1
00:00:00,000 --> 00:00:04,120
Hello everyone and welcome to another episode of Microsoft Knowledge Nuggets here on M365.

2
00:00:04,120 --> 00:00:06,160
FM, I'm your host, Mirko Peters.

3
00:00:06,160 --> 00:00:09,800
Today's topic is one that almost everyone has heard of, but few people truly understand

4
00:00:09,800 --> 00:00:10,800
AI agents.

5
00:00:10,800 --> 00:00:15,240
What exactly is an AI agent and do you really need to be a developer to build one?

6
00:00:15,240 --> 00:00:20,320
Those are fair questions, especially when you hear terms like declarative agent and orchestration thrown around.

7
00:00:20,320 --> 00:00:22,680
Here's the short answer, no, you don't need to code.

8
00:00:22,680 --> 00:00:26,240
By the end of this episode, you'll understand what Microsoft agent builder is, how it works,

9
00:00:26,240 --> 00:00:29,800
and exactly when to use it versus co-pilot studio or AI Foundry.

10
00:00:29,800 --> 00:00:32,080
Grab your coffee and let's dive in.

11
00:00:32,080 --> 00:00:37,040
What agent builder actually is, Microsoft agent builder is a no-code tool that lives inside

12
00:00:37,040 --> 00:00:39,160
Microsoft 365 co-pilot.

13
00:00:39,160 --> 00:00:42,400
It lets you create what Microsoft calls declarative agents.

14
00:00:42,400 --> 00:00:46,240
In plain English, that means you design a personal AI assistant just by describing what

15
00:00:46,240 --> 00:00:48,720
you want, you don't write code, you write sentences.

16
00:00:48,720 --> 00:00:50,360
So how does it actually work?

17
00:00:50,360 --> 00:00:54,360
Open co-pilot in your browser, look at the sidebar on the left and you'll see a section called

18
00:00:54,360 --> 00:00:55,360
agents.

19
00:00:55,360 --> 00:00:58,120
Click New Agent and you get two tabs at the top.

20
00:00:58,120 --> 00:00:59,520
Describe and configure.

21
00:00:59,520 --> 00:01:01,360
The Describe tab is the magic one.

22
00:01:01,360 --> 00:01:05,920
You type something like "I need an agent that helps onboard new team members" and co-pilot

23
00:01:05,920 --> 00:01:10,640
builds the agent for you, generating a name, description, instructions and even suggested

24
00:01:10,640 --> 00:01:11,640
prompts.

25
00:01:11,640 --> 00:01:12,720
You talk, it builds.

26
00:01:12,720 --> 00:01:15,240
The configure tab is for tweaking things manually.

27
00:01:15,240 --> 00:01:19,240
Change the name, upload an icon, rewrite the instructions, add knowledge sources.

28
00:01:19,240 --> 00:01:23,040
Most people start with describe, then switch to configure to polish things up.

29
00:01:23,040 --> 00:01:24,840
Think of describe as the fast lane.

30
00:01:24,840 --> 00:01:28,120
You just tell it what you need and the agent appears in seconds.

31
00:01:28,120 --> 00:01:31,760
The user is for fine tuning, like adjusting the furniture after the room is built.

32
00:01:31,760 --> 00:01:35,240
But here's the thing, agent builder is not a separate website.

33
00:01:35,240 --> 00:01:38,680
You don't go to co-pilot studio, Microsoft.com or some other portal.

34
00:01:38,680 --> 00:01:42,200
It lives right inside the co-pilot sidebar in the same place you chat with co-pilot every

35
00:01:42,200 --> 00:01:43,960
day and that's by design.

36
00:01:43,960 --> 00:01:47,360
Microsoft wants this to feel like a natural extension of your workflow, not a separate tool

37
00:01:47,360 --> 00:01:48,360
you have to learn.

38
00:01:48,360 --> 00:01:52,320
Most users never notice the complexity behind the scenes, but that's the beauty of it.

39
00:01:52,320 --> 00:01:53,600
Think of it like an office building.

40
00:01:53,600 --> 00:01:56,720
Imagine you work in a big building with rooms for different things.

41
00:01:56,720 --> 00:01:58,680
Building rooms, storage rooms, break rooms.

42
00:01:58,680 --> 00:02:02,360
In the old days, if you needed a new room, you'd have to hire an architect, draw up plans,

43
00:02:02,360 --> 00:02:04,360
get permits and wait weeks for construction.

44
00:02:04,360 --> 00:02:06,920
That's what building an AI agent used to feel like.

45
00:02:06,920 --> 00:02:08,320
Agent builder is the opposite.

46
00:02:08,320 --> 00:02:11,840
It's like walking up to the building manager and saying, "I need a small room on the third

47
00:02:11,840 --> 00:02:13,880
floor with a whiteboard and good lighting."

48
00:02:13,880 --> 00:02:16,280
The manager says, "Got it, and it's ready in minutes.

49
00:02:16,280 --> 00:02:18,240
You don't design the plumbing or the wiring.

50
00:02:18,240 --> 00:02:21,720
You just describe the room you need and the building handles the rest."

51
00:02:21,720 --> 00:02:23,040
That's the core idea.

52
00:02:23,040 --> 00:02:26,960
Agent builder takes your description and turns it into a working agent, handling all the

53
00:02:26,960 --> 00:02:31,640
technical stuff behind the scenes, the AI model, the grounding, the permissions.

54
00:02:31,640 --> 00:02:33,520
You just tell it what you want.

55
00:02:33,520 --> 00:02:34,760
The creation flow.

56
00:02:34,760 --> 00:02:35,760
Step by step.

57
00:02:35,760 --> 00:02:37,680
What does this actually look like in practice?

58
00:02:37,680 --> 00:02:38,680
Let's walk through it.

59
00:02:38,680 --> 00:02:40,520
You start from the co-pilot sidebar.

60
00:02:40,520 --> 00:02:42,680
Click New Agent and you see two tabs.

61
00:02:42,680 --> 00:02:44,280
Describe and configure.

62
00:02:44,280 --> 00:02:45,880
Most people land on Describe first.

63
00:02:45,880 --> 00:02:47,120
That's where the magic happens.

64
00:02:47,120 --> 00:02:49,480
In the Describe tab, you type what you need.

65
00:02:49,480 --> 00:02:53,120
Like, "I need an agent that helps onboard new team members."

66
00:02:53,120 --> 00:02:54,760
Hit Send and co-pilot starts working.

67
00:02:54,760 --> 00:02:58,880
It thinks for a few seconds, then comes back with a complete agent, including a name, description,

68
00:02:58,880 --> 00:03:02,000
instructions, and suggested prompts users can click when they start chatting.

69
00:03:02,000 --> 00:03:04,440
It's all generated from that one sentence you typed.

70
00:03:04,440 --> 00:03:08,120
You can accept it as is or switch to configure and tweak things manually.

71
00:03:08,120 --> 00:03:10,040
The configure tab is where you get hands-on.

72
00:03:10,040 --> 00:03:14,000
You can change the agent's name, upload a custom icon, or rewrite the description.

73
00:03:14,000 --> 00:03:16,360
The instructions field is the most important part.

74
00:03:16,360 --> 00:03:18,720
This is where you tell the agent how to behave.

75
00:03:18,720 --> 00:03:20,720
And something like "You are an HR assistant."

76
00:03:20,720 --> 00:03:23,040
Only answer questions using the provided documents.

77
00:03:23,040 --> 00:03:26,200
If you don't know the answer, tell the user to contact HR directly.

78
00:03:26,200 --> 00:03:29,360
That's your instruction and the agent follows it.

79
00:03:29,360 --> 00:03:30,840
Then there's the Knowledge section.

80
00:03:30,840 --> 00:03:33,520
This is where you connect the agent to actual data.

81
00:03:33,520 --> 00:03:36,760
Without knowledge sources, the agent is just a chatbot with a personality.

82
00:03:36,760 --> 00:03:38,960
It doesn't know anything about your company.

83
00:03:38,960 --> 00:03:43,960
So you add sources, SharePoint sites, specific files from one drive, upload a document from

84
00:03:43,960 --> 00:03:48,320
your computer, public URLs, Teams chats, or Outlook emails.

85
00:03:48,320 --> 00:03:52,760
The agent reads those sources and answers questions based on what it finds.

86
00:03:52,760 --> 00:03:55,200
Below that, there's a capabilities section with toggles.

87
00:03:55,200 --> 00:03:59,040
You can let the agent generate documents, write code, or create images.

88
00:03:59,040 --> 00:04:02,400
Flip those on if you want the agent to produce word docs or PowerPoint slides.

89
00:04:02,400 --> 00:04:04,840
They're off by default, but easy to enable.

90
00:04:04,840 --> 00:04:06,600
Then there are suggested prompts.

91
00:04:06,600 --> 00:04:09,400
Pre-written questions that show up when someone opens the agent.

92
00:04:09,400 --> 00:04:13,920
Things like "What are the steps for new hire paperwork?" or find the travel policy document.

93
00:04:13,920 --> 00:04:17,600
You can write your own or let co-pilot generate them based on your description.

94
00:04:17,600 --> 00:04:19,760
And everything looks good, click "Create".

95
00:04:19,760 --> 00:04:24,280
The agent appears in your co-pilot sidebar alongside any other agents you've built or installed,

96
00:04:24,280 --> 00:04:26,160
and you can start chatting with it immediately.

97
00:04:26,160 --> 00:04:29,360
If you need to change something later, just click the three dots next to the agent and

98
00:04:29,360 --> 00:04:30,520
choose "Edit".

99
00:04:30,520 --> 00:04:34,080
It opens the "Configure" tab again, and you can tweak anything.

100
00:04:34,080 --> 00:04:39,240
Instructions, knowledge sources, prompts, update it, and the changes take effect right away.

101
00:04:39,240 --> 00:04:40,680
Where the smarts come from?

102
00:04:40,680 --> 00:04:41,680
Knowledge sources.

103
00:04:41,680 --> 00:04:45,440
Now you've got a working agent, but where does its intelligence actually come from?

104
00:04:45,440 --> 00:04:46,440
Here's the thing.

105
00:04:46,440 --> 00:04:48,120
A agent doesn't know anything by default.

106
00:04:48,120 --> 00:04:49,920
It's like a new employee on their first day.

107
00:04:49,920 --> 00:04:53,280
They might be smart, but they don't know your company's policies, your products, or your

108
00:04:53,280 --> 00:04:54,280
customers.

109
00:04:54,280 --> 00:04:58,240
They need training materials and for an agent, those training materials are called "knowledge

110
00:04:58,240 --> 00:04:59,240
sources".

111
00:04:59,240 --> 00:05:03,040
Without grounding data, your agent is just a generic chatbot with a personality.

112
00:05:03,040 --> 00:05:06,960
It can hold a conversation, but won't know anything specific about your business.

113
00:05:06,960 --> 00:05:10,520
That's why the knowledge section is the most important part of building an agent.

114
00:05:10,520 --> 00:05:13,000
You have several types of sources you can connect.

115
00:05:13,000 --> 00:05:14,520
SharePoint sites are the most common.

116
00:05:14,520 --> 00:05:16,920
Think of them as your company's document libraries.

117
00:05:16,920 --> 00:05:21,040
One drive files work too if you have a specific document stored in your personal cloud.

118
00:05:21,040 --> 00:05:23,440
You can upload files directly from your computer.

119
00:05:23,440 --> 00:05:28,440
PDFs, Word documents, PowerPoints, Excel spreadsheets, public URLs or another option so you

120
00:05:28,440 --> 00:05:31,520
can point the agent to your company website or a public knowledge base.

121
00:05:31,520 --> 00:05:35,840
And if you have a copilot license, you can also connect Teams chats and Outlook emails.

122
00:05:35,840 --> 00:05:37,760
Now there's a toggle that matters a lot.

123
00:05:37,760 --> 00:05:39,800
Only use specified sources.

124
00:05:39,800 --> 00:05:43,880
Flip this on, and the agent will only answer from the documents and sites you've provided.

125
00:05:43,880 --> 00:05:46,600
But don't use the general knowledge of the AI model to make things up.

126
00:05:46,600 --> 00:05:48,600
This is your best defense against hallucinations.

127
00:05:48,600 --> 00:05:52,720
If you're building an HR agent that answers from your employee handbook, you want this turned

128
00:05:52,720 --> 00:05:56,120
on like you don't want the agent inventing a policy that doesn't exist.

129
00:05:56,120 --> 00:05:57,480
There are limits to keep in mind.

130
00:05:57,480 --> 00:06:02,160
Microsoft says individual files can be up to 512 megabytes for most document types.

131
00:06:02,160 --> 00:06:04,200
With Excel files kept at 30 megabytes.

132
00:06:04,200 --> 00:06:07,880
In practical terms, you're looking at about 100 SharePoint files per agent and around

133
00:06:07,880 --> 00:06:09,560
50 one drive files.

134
00:06:09,560 --> 00:06:12,440
Those aren't hard limits written in stone, but they're good guidelines for planning.

135
00:06:13,440 --> 00:06:15,960
Here's the part that surprises most people.

136
00:06:15,960 --> 00:06:19,440
The agent respects your existing Microsoft 365 permissions.

137
00:06:19,440 --> 00:06:20,560
It doesn't get special access.

138
00:06:20,560 --> 00:06:24,920
If you add a SharePoint site as a knowledge source, the agent can only return information

139
00:06:24,920 --> 00:06:28,160
that the person asking the question is already allowed to see.

140
00:06:28,160 --> 00:06:32,080
So if a user can't normally view a specific document in SharePoint, the agent won't show it

141
00:06:32,080 --> 00:06:33,080
to them either.

142
00:06:33,080 --> 00:06:37,080
This is built in security, and it works automatically.

143
00:06:37,080 --> 00:06:38,400
The template shortcut.

144
00:06:38,400 --> 00:06:40,000
Welcome back to Microsoft Knowledge Nuggets.

145
00:06:40,000 --> 00:06:41,440
I'm Mirko Peters.

146
00:06:41,440 --> 00:06:44,720
But what if you want to build something faster than starting from scratch?

147
00:06:44,720 --> 00:06:46,240
That's where templates come in.

148
00:06:46,240 --> 00:06:49,240
Microsoft now pushes template first creation pretty heavily.

149
00:06:49,240 --> 00:06:53,000
Instead of staring at a blank described tab, you pick from pre-configured agents designed

150
00:06:53,000 --> 00:06:54,320
for common scenarios.

151
00:06:54,320 --> 00:06:57,360
Career coach helps people navigate professional development.

152
00:06:57,360 --> 00:06:59,600
Researcher digs deep across multiple sources.

153
00:06:59,600 --> 00:07:01,560
Analyst focuses on data and spreadsheets.

154
00:07:01,560 --> 00:07:04,760
And there are industry-specific ones too, depending on your organization.

155
00:07:04,760 --> 00:07:06,520
Microsoft keeps adding more over time.

156
00:07:06,520 --> 00:07:07,920
The workflow is simple.

157
00:07:07,920 --> 00:07:09,800
Pick a template close to what you need.

158
00:07:09,800 --> 00:07:10,960
Then customize it.

159
00:07:10,960 --> 00:07:12,640
Swap in your own knowledge sources.

160
00:07:12,640 --> 00:07:13,880
Tweak the instructions.

161
00:07:13,880 --> 00:07:15,400
Adjust the suggested prompts.

162
00:07:15,400 --> 00:07:19,280
The template gives you a solid starting structure, but the real value comes from the data

163
00:07:19,280 --> 00:07:20,280
you connect.

164
00:07:20,280 --> 00:07:21,280
Think of it like cooking.

165
00:07:21,280 --> 00:07:24,440
Starting from a recipe is way easier than inventing a dish from scratch.

166
00:07:24,440 --> 00:07:28,600
The recipe gives you the proportions, the techniques, the order of operations.

167
00:07:28,600 --> 00:07:31,520
You still need to add your own ingredients, and that's your company's data.

168
00:07:31,520 --> 00:07:33,960
But you don't have to figure out the structure yourself.

169
00:07:33,960 --> 00:07:37,200
Templates are great for business users who want speed without complexity.

170
00:07:37,200 --> 00:07:40,920
If you're not a developer and don't want to learn prompt engineering, a template gets

171
00:07:40,920 --> 00:07:42,680
you a working agent in minutes.

172
00:07:42,680 --> 00:07:46,200
You just need to know what you want the agent to do and where the relevant information

173
00:07:46,200 --> 00:07:47,200
lives.

174
00:07:47,200 --> 00:07:50,160
Just remember, a template is a starting point, not a finished product.

175
00:07:50,160 --> 00:07:53,640
The agent is only as useful as the knowledge sources you connect to it.

176
00:07:53,640 --> 00:07:57,920
A career coach template with no company data attached is just a generic career advice

177
00:07:57,920 --> 00:08:02,080
board, connected to your internal training documents and job postings, and suddenly it

178
00:08:02,080 --> 00:08:04,960
becomes a tool that actually helps your employees.

179
00:08:04,960 --> 00:08:05,960
The spectrum.

180
00:08:05,960 --> 00:08:09,080
Agent Builda versus Copilot Studio versus AI Foundry.

181
00:08:09,080 --> 00:08:11,320
Now here's the question everyone asks.

182
00:08:11,320 --> 00:08:15,000
When should you use Agent Builda versus the other tools Microsoft offers?

183
00:08:15,000 --> 00:08:18,240
Because Microsoft gives you three ways to build agents and they're not the same thing.

184
00:08:18,240 --> 00:08:19,240
Think of it as a spectrum.

185
00:08:19,240 --> 00:08:23,840
On one end, you have Agent Builda, the simplest, fastest, most accessible option in the middle,

186
00:08:23,840 --> 00:08:28,400
Copilot Studio with more power, more complexity, more control, and on the far end, AI Foundry,

187
00:08:28,400 --> 00:08:31,600
the full developer platform where you can do almost anything, but you'd better know

188
00:08:31,600 --> 00:08:32,600
what you're doing.

189
00:08:32,600 --> 00:08:33,600
Start with Agent Builda.

190
00:08:33,600 --> 00:08:38,080
This is for quick internal Q&A assistance grounded in Microsoft 365 content.

191
00:08:38,080 --> 00:08:39,240
No code required.

192
00:08:39,240 --> 00:08:42,640
You describe what you want, connect some knowledge sources, and you're done.

193
00:08:42,640 --> 00:08:43,640
But there are trade-offs.

194
00:08:43,640 --> 00:08:46,280
Agent Builda agents stay inside your organization.

195
00:08:46,280 --> 00:08:48,120
You can't put them on a public website.

196
00:08:48,120 --> 00:08:50,760
You can't connect them to Salesforce or Service now.

197
00:08:50,760 --> 00:08:54,360
They answer questions from your SharePoint sites and documents, and that's about it.

198
00:08:54,360 --> 00:08:56,440
If that's all you need, Agent Builda is perfect.

199
00:08:56,440 --> 00:08:58,160
Copilot Studio is the next step up.

200
00:08:58,160 --> 00:09:01,160
This is a low-code platform for enterprise-grade agents.

201
00:09:01,160 --> 00:09:05,560
It supports workflows, real multi-step processes with branching and conditional logic.

202
00:09:05,560 --> 00:09:10,360
It has connectors to external systems like Salesforce, Dataverse, and hundreds of other services.

203
00:09:10,360 --> 00:09:14,200
You can publish agents to websites, teams, Slack, even custom apps.

204
00:09:14,200 --> 00:09:15,840
But with that power comes complexity.

205
00:09:15,840 --> 00:09:16,840
You need licensing.

206
00:09:16,840 --> 00:09:18,000
You need governance.

207
00:09:18,000 --> 00:09:19,840
Your IT department needs to be involved.

208
00:09:19,840 --> 00:09:22,440
It's not something a business user sets up in five minutes.

209
00:09:22,440 --> 00:09:24,800
AI Foundry is the far end of the spectrum.

210
00:09:24,800 --> 00:09:27,640
This is Microsoft's professional AI development platform.

211
00:09:27,640 --> 00:09:28,760
Pro developers only.

212
00:09:28,760 --> 00:09:32,960
You can choose from thousands of AI models, not just Microsoft's, but Claude, Grog, even

213
00:09:32,960 --> 00:09:34,360
specialized healthcare models.

214
00:09:34,360 --> 00:09:37,920
You can fine tune them, customize them, build entirely custom solutions.

215
00:09:37,920 --> 00:09:40,520
But for most business users, AI Foundry is overkill.

216
00:09:40,520 --> 00:09:43,680
If you're asking, should I use Agent Builda or AI Foundry?

217
00:09:43,680 --> 00:09:45,560
The answer is Agent Builda.

218
00:09:45,560 --> 00:09:48,200
So when do you move from Agent Builda to Copilot Studio?

219
00:09:48,200 --> 00:09:49,200
Three main reasons.

220
00:09:49,200 --> 00:09:51,480
Reason one, you need an external facing agent.

221
00:09:51,480 --> 00:09:53,880
Agent Builda agents only work inside your organization.

222
00:09:53,880 --> 00:09:58,320
If you want a customer service bot on your public website, you need Copilot Studio.

223
00:09:58,320 --> 00:10:01,320
Reason two, you need exotic data sources.

224
00:10:01,320 --> 00:10:06,000
Agent Builda connects to SharePoint, OneDrive, WebContent, and Microsoft 365 data.

225
00:10:06,000 --> 00:10:11,120
But if you need to pull data from a SQL database, Salesforce, or an MCP server, you need Copilot

226
00:10:11,120 --> 00:10:12,120
Studio.

227
00:10:12,120 --> 00:10:14,160
Reason three, you need agentic actions.

228
00:10:14,160 --> 00:10:18,280
This means the agent doesn't just answer questions, it does things, it creates tickets, sends

229
00:10:18,280 --> 00:10:21,200
emails, updates records, triggers workflows.

230
00:10:21,200 --> 00:10:22,520
Agent Builda can't do that.

231
00:10:22,520 --> 00:10:23,520
Copilot Studio can.

232
00:10:23,520 --> 00:10:25,440
Here's a real world way to think about it.

233
00:10:25,440 --> 00:10:30,560
An HR FAQ bot that answers questions from your employee handbook is Agent Builda territory.

234
00:10:30,560 --> 00:10:35,320
A helpdesk agent that reads a support request creates a ticket in service now, roots it to

235
00:10:35,320 --> 00:10:39,080
the right team, and sends a confirmation email is Copilot Studio.

236
00:10:39,080 --> 00:10:44,080
A medical diagnosis assistant that uses specialized healthcare models and needs to comply with

237
00:10:44,080 --> 00:10:46,120
strict regulations is AI Foundry.

238
00:10:46,120 --> 00:10:48,600
The bottom line is simple, start with Agent Builda.

239
00:10:48,600 --> 00:10:52,560
It's free with your Copilot license, it takes minutes to set up, and it handles a surprising

240
00:10:52,560 --> 00:10:54,280
number of use cases.

241
00:10:54,280 --> 00:10:57,960
Only level up when you hit a wall, when you need external access, different data sources

242
00:10:57,960 --> 00:10:59,560
or actual actions.

243
00:10:59,560 --> 00:11:01,920
Most people never need to leave Agent Builda.

244
00:11:01,920 --> 00:11:06,520
The hard limits, what Agent Builda can't do, but Agent Builda does have limits.

245
00:11:06,520 --> 00:11:10,120
Let's talk about where it stops, because knowing when a tool isn't right for the job is

246
00:11:10,120 --> 00:11:12,280
just as important as knowing when it is.

247
00:11:12,280 --> 00:11:15,360
One important limit, these agents stay inside Copilot.

248
00:11:15,360 --> 00:11:19,200
You open Copilot in your browser, see your agent in the sidebar and chat with it there,

249
00:11:19,200 --> 00:11:22,400
but you cannot pull it into a team's conversation or add it to a channel.

250
00:11:22,400 --> 00:11:26,720
If you need an agent that lives inside teams, you have outgrown Agent Builda.

251
00:11:26,720 --> 00:11:30,120
The limit, no external facing agents.

252
00:11:30,120 --> 00:11:33,560
Everything you build in Agent Builda stays internal to your organization.

253
00:11:33,560 --> 00:11:38,160
You can share agents with co-workers, but you cannot put one on your public website, embedded

254
00:11:38,160 --> 00:11:41,800
in a customer portal, or let anyone outside your company use it.

255
00:11:41,800 --> 00:11:46,120
If your goal is a customer service bot on your homepage, you need Copilot Studio.

256
00:11:46,120 --> 00:11:50,000
Then there's complexity, Agent Builda is designed for Q&A, not automation.

257
00:11:50,000 --> 00:11:54,680
There is no branching logic, no conditional steps, no if this then that sequences.

258
00:11:54,680 --> 00:11:58,640
You cannot connect it to Power Automate to trigger actions when someone asks the question.

259
00:11:58,640 --> 00:12:00,560
The agent answers questions and that is it.

260
00:12:00,560 --> 00:12:05,520
If you need a multi-step process, like an agent that reads a request, checks inventory, creates

261
00:12:05,520 --> 00:12:08,040
an order, and sends a confirmation.

262
00:12:08,040 --> 00:12:10,000
Agent Builda will not do it.

263
00:12:10,000 --> 00:12:11,600
External connections are also off the table.

264
00:12:11,600 --> 00:12:16,440
You cannot connect Agent Builda to Salesforce, ServiceNow, or any other third-party system.

265
00:12:16,440 --> 00:12:21,480
It works with Microsoft 365 content, SharePoint, OneDrive, Teams, Email, and Public Web content.

266
00:12:21,480 --> 00:12:22,680
That is the full list.

267
00:12:22,680 --> 00:12:27,240
If your agent needs to pull data from a CRM or update a record in an external database,

268
00:12:27,240 --> 00:12:28,680
you need a different tool.

269
00:12:28,680 --> 00:12:30,240
Compliance features are missing too.

270
00:12:30,240 --> 00:12:33,600
Agent Builda does not support lockbox or customer managed keys.

271
00:12:33,600 --> 00:12:37,680
Those controls that let organizations manage encryption keys or require explicit approval

272
00:12:37,680 --> 00:12:40,560
before Microsoft engineers can access data.

273
00:12:40,560 --> 00:12:44,400
If you work in a highly regulated industry where those are mandatory, Agent Builda is not

274
00:12:44,400 --> 00:12:45,400
the right choice.

275
00:12:45,400 --> 00:12:48,560
Here is a licensing gotcha that catches people during testing.

276
00:12:48,560 --> 00:12:52,240
SharePoint grounding requires a proper Microsoft 365 Copilot license.

277
00:12:52,240 --> 00:12:57,080
You cannot build an agent that reads from SharePoint in a developer-only tenant or a trial account.

278
00:12:57,080 --> 00:13:00,560
The agent will work, but it won't be able to access your SharePoint data.

279
00:13:00,560 --> 00:13:02,640
And finally, practical source limits.

280
00:13:02,640 --> 00:13:08,400
Per agent you are looking at roughly 100 SharePoint files, 51 Drive files, 20 uploaded documents,

281
00:13:08,400 --> 00:13:10,960
full public URLs, and five Teams chats.

282
00:13:10,960 --> 00:13:14,600
Those aren't hard Microsoft limits written in stone, but they're what the community consistently

283
00:13:14,600 --> 00:13:15,600
reports.

284
00:13:15,600 --> 00:13:19,120
If your agent needs to search across hundreds of documents, you will hit a ceiling.

285
00:13:19,120 --> 00:13:20,280
So here is the honest truth.

286
00:13:20,280 --> 00:13:22,760
If you need any of these things, you have outgrown Agent Builda.

287
00:13:22,760 --> 00:13:23,760
That is not a failure.

288
00:13:23,760 --> 00:13:26,840
Agent Builda is made for a specific job and it does that job well.

289
00:13:26,840 --> 00:13:31,480
But when your needs get more complex, it is time to look at Copilot Studio.

290
00:13:31,480 --> 00:13:33,560
Security and governance, the simple version.

291
00:13:33,560 --> 00:13:35,840
So how does security work with Agent Builda?

292
00:13:35,840 --> 00:13:37,280
Here is the simple version.

293
00:13:37,280 --> 00:13:40,600
Agent Builda respects your existing Microsoft 365 permissions.

294
00:13:40,600 --> 00:13:41,600
That is the key point.

295
00:13:41,600 --> 00:13:45,760
When you add a SharePoint site as a knowledge source, the agent does not get special access.

296
00:13:45,760 --> 00:13:48,840
They can only return information that the person asking the question is already allowed

297
00:13:48,840 --> 00:13:49,840
to see.

298
00:13:49,840 --> 00:13:52,640
Other user cannot view a specific document in SharePoint.

299
00:13:52,640 --> 00:13:54,360
The agent will not show it to them either.

300
00:13:54,360 --> 00:13:56,280
The permissions are inherited automatically.

301
00:13:56,280 --> 00:13:58,840
This is built in and works without any extra configuration.

302
00:13:58,840 --> 00:14:01,840
You do not need to set up separate access rules for your agents.

303
00:14:01,840 --> 00:14:05,600
They follow the same rules your organization already has in place.

304
00:14:05,600 --> 00:14:08,800
The only use specified sources toggle is very useful.

305
00:14:08,800 --> 00:14:09,800
Turn it on.

306
00:14:09,800 --> 00:14:13,500
And the agent only answers from the documents and sites you explicitly provide.

307
00:14:13,500 --> 00:14:15,840
It ignores the general knowledge of the AI model.

308
00:14:15,840 --> 00:14:18,480
This is your primary defense against hallucinations.

309
00:14:18,480 --> 00:14:22,560
The agent cannot make up a policy if it is forced to use only your employee handbook.

310
00:14:22,560 --> 00:14:25,280
Simple, effective, and we always recommend it.

311
00:14:25,280 --> 00:14:26,880
One thing to watch for.

312
00:14:26,880 --> 00:14:29,840
Tenant level policies can override what you see in the UI.

313
00:14:29,840 --> 00:14:33,840
If your IT admin has disabled web search and copilot through a tenant policy, the web content

314
00:14:33,840 --> 00:14:37,360
toggle in Agent Builda might still appear enabled, but it won't actually work.

315
00:14:37,360 --> 00:14:38,440
The policy wins.

316
00:14:38,440 --> 00:14:42,480
So if you are building an agent and wondering why web content isn't working, check with your

317
00:14:42,480 --> 00:14:43,600
admin first.

318
00:14:43,600 --> 00:14:47,000
You do not need to become a security expert to use Agent Builda.

319
00:14:47,000 --> 00:14:49,920
You will understand that the agent inherits your organization's rules.

320
00:14:49,920 --> 00:14:53,040
It does not create new permissions or bypass existing ones.

321
00:14:53,040 --> 00:14:55,880
If your data is secure and share point, it is secure in your agent.

322
00:14:55,880 --> 00:15:00,800
For organizations that need more, monitoring, lifecycle management, identity controls, there

323
00:15:00,800 --> 00:15:02,840
is a tool called Agent365.

324
00:15:02,840 --> 00:15:05,680
It is Microsoft's governance layer for AI agents.

325
00:15:05,680 --> 00:15:07,920
But that is a separate topic for a future episode.

326
00:15:07,920 --> 00:15:11,320
For most people using Agent Builda, the built-in security is enough.

327
00:15:11,320 --> 00:15:15,440
So that's Microsoft Agent Builda in plain English and the easiest way to create a helpful

328
00:15:15,440 --> 00:15:18,240
AI assistant for your team without writing any code.

329
00:15:18,240 --> 00:15:19,840
Here is your one task this week.

330
00:15:19,840 --> 00:15:24,280
Open Copilot, click New Agent, start with the Described tab, pick a template or describe

331
00:15:24,280 --> 00:15:27,120
what you need, connect a SharePoint site and test it.

332
00:15:27,120 --> 00:15:29,360
You'll have a working agent in under 10 minutes.

333
00:15:29,360 --> 00:15:33,520
If this helped, subscribe on your favorite podcast platform and share it with someone starting

334
00:15:33,520 --> 00:15:34,520
their AI journey.

335
00:15:34,520 --> 00:15:37,280
I'm Mirko Peters and I'll catch you on the next episode.