:04:39,840
It's a production-ready product.

107
00:04:39,840 --> 00:04:42,480
Microsoft is saying, "Build your real systems on this."

108
00:04:42,480 --> 00:04:43,480
It's also open source.

109
00:04:43,480 --> 00:04:47,960
You can find it on GitHub with over 7,000 stars and more than 100 contributors.

110
00:04:47,960 --> 00:04:49,040
The community is active.

111
00:04:49,040 --> 00:04:52,720
There are weekly office hours where you can talk to the engineering team directly.

112
00:04:52,720 --> 00:04:54,480
That's not something you get with a closed product.

113
00:04:54,480 --> 00:04:57,880
The framework supports Python and Bennet as first-class languages,

114
00:04:57,880 --> 00:05:00,720
whether you're a CI developer or a Python developer,

115
00:05:00,720 --> 00:05:01,720
you get the same experience.

116
00:05:01,720 --> 00:05:04,760
That's a big deal because many AI frameworks are Python only.

117
00:05:04,760 --> 00:05:08,480
Microsoft made a deliberate choice to treat both languages equally.

118
00:05:08,480 --> 00:05:09,960
And here's something that might surprise you.

119
00:05:09,960 --> 00:05:12,640
The framework supports over seven different model providers.

120
00:05:12,640 --> 00:05:15,840
Microsoft Foundry, Azure OpenAI, OpenAI, Anthropic,

121
00:05:15,840 --> 00:05:19,840
Cloud, Amazon Bedrock, Google Gemini, even local models through Alama.

122
00:05:19,840 --> 00:05:23,000
You write your agent logic once and it works across all of them.

123
00:05:23,000 --> 00:05:24,480
You're not locked into one provider,

124
00:05:24,480 --> 00:05:26,400
switch from OpenAI to Anthropic,

125
00:05:26,400 --> 00:05:29,560
and you change your configuration value, not your entire code base.

126
00:05:29,560 --> 00:05:31,480
The framework has two major components.

127
00:05:31,480 --> 00:05:32,880
Agents are the building blocks,

128
00:05:32,880 --> 00:05:35,600
individual reasoning units that can think and act.

129
00:05:35,600 --> 00:05:38,680
Workflows are the orchestration layer that ties multiple agents together.

130
00:05:38,680 --> 00:05:42,600
You can have one agent doing one job or a whole team working on a complex task.

131
00:05:42,600 --> 00:05:45,520
There's also something called declarative definitions.

132
00:05:45,520 --> 00:05:47,560
You define an agent's behavior in a YAML file.

133
00:05:47,560 --> 00:05:50,280
That means your agent configuration is version controllable,

134
00:05:50,280 --> 00:05:52,520
put it in Git, review it in pull requests,

135
00:05:52,520 --> 00:05:54,520
deploy it through CI/CD pipelines.

136
00:05:54,520 --> 00:05:56,720
For enterprise teams, that's a huge advantage.

137
00:05:56,720 --> 00:05:58,360
But here's what I want you to understand.

138
00:05:58,360 --> 00:06:01,560
This framework isn't just for developers building complex systems.

139
00:06:01,560 --> 00:06:03,440
It's designed for enterprise scenarios,

140
00:06:03,440 --> 00:06:06,400
security compliance, observability, human-in-the-loop approvals,

141
00:06:06,400 --> 00:06:07,880
these are built-in, not bolted on.

142
00:06:07,880 --> 00:06:12,200
And that's what makes it different from the experimental frameworks that came before.

143
00:06:12,200 --> 00:06:15,080
The building blocks, agents, tools, memory, workflows.

144
00:06:15,080 --> 00:06:17,320
Let's break it down into the four core concepts.

145
00:06:17,320 --> 00:06:21,040
Get these right and you understand 90% of what this framework does.

146
00:06:21,040 --> 00:06:23,000
First up agents, they're the fundamental unit.

147
00:06:23,000 --> 00:06:26,400
Each agent has a name, a set of instructions and access to tools.

148
00:06:26,400 --> 00:06:28,800
It's a reasoning unit that can think and act.

149
00:06:28,800 --> 00:06:31,840
You tell it who it is, what it should do, and what resources it can use.

150
00:06:31,840 --> 00:06:33,280
Everything else flows from that.

151
00:06:33,280 --> 00:06:36,240
Next tools, these are the hands of the agent functions,

152
00:06:36,240 --> 00:06:38,760
APIs or external services the agent can call,

153
00:06:38,760 --> 00:06:42,600
weather data, calendar access, database queries, file operations.

154
00:06:42,600 --> 00:06:44,880
The agent doesn't know how to do these things by itself.

155
00:06:44,880 --> 00:06:47,880
It needs tools to reach out into the world and take action.

156
00:06:47,880 --> 00:06:49,720
You can write simple functions yourself,

157
00:06:49,720 --> 00:06:52,840
or they can be MCP servers, model context protocol.

158
00:06:52,840 --> 00:06:55,880
That's a standard way to connect agents to external services.

159
00:06:55,880 --> 00:06:57,280
Think of it as a universal plug.

160
00:06:57,280 --> 00:07:02,040
If a service supports MCP, your agent can use it without custom integration code,

161
00:07:02,040 --> 00:07:05,800
then there's memory, this is the brain's storage, two types exist.

162
00:07:05,800 --> 00:07:07,920
Session level memory is short term.

163
00:07:07,920 --> 00:07:10,480
It remembers what happened during a single conversation.

164
00:07:10,480 --> 00:07:12,120
Context providers are long term.

165
00:07:12,120 --> 00:07:15,360
They remember preferences and facts across multiple conversations.

166
00:07:15,360 --> 00:07:17,920
Memories what makes an agent feel intelligent.

167
00:07:17,920 --> 00:07:20,520
Without it, every interaction starts from scratch.

168
00:07:20,520 --> 00:07:23,920
You introduce yourself, and the next time you talk, it's forgotten.

169
00:07:23,920 --> 00:07:27,760
With memory, the agent knows who you are, what you've discussed, and what you care about.

170
00:07:27,760 --> 00:07:31,920
It's the difference between talking to a stranger and talking to a colleague who knows your work.

171
00:07:31,920 --> 00:07:33,120
Finally, workflows.

172
00:07:33,120 --> 00:07:36,680
This is the orchestration layer where you define multiple agents working together.

173
00:07:36,680 --> 00:07:38,360
Workflows can be sequential.

174
00:07:38,360 --> 00:07:39,960
Step A, then step B.

175
00:07:39,960 --> 00:07:41,200
They can be concurrent.

176
00:07:41,200 --> 00:07:44,920
Multiple agents working in parallel on different parts of the same problem.

177
00:07:44,920 --> 00:07:46,120
Or they can be conditional.

178
00:07:46,120 --> 00:07:47,560
If this happens, go here.

179
00:07:47,560 --> 00:07:49,360
If that happens, go there.

180
00:07:49,360 --> 00:07:50,960
Workflows also support checkpointing.

181
00:07:50,960 --> 00:07:54,640
If something fails halfway through, the system can resume from where it left off.

182
00:07:54,640 --> 00:07:55,760
It doesn't start over.

183
00:07:55,760 --> 00:07:59,840
That's important for long-running tasks that might take hours or involve multiple steps.

184
00:07:59,840 --> 00:08:01,880
Human and the loop patterns are built in two.

185
00:08:01,880 --> 00:08:05,640
Agents can pause and ask for approval before taking sensitive actions.

186
00:08:05,640 --> 00:08:10,160
If an agent is about to send an email or approve a payment, it can stop and wait for a human

187
00:08:10,160 --> 00:08:11,160
to say yes.

188
00:08:11,160 --> 00:08:12,640
That's essential for regulated industries.

189
00:08:12,640 --> 00:08:14,080
Here's a real-world example.

190
00:08:14,080 --> 00:08:16,320
Imagine a sales preparation agent.

191
00:08:16,320 --> 00:08:18,160
You have a meeting with a client tomorrow.

192
00:08:18,160 --> 00:08:20,400
You tell the agent, help me prepare.

193
00:08:20,400 --> 00:08:22,400
It checks your calendar to confirm the time.

194
00:08:22,400 --> 00:08:24,560
It pulls CRM data to see the client's history.

195
00:08:24,560 --> 00:08:27,640
It searches marketing materials for relevant case studies.

196
00:08:27,640 --> 00:08:30,640
It generates a briefing document and a PowerPoint presentation.

197
00:08:30,640 --> 00:08:32,840
Or without you doing anything, that's not a chatbot.

198
00:08:32,840 --> 00:08:33,840
That's an agent.

199
00:08:33,840 --> 00:08:34,840
Why should you care?

200
00:08:34,840 --> 00:08:35,840
What's practical value?

201
00:08:35,840 --> 00:08:37,400
You understand what the framework is.

202
00:08:37,400 --> 00:08:38,400
But here's the thing.

203
00:08:38,400 --> 00:08:41,280
Why should you care if you're not a developer building AI systems?

204
00:08:41,280 --> 00:08:44,600
The answer is that this framework doesn't just live in a developer's terminal.

205
00:08:44,600 --> 00:08:47,360
It powers tools that everyday business users interact with.

206
00:08:47,360 --> 00:08:51,360
Agents built with the framework can be published directly to Microsoft Teams and Copilot chat.

207
00:08:51,360 --> 00:08:54,160
That means your organization's agents live where you already work.

208
00:08:54,160 --> 00:08:57,960
You don't need a separate app to install and you log in to remember or a learning curve.

209
00:08:57,960 --> 00:09:00,000
You just open Teams and the agent is there.

210
00:09:00,000 --> 00:09:03,040
For IT and business decision makers, this matters even more.

211
00:09:03,040 --> 00:09:05,840
The framework provides enterprise controls that actually matter.

212
00:09:05,840 --> 00:09:10,080
You get identity management through Entra ID, access control, so agents only see what they're

213
00:09:10,080 --> 00:09:15,360
supposed to audit trails so you know what every agent did and when and data governance.

214
00:09:15,360 --> 00:09:17,840
So sensitive information stays protected.

215
00:09:17,840 --> 00:09:19,800
And here's a problem you might not have thought about.

216
00:09:19,800 --> 00:09:21,520
Shadow AI is real.

217
00:09:21,520 --> 00:09:26,240
Employees are using unsanctioned AI tools every day, pasting company data into public chatbots,

218
00:09:26,240 --> 00:09:29,640
uploading sensitive documents to unknown servers and hoping for the best.

219
00:09:29,640 --> 00:09:33,680
IT teams know this is happening but they can't stop it by banning tools.

220
00:09:33,680 --> 00:09:35,120
People will find a way.

221
00:09:35,120 --> 00:09:40,760
This framework gives IT a way to offer approved secure agents that still feel modern.

222
00:09:40,760 --> 00:09:45,040
Instead of telling employees don't use AI, you can say use these agents instead.

223
00:09:45,040 --> 00:09:48,720
They're safe, they're governed, and they work with your data.

224
00:09:48,720 --> 00:09:51,440
There's also the problem of agents sprawl and it's already happening.

225
00:09:51,440 --> 00:09:54,040
Teams are creating agents faster than anyone can track.

226
00:09:54,040 --> 00:09:57,920
The framework gives you a single place to manage, monitor and govern all of them.

227
00:09:57,920 --> 00:10:01,440
You can see which agents are being used, which ones are collecting dust and which ones need

228
00:10:01,440 --> 00:10:02,440
attention.

229
00:10:02,440 --> 00:10:04,040
Cost visibility is another big one.

230
00:10:04,040 --> 00:10:08,240
You can see exactly how much each agent costs to run, which agents are expensive, which

231
00:10:08,240 --> 00:10:10,520
ones are efficient and where to optimize.

232
00:10:10,520 --> 00:10:13,720
That's not something you get with a collection of random tools and scripts.

233
00:10:13,720 --> 00:10:15,040
The bottom line is this.

234
00:10:15,040 --> 00:10:19,080
This framework makes AI agents practical, safe and manageable for real organizations.

235
00:10:19,080 --> 00:10:21,840
There's not just about building cool technology.

236
00:10:21,840 --> 00:10:25,760
It's about building something that works in a real business with real constraints.

237
00:10:25,760 --> 00:10:27,640
Security, compliance, budget, governance.

238
00:10:27,640 --> 00:10:31,360
If you're responsible for any of those things, this framework matters to you.

239
00:10:31,360 --> 00:10:33,120
How it fits into the bigger picture.

240
00:10:33,120 --> 00:10:35,400
So where does the agent framework actually fit?

241
00:10:35,400 --> 00:10:36,880
It's not a standalone product.

242
00:10:36,880 --> 00:10:38,680
It's one piece of a much bigger puzzle.

243
00:10:38,680 --> 00:10:40,280
Let's look at how everything connects.

244
00:10:40,280 --> 00:10:42,160
Start with Azure AI Foundry.

245
00:10:42,160 --> 00:10:43,200
Think of it this way.

246
00:10:43,200 --> 00:10:47,640
The agent framework is how you build your agents and foundry is where they live and run.

247
00:10:47,640 --> 00:10:52,400
It handles all the heavy lifting, infrastructure, scaling, monitoring, security.

248
00:10:52,400 --> 00:10:55,960
You build locally, you deploy to Foundry and your agent runs in a production environment

249
00:10:55,960 --> 00:10:58,520
with enterprise controls, simple.

250
00:10:58,520 --> 00:11:02,280
Now if you're not a developer, you've got another option, co-pilot studio.

251
00:11:02,280 --> 00:11:03,560
This is the low-code path.

252
00:11:03,560 --> 00:11:06,200
You use a visual interface to define what your agent does,

253
00:11:06,200 --> 00:11:09,040
connect it to data sources and publish it to teams.

254
00:11:09,040 --> 00:11:10,280
No code required.

255
00:11:10,280 --> 00:11:13,040
The agent framework is for developers who want full control.

256
00:11:13,040 --> 00:11:14,480
They complement each other.

257
00:11:14,480 --> 00:11:16,480
Deep customization, use the framework.

258
00:11:16,480 --> 00:11:17,600
Need to move fast.

259
00:11:17,600 --> 00:11:19,520
Use co-pilot studio.

260
00:11:19,520 --> 00:11:23,520
Then there's Microsoft 365 co-pilot, the user facing side.

261
00:11:23,520 --> 00:11:26,720
Agents built with the framework can show up right inside co-pilot,

262
00:11:26,720 --> 00:11:29,400
so users interact with them through a familiar interface,

263
00:11:29,400 --> 00:11:31,960
the same co-pilot they already use for work.

264
00:11:31,960 --> 00:11:34,720
No new tool to learn, no new workflow to adopt.

265
00:11:34,720 --> 00:11:37,520
But here's where the real value shines, the intelligence layer.

266
00:11:37,520 --> 00:11:38,600
It has three parts.

267
00:11:38,600 --> 00:11:42,600
Work IQ connects to your personal data, email, calendar, chats.

268
00:11:42,600 --> 00:11:47,240
Foundry IQ connects to organizational knowledge, documents, policies, shared resources,

269
00:11:47,240 --> 00:11:49,600
and fabric IQ connects to business data.

270
00:11:49,600 --> 00:11:52,480
Sales numbers, customer records, operational metrics.

271
00:11:52,480 --> 00:11:54,160
Agents can pull from all three.

272
00:11:54,160 --> 00:11:55,640
That's what makes them truly useful.

273
00:11:55,640 --> 00:11:57,360
They don't just answer general questions.

274
00:11:57,360 --> 00:11:58,480
They have context.

275
00:11:58,480 --> 00:12:02,400
They know who you are, what you're working on, and what data matters to your business.

276
00:12:02,400 --> 00:12:06,120
When an agent prepares a briefing document for a sales meeting, it's not guessing.

277
00:12:06,120 --> 00:12:09,240
It's pulling from your calendar, your CRM, your marketing materials,

278
00:12:09,240 --> 00:12:10,840
and your company's knowledge base.

279
00:12:10,840 --> 00:12:11,840
All in one request.

280
00:12:11,840 --> 00:12:14,000
And it all works with Microsoft's security model.

281
00:12:14,000 --> 00:12:15,240
Enter ID for identity,

282
00:12:15,240 --> 00:12:16,640
Intune for device management,

283
00:12:16,640 --> 00:12:18,200
Defender for threat protection.

284
00:12:18,200 --> 00:12:22,040
If your organization already uses these tools, the framework fits right in.

285
00:12:22,040 --> 00:12:24,040
No need to build a separate security layer.

286
00:12:24,040 --> 00:12:27,680
For organizations already using Microsoft 365 Azure and Power Platform,

287
00:12:27,680 --> 00:12:29,640
this framework is the natural next step.

288
00:12:29,640 --> 00:12:31,000
You're not starting from scratch.

289
00:12:31,000 --> 00:12:32,600
You're extending what you already have.

290
00:12:32,600 --> 00:12:34,320
The long term vision, simple.

291
00:12:34,320 --> 00:12:36,280
Agents become as common as apps.

292
00:12:36,280 --> 00:12:39,320
Every business process gets an agent layer, not replacing people,

293
00:12:39,320 --> 00:12:43,600
but helping them work faster with better information and fewer repetitive tasks.

294
00:12:43,600 --> 00:12:45,960
This framework is how you build that layer.

295
00:12:45,960 --> 00:12:48,000
Getting started, your first step.

296
00:12:48,000 --> 00:12:49,000
Here's the good news.

297
00:12:49,000 --> 00:12:51,400
You don't need to be an AI expert to get started.

298
00:12:51,400 --> 00:12:52,400
There are two paths.

299
00:12:52,400 --> 00:12:54,400
The no-code path is co-pilot studio.

300
00:12:54,400 --> 00:12:57,720
Visual interface, drag, drop, configure, no code required.

301
00:12:57,720 --> 00:12:59,840
You can have something working in hours, not days.

302
00:12:59,840 --> 00:13:03,520
If your goal is to solve a specific business problem fast, this is your path.

303
00:13:03,520 --> 00:13:05,400
The code path is the agent framework itself.

304
00:13:05,400 --> 00:13:07,880
You install the packages for Python or Punch and Net,

305
00:13:07,880 --> 00:13:09,520
and you use the VS Code extension.

306
00:13:09,520 --> 00:13:13,440
And the extension is worth talking about because it includes AI-powered skills.

307
00:13:13,440 --> 00:13:14,760
You tell it what you want.

308
00:13:14,760 --> 00:13:18,320
Create a sales prep agent that checks my calendar and pulls CRM data,

309
00:13:18,320 --> 00:13:20,480
and it scaffolds the entire project for you.

310
00:13:20,480 --> 00:13:22,400
The code configuration files everything.

311
00:13:22,400 --> 00:13:23,400
It's not a blank page.

312
00:13:23,400 --> 00:13:24,520
It's a running start.

313
00:13:24,520 --> 00:13:27,520
The extension also includes something called the agent inspector.

314
00:13:27,520 --> 00:13:30,840
This is a visual tool that shows you what your agent is doing in real time, step by step.

315
00:13:30,840 --> 00:13:31,920
You can see it thinking.

316
00:13:31,920 --> 00:13:33,600
You can see which tools it's calling.

317
00:13:33,600 --> 00:13:34,920
You can see where it's getting stuck.

318
00:13:34,920 --> 00:13:37,080
It's like having a debugger that speaks plain English.

319
00:13:37,080 --> 00:13:40,720
If you're already using semantic kernel or autogen, there are migration assistants that

320
00:13:40,720 --> 00:13:44,560
analyze your existing code and generate a step by step migration plan.

321
00:13:44,560 --> 00:13:46,040
You don't have to start from scratch.

322
00:13:46,040 --> 00:13:47,720
The framework is open source.

323
00:13:47,720 --> 00:13:52,600
The GitHub repo is at GitHub, COM, Microsoft, agent framework, samples, documentation,

324
00:13:52,600 --> 00:13:54,160
discussion forums all there.

325
00:13:54,160 --> 00:13:55,440
The community is active.

326
00:13:55,440 --> 00:13:59,080
There are weekly office hours where you can talk to the engineering team directly.

327
00:13:59,080 --> 00:14:00,880
That's not something you get with most frameworks.

328
00:14:00,880 --> 00:14:02,160
The key message is this.

329
00:14:02,160 --> 00:14:03,320
Starts small.

330
00:14:03,320 --> 00:14:05,680
Build a single agent that does one useful thing.

331
00:14:05,680 --> 00:14:07,040
Maybe it checks your calendar.

332
00:14:07,040 --> 00:14:08,520
Maybe it summarizes your emails.

333
00:14:08,520 --> 00:14:10,080
Maybe it looks up customer information.

334
00:14:10,080 --> 00:14:11,080
See how it feels.

335
00:14:11,080 --> 00:14:12,080
Then scale up.

336
00:14:12,080 --> 00:14:14,120
The most important thing is to understand the concepts.

337
00:14:14,120 --> 00:14:15,920
And that's what this episode was about.

338
00:14:15,920 --> 00:14:18,280
Let me wrap it up by tying these pieces together.

339
00:14:18,280 --> 00:14:20,000
An agent isn't a chatbot.

340
00:14:20,000 --> 00:14:23,160
It's a system that thinks, decides, and acts on its own.

341
00:14:23,160 --> 00:14:26,960
The framework exists because building agents the old way meant juggling incompatible tools

342
00:14:26,960 --> 00:14:29,040
and frameworks that didn't talk to each other.

343
00:14:29,040 --> 00:14:30,800
That fragmentation made everything confusing.

344
00:14:30,800 --> 00:14:35,280
So Microsoft built one way to build, one way to deploy, and one way to manage.

345
00:14:35,280 --> 00:14:39,360
No more switching between frameworks or stitching together mismatched pieces.

346
00:14:39,360 --> 00:14:41,200
For beginners, here's the simple takeaway.

347
00:14:41,200 --> 00:14:42,920
AI agents aren't magic.

348
00:14:42,920 --> 00:14:47,080
They're structured systems that reason, use tools, and remember context.

349
00:14:47,080 --> 00:14:49,200
Microsoft gave you a complete framework to build them.

350
00:14:49,200 --> 00:14:50,760
And it's production ready right now.

351
00:14:50,760 --> 00:14:52,040
This isn't hype.

352
00:14:52,040 --> 00:14:56,120
Agents are becoming a normal part of how we work just like apps and websites did years ago.

353
00:14:56,120 --> 00:14:57,200
The tools are here.

354
00:14:57,200 --> 00:15:01,000
The community is active, and your first agent is closer than you think.

355
00:15:01,000 --> 00:15:05,560
If you want to explore more, start with the GitHub repo and the VS Code extension.

356
00:15:05,560 --> 00:15:08,760
Subscribe on your favorite podcast platform and share this with someone starting their