Latest / Microsoft Cloud IT Pro Podcast / Episode 431: Agent Governance Is the New App Governance
Transcript
- 0:03- Welcome to episode 431
- 0:05- of the Microsoft Cloud IT Pro podcast recorded
- 0:08- live on 06/26/2026.
- 0:11- This is a show about Microsoft three sixty
- 0:13- five and Azure from the perspective of IT
- 0:15- pros and end users, where we discuss the
- 0:17- topic or recent news and how it relates
- 0:20- to you. In this episode, Ben and Scott
- 0:22- get into why agent governance is quickly becoming
- 0:25- one of the most pressing challenges for Microsoft
- 0:28- three sixty five administrators,
- 0:30- As agentic workflows become more common across Copilot,
- 0:33- Copilot Studio, and third party platforms,
- 0:36- organizations are dealing with a sprawl problem that
- 0:38- looks familiar but hits differently. Let's dive in
- 0:42- as we talk about Copilot agents and the
- 0:44- similarities and differences to enterprise apps.
- 0:49- Can I give you a quick rant? I
- 0:50- don't know. Is it gonna make me start
- 0:52- ranting? Is this gonna be the episode where
- 0:54- Scott rants and Ben gets ranting? Alright. What's
- 0:57- your rant? I happen to have an email
- 1:00- address, an Outlook dot com email address that
- 1:03- is a very short
- 1:05- handle, and it's a very common thing.
- 1:08- But it's kinda like a a vanity handle.
- 1:11- Whatever. But it would be something that's pretty
- 1:13- common,
- 1:14- especially
- 1:14- if your name was Scott and your last
- 1:18- initial started with an h.
- 1:20- So
- 1:21- I have this thing
- 1:23- where
- 1:24- I don't understand
- 1:25- why
- 1:26- people
- 1:27- use this email address, which is my email
- 1:30- address, to be very clear. Like, it's all
- 1:31- locked down to me. I've got all my
- 1:34- YubiKeys,
- 1:35- authenticator.
- 1:36- Like, I closely monitor the logins because I
- 1:38- do get these weird, like, who somebody's trying
- 1:40- to log in to your account kinds of
- 1:42- things, right, with password sprays and everything like
- 1:45- that. But so so it's a real email
- 1:47- address. I use it. Like, it's a real
- 1:49- email address. Unfortunately,
- 1:51- I have this issue where other people
- 1:53- are also using it as a real email
- 1:56- address. Like, I had somebody the other day
- 1:58- went and signed up for a Walmart account.
- 2:00- So they signed up for a Walmart account.
- 2:02- Somehow, they were able to sign up for
- 2:04- a Walmart account, authenticate with their phone number
- 2:07- to that, like, they didn't need to get,
- 2:08- like, the access code out of email or
- 2:10- things like that. And then now they've got
- 2:12- a Walmart account, and they're making purchases on
- 2:14- it. And I'm getting, like, the invoices and
- 2:16- the receipts.
- 2:18- Normally, what I can do is when I
- 2:19- get these kinds of things, I go and
- 2:21- I
- 2:22- I say, hey. I forgot my password, and
- 2:25- I request the account to be deleted, or
- 2:27- I just change the password to something that's,
- 2:29- like, black holes. Like, I'm not a smart
- 2:31- person and use somebody else's email address to
- 2:33- sign up for this account. Like like, it's
- 2:35- just this big long string that can be
- 2:37- crammed in there. But the problem I'm running
- 2:40- into lately is there are sites like Walmart.
- 2:43- I've got PayPal kicking around, a couple of
- 2:45- banks that people have signed up for, like,
- 2:48- stuff that I should not be getting. Right?
- 2:49- So but I can't go in, and I
- 2:51- can't request the deletion of the account. So,
- 2:54- like, Walmart, you go in and you say,
- 2:56- hey. Here's my email address. Send me a
- 2:58- code. It says, great. I sent you a
- 2:59- code, but you have to authenticate with your
- 3:01- phone. Well, I don't have my phone. Right?
- 3:03- Because it's not my phone. It's somebody else's
- 3:06- somebody else's phone that that put in there.
- 3:09- So I have this email address that more
- 3:10- and more, I cannot use for anything personal
- 3:13- because there's so many other people in the
- 3:15- world with the same first name and last
- 3:16- initial
- 3:17- that are using it for all their stuff
- 3:19- and getting it pushed through. I I have,
- 3:21- like, Walmart account, bank accounts.
- 3:23- I have a PayPal account in The UK
- 3:25- that somebody signed up for, like, can't get
- 3:27- into. I had this week, somebody did Instacart,
- 3:31- and another person did Grubhub and signed up
- 3:34- for Grubhub Plus. Cool.
- 3:38- It's
- 3:39- really weird.
- 3:41- And I would understand, like, common email address,
- 3:44- like,
- 3:45- first, last initial, blah blah blah. Like, somebody's
- 3:47- just pumping it in to
- 3:49- sign in to Wi Fi somewhere randomly. Right?
- 3:51- And they don't care and, like, hey. I'll
- 3:53- go spam this other person kinda thing. But
- 3:55- these are, like, legit
- 3:57- accounts. These are legitimate
- 3:59- things. People are buying stuff. They're getting bank
- 4:02- accounts.
- 4:03- There's some guy in Iowa
- 4:05- who runs an entire survey business, and I'm
- 4:09- getting all his invoices
- 4:11- and all the surveys
- 4:13- that he's, like, sending back and forth to
- 4:15- the companies that he's working with. Like, I
- 4:17- don't understand
- 4:18- how
- 4:19- this is the state of the world. It's
- 4:21- so strange to me. You said you weren't
- 4:23- gonna get me ranting, but this is gonna
- 4:25- be an interesting one, I think, that's going
- 4:27- to come up. Not because part you mentioned
- 4:30- it in here. It was like a sub
- 4:31- bullet of what you said when you talked
- 4:33- about you needed your phone for authentication.
- 4:36- Is
- 4:37- my problem is I have password less set
- 4:40- up a lot of my accounts because I
- 4:41- don't wanna type in a password. And I'm
- 4:43- curious if you start seeing this. The amount
- 4:45- of prompts
- 4:46- that I get an authenticator
- 4:48- because
- 4:50- and I've had a client that had this
- 4:51- too with one of their email accounts because
- 4:53- nobody needs a password because it relies on
- 4:56- the push notification.
- 4:58- I get, like, I get spammed with authenticator
- 5:01- requests, and it's not approve or deny.
- 5:03- It's number matching once. And I'm like, okay.
- 5:05- I don't think this was me. Sometimes it's
- 5:07- close to the same time I do it.
- 5:09- But now I started seeing it's, like, from
- 5:11- Germany, and it's from The US, and it's
- 5:13- from here, and it's from there as and
- 5:15- there's no way to stop it. Like, I
- 5:17- can't go change a password. I can't go
- 5:20- change my authenticator
- 5:22- app. Like, in this world of passwordless,
- 5:25- I feel like these authenticator
- 5:27- spamming
- 5:28- number matching things
- 5:30- are just going to get annoying. Like They
- 5:32- are very annoying. So so I have the
- 5:34- same problem, again, on my Outlook account, but
- 5:37- mine aren't number matches.
- 5:38- So the problem is the way Microsoft does
- 5:41- it, e even if somebody happens to find
- 5:43- a way in, like, Microsoft's rationale is, like,
- 5:46- hey. We're still sending you the note the
- 5:48- push notification. Right? So, therefore, it's up to
- 5:50- you to approve or deny. But I don't
- 5:51- know if the push notification
- 5:53- like, I know which ones are from me
- 5:55- in the moment as a number match. Like,
- 5:56- that's very clear. Yep. You might have something
- 5:58- in the background. So, like, for example, for
- 6:00- me, I sign into my outlook.com
- 6:02- account through Outlook on my desktop. Makes sense.
- 6:06- Right? It might be Outlook just trying to
- 6:08- reauthenticate. It might be my iPhone trying to
- 6:10- reauthenticate. It might be my tablet or another
- 6:12- computer. Like, could be some service I signed
- 6:15- up for, and I gave permission
- 6:16- to read my But you don't know what
- 6:18- it is, and
- 6:20- Microsoft
- 6:21- in they they used to publish all the
- 6:24- security information. So you could go in and
- 6:26- you go to, like, account.microsoft,
- 6:28- and you could see all the sign ins,
- 6:29- and you could see where they were coming
- 6:31- from. That got so noisy that they stopped
- 6:34- publishing those. So now I get all these
- 6:37- push notifications.
- 6:38- I get all these authenticator prompts, and I
- 6:40- have no
- 6:42- way to even know where they're coming from.
- 6:44- Like, I don't even know if they're coming
- 6:45- from Russia or China or The United States
- 6:48- or anything else anywhere. Because, like, I'll I'll
- 6:50- tell you, like, if I go and look
- 6:51- I was just doing this before I hopped
- 6:52- on because I was trying to clean up
- 6:54- some of these emails out of that mailbox.
- 6:56- So I went and looked. The only login
- 6:58- that shows up in when I go to
- 7:00- account.microsoft.com
- 7:02- for that
- 7:03- Microsoft account, the only login that I see
- 7:06- is one login from today. Okay. I authenticate
- 7:09- to that account a ton of times today
- 7:11- and from my iPhone. And so, like, my
- 7:14- iPhone's running Defender,
- 7:16- and I have a VPN client on there.
- 7:18- So, theoretically, my iPhone should be showing up
- 7:21- as a separate login, like, being routed through,
- 7:24- like, Defender and, like, Seattle where the VPN
- 7:27- location is. Or if I hop on to
- 7:29- Tailscale and I happen to have an exit
- 7:31- node in another location,
- 7:33- shouldn't, like, the login thing when I get
- 7:34- reprompted for auths, shouldn't it show me that
- 7:36- I relogged in from a new location? Like,
- 7:38- they they just, like, masked all that information
- 7:40- and hid it away. I think it was
- 7:42- I get why it was noise for customers.
- 7:44- I bet it was a ton of telemetry
- 7:46- for them to collect as well and have
- 7:47- to kinda store and push out and surface
- 7:50- and query and
- 7:51- and all those kinds of things. But in
- 7:53- this world of, like, passwordless,
- 7:56- you tie a phone number to your account,
- 7:58- it's no longer
- 7:59- just a password kinda thing. It is it's
- 8:02- rough, especially when, at least in my case,
- 8:05- for people who haven't gone through this, it
- 8:06- stinks when somebody else is using your email
- 8:08- address. It's really annoying. Thanks for that because
- 8:10- I thought I was going crazy because I
- 8:11- did the same thing. When I started getting
- 8:14- error as I've been getting all these random
- 8:17- prompts, I wanted to go see, like, where
- 8:19- are they coming from? Is there any way
- 8:20- I can report these? And I had the
- 8:23- same thing. I'm like, I thought I used
- 8:25- to see all the login attempts, and now
- 8:27- I don't see anything. And I was like,
- 8:29- maybe I was getting entry confused with my
- 8:32- personal account, but they did used to have
- 8:34- that, and they took it away. They very
- 8:36- much cut it back,
- 8:38- and they removed a bunch of controls there.
- 8:40- Like, they used to have stuff for, like,
- 8:42- reporting things Yep. And all that. So it's
- 8:45- down to the point now where even if
- 8:46- you go into the docs for, like, reading
- 8:49- about, like, what recent activity is,
- 8:52- they go so far as to say, if
- 8:54- you get an email about unusual activity and
- 8:57- you're not sure
- 8:58- if it's from Microsoft, just go ahead and,
- 9:00- like, sign into your account. And even if
- 9:03- you get one about unusual activity that is
- 9:04- from Microsoft, when you sign into your account,
- 9:07- the detail's not there. There there is no
- 9:09- detail there. It's just it is completely
- 9:13- opaque. I I don't know how they could
- 9:15- make it better versus going back to the
- 9:17- way it used to be. Like, show me
- 9:19- every time you send me a push notification,
- 9:21- show me why. Like like, give me a
- 9:23- little bit of insight into where that was
- 9:25- coming from. Like, because if I could even
- 9:27- see, like, a user agent string and a
- 9:29- timestamp, like, if I could see, like, oh,
- 9:31- that was Outlook from macOS
- 9:33- at 03:25PM
- 9:35- on Friday. Oh, that was me. Like, I
- 9:37- was just in Outlook.
- 9:39- And, oh, yeah. I did get a little
- 9:41- error pop up or things like that. So
- 9:43- yeah.
- 9:44- Any anyhoo, that that's my rant. Somebody else
- 9:47- having your email address sucks.
- 9:49- Passwordless
- 9:50- is
- 9:51- super annoying in a world of opaqueness.
- 9:54- I think you should like, I think the
- 9:55- vendor should air on on on the side
- 9:57- of verbosity
- 9:58- and let folks know, but, unfortunately,
- 10:01- they do not do that. Yes. And report,
- 10:03- like, if I report that I'm not logging
- 10:05- in from this particular one, I'm surprised how
- 10:08- many keep coming through that appear to be
- 10:10- coming from the same location. Mhmm. And I
- 10:12- don't know how you get around it. Because,
- 10:14- ideally, I would say, you know what? I'm
- 10:15- not in Germany, and, frankly, I'm not in
- 10:17- any country but The US. Any push notification
- 10:20- from anything other than The US should be
- 10:22- denied. Now if I travel,
- 10:24- I'd want a way to open that back
- 10:26- up. But for now, like, I should not
- 10:29- even be having to deal with those push
- 10:31- notifications.
- 10:32- It's interesting. It's weird. Actually, for my Microsoft
- 10:34- account,
- 10:35- I think I'm at the point where, like,
- 10:37- at least for my consumer account because of,
- 10:39- like, the state it's in. I was gonna
- 10:41- say consumer accounts is where I have this
- 10:42- problem more than my business one. Definitely the
- 10:45- most there. But for that, I would actually
- 10:48- prefer to be able to turn off authenticator.
- 10:50- Like, if I have another passwordless
- 10:52- mechanism,
- 10:53- so if I've got biometrics just through my
- 10:56- phone with a passkey
- 10:58- or I have a YubiKey or something, give
- 11:00- me the option to turn off authenticator.
- 11:02- You let me turn off my phone, like,
- 11:04- my phone number and OTP.
- 11:06- Let me turn off authenticator. Like, don't force
- 11:10- I'm sorry. Like, that piece of garbage application,
- 11:13- like, at down my throat and make it
- 11:15- the only the only pass in.
- 11:19- If you're managing file storage today, you've probably
- 11:22- felt the pressure. Refresh cycles are harder to
- 11:24- plan, costs and complexity are increasing, especially as
- 11:28- file workloads scale across regions. That's why Microsoft
- 11:32- customers trust Nasuni. As an Azure certified software,
- 11:36- Nasuni provides a global file system designed for
- 11:39- enterprise scale environments,
- 11:40- keeping file data secure, governed, and accessible across
- 11:44- Azure environments without added infrastructure complexity.
- 11:47- It's the foundation to keep your unstructured data
- 11:50- stable even as your infrastructure evolves. To see
- 11:54- how much you could save while keeping your
- 11:56- initiatives on track, visit nasuni.com/tco.
- 12:01- That's nasuni.com/tco.
- 12:07- You know what else people are having to
- 12:08- deal with, Scott, in managing outside of authentication
- 12:12- requests? This was such a good transition to
- 12:13- our main topic for the day. It was.
- 12:15- Good job.
- 12:16- And agents. Agents are everywhere. Agents are everywhere.
- 12:20- So this is an interesting one. Like like,
- 12:22- my day job,
- 12:24- I think a lot about
- 12:26- SDKs and rest APIs
- 12:28- and things like that. And even in, like,
- 12:31- my day job, I spend more and more
- 12:32- of my time thinking about how agents interact
- 12:35- with those things over humans as more and
- 12:38- more stuff gets handed off. And I'm definitely,
- 12:41- like, in that ecosystem. I think you're in
- 12:43- that ecosystem as well, kind of being tied
- 12:46- into even just, like, generic Microsoft three sixty
- 12:49- five. So if I'm just in Microsoft three
- 12:51- sixty five, let's say Copilot, I've got a
- 12:53- researcher agent. I've got an analyst agent. I
- 12:56- can connect to
- 12:57- to SAP. I can connect to consumer services
- 13:01- like Canva,
- 13:02- all these different things. And now that Cowork
- 13:05- has g eight, I think you start to
- 13:07- see
- 13:08- a a little bit more probably usage
- 13:10- and uptick there. It's becoming, like, more prolific,
- 13:14- I think, to see
- 13:16- agentic workflows,
- 13:18- whether those are, like, true, like, bounded agents,
- 13:21- they're just kind of fluffy declarative agents in
- 13:24- m three sixty five Copilot, whatever happens to
- 13:26- be, but agentic workflows
- 13:29- augmenting
- 13:30- more and more of just the day to
- 13:32- day of what's going on there. So that
- 13:34- raises an interesting question. In a world of
- 13:38- agents
- 13:39- and likely more agents than humans that are
- 13:42- in your organization because there's an amplification factor
- 13:45- of I'm one person, and I might use
- 13:47- 20 agents.
- 13:48- How do you kinda know what's going on
- 13:50- out there? How do you monitor?
- 13:53- How do you respond? And I think, largely,
- 13:55- like, there's an argument to be made to
- 13:57- say is that
- 13:58- agentic governance or agent governance
- 14:01- is
- 14:02- the new
- 14:03- application governance space. So much like you might
- 14:06- have been in, like, a world of MAM
- 14:09- and managing to the applications
- 14:11- that your users could use, now you're in
- 14:13- the world of agents, and you also have
- 14:15- to manage to the agents that your users
- 14:18- can use. Or maybe you're in a fully
- 14:19- ungated world, and you don't know what agents
- 14:23- your users
- 14:25- are using out there. And it's such a
- 14:27- wide landscape
- 14:28- and wide surface area for agents between Microsoft
- 14:32- three sixty five Copilot.
- 14:34- You got Copilot Studio.
- 14:35- There's the whole just broader
- 14:38- Microsoft
- 14:39- security stack,
- 14:41- and you kinda gotta get out there and
- 14:43- find a way to
- 14:45- govern agents
- 14:46- either as applications or as,
- 14:49- I guess, artifacts of your enterprise
- 14:51- and kinda part of your
- 14:53- core business. It's interesting because I in some
- 14:56- respects, I always thought about agent governance as
- 14:59- also
- 15:01- the new
- 15:02- almost people governance
- 15:04- to your point, but agents are also apps
- 15:06- and agents have to deal with sprawl. It's
- 15:08- governing agents is turning into I feel like
- 15:12- a massive thing. Because to your point with
- 15:13- apps, you have, like, who can access this
- 15:15- agent? Kinda like you'd have to manage apps.
- 15:18- Who can log into it? What can they
- 15:20- do with this agent? But then you also
- 15:22- have agents running
- 15:24- similar to people where it's the data issue
- 15:26- is what data can these agents access and
- 15:28- which agents can access which data, and how
- 15:30- do you protect data from agents. And then
- 15:32- you have
- 15:34- the sprawl aspect of it of, again, people
- 15:37- just going out and adding agents, and
- 15:39- you have, like, ghost agents now. The whole
- 15:43- in some respects, it's I would say it's
- 15:45- turning into the one of the more complicated
- 15:48- things to govern because there's so many different
- 15:51- aspects to it in terms of what agents
- 15:54- can do and how people access them and
- 15:56- how they do things and even how agents
- 15:58- manage agents. You're getting into these things now
- 16:01- where you have agent orchestrators that manage
- 16:03- sub agents.
- 16:05- And agent governance, I feel like, is it's
- 16:08- already started, but I feel like it's just
- 16:09- gonna continue to explode
- 16:12- over the next, I don't know, months, years
- 16:15- as you try to figure out how to
- 16:17- govern this? So it's a very broad area.
- 16:19- I think in and it's not probably unfamiliar
- 16:22- to most administrators. Like, you're used to governing
- 16:25- specific aspects of your platform. So let's say
- 16:28- maybe, like, app consent.
- 16:30- Hey. I created a new I needed a
- 16:32- new enterprise app registration created that that came
- 16:35- with an owner, came with a specific set
- 16:37- of permissions
- 16:39- on the graph, things like that. Maybe you
- 16:41- had a Teams app, again that had its
- 16:43- own kind of construct. You had SharePoint that
- 16:46- had its own construct. So we're very used
- 16:48- to dealing with the individual constructs,
- 16:51- but when they start to get mishmashed into
- 16:53- this one
- 16:55- broader bucket, which has a whole bunch of
- 16:57- different subcomponents,
- 16:58- it becomes
- 17:00- really hard. And I think this is where,
- 17:02- like, Microsoft comes in with things like Agent
- 17:04- three sixty five and says, hey. You need
- 17:06- more of a control plane here for agents
- 17:09- because it's an identity
- 17:10- problem. So it's not just who is the
- 17:13- agent, it's who and or what is the
- 17:16- agent. Like, who created it is one question.
- 17:19- Who can use it or who consume it
- 17:21- is another question. What permissions does it have?
- 17:24- Does it have different permissions and different modes?
- 17:26- K. That that that's just identity. Right? You
- 17:29- have five questions right there.
- 17:31- You've got this
- 17:33- data access issue
- 17:35- or kinda
- 17:36- lack of visibility or something that you need
- 17:37- to kinda figure out is,
- 17:39- what's the data that these agents can retrieve?
- 17:42- What are they allowed to do with that
- 17:43- data? Do they transform it? Do they pass
- 17:45- it on to another agent or another process?
- 17:48- Are they just summarizing it? Like, when they
- 17:50- go and do the retrieval,
- 17:52- let's say you retrieved a bunch of sensitive
- 17:54- business data, and then you're telling that agent
- 17:56- to summarize it in a table. Well, it
- 17:58- might be interacting with, like, an LLM, so
- 18:01- it could be chat GPT
- 18:03- or an anthropic model or deep seek or
- 18:06- whatever
- 18:07- out out on the sides. Now you gotta
- 18:08- worry about not only what data it can
- 18:11- retrieve, what it can summarize, what it can
- 18:12- transform, but you also have to worry about
- 18:14- leaks, like like, pretty big.
- 18:16- You have an application problem
- 18:18- because you don't always have visibility into
- 18:22- not only, like, what are the connectors and
- 18:24- tools and actions that these agents use, but,
- 18:26- also, like, what are the raw APIs?
- 18:28- I've seen some things when I'm interacting with
- 18:31- them because the LLM guides them into it.
- 18:33- They'll say like, oh, hey. I don't see
- 18:35- a tool for blah blah blah, but I
- 18:36- read the documentation for this thing, and I
- 18:38- think it's got a REST API. I'm just
- 18:40- gonna, like, go wild and
- 18:42- see what I can get out of it
- 18:43- with, like, an external service call. And, again,
- 18:45- you're back to data leaks, all that kind
- 18:47- of stuff. You've got a life cycle problem.
- 18:50- So
- 18:51- we had the identity problem of, like, who
- 18:53- created it, who can use it. There's also
- 18:54- a life cycle problem in who approves it,
- 18:57- who owns it, like, how do you attribute
- 18:59- ownership to an agent,
- 19:02- when and or if should it be retired,
- 19:05- what happens when the person who made the
- 19:06- agent, like, if it's a custom agent, what
- 19:08- happens when they leave your organization? That's a
- 19:10- big problem with things like power automate flows
- 19:13- and everything else that's out there, but that
- 19:15- was just a single space. Now you're in
- 19:17- a multispace kinda thing. And then there's just
- 19:19- all the security stuff. So
- 19:21- bad actors, malicious prompts,
- 19:24- prompt injection,
- 19:25- risky connectors,
- 19:29- maybe there's, like, a mismatch in privileges and
- 19:32- configuration of an agent, so you get this,
- 19:34- like, creep aspect
- 19:35- that goes in there. So so it's really
- 19:37- hard because
- 19:39- the question,
- 19:40- I think, used to be, like, what are
- 19:42- the apps that have access to my tenant?
- 19:45- And you could kinda go scope that down,
- 19:47- and you could run-in,
- 19:48- and you could start to wrap your head
- 19:50- around that.
- 19:51- But now the question
- 19:53- is
- 19:54- really different for administrators
- 19:57- in these environments where you have to ask
- 19:59- not only what are the apps, but now
- 20:01- it shifts to
- 20:02- what agents
- 20:03- can access my tenant, what agents can act
- 20:07- across my tenant. So it's not just about
- 20:09- access in a single area, but it's also
- 20:11- about, like, across the spread of tools, Teams,
- 20:14- SharePoint,
- 20:15- Exchange, all those things. What are the tools
- 20:18- they expose? What are the tools they actually
- 20:21- use?
- 20:21- And what's the business data that they have
- 20:24- access to and that they can
- 20:27- reason over? So it just makes it so
- 20:29- much more of, like, a heady thing. I
- 20:31- think Microsoft is approaching this in a manner
- 20:34- of saying, hey. We have things like Microsoft
- 20:37- Agent three sixty five, which start to lay
- 20:40- a groundwork
- 20:41- for a control plane for
- 20:43- management
- 20:44- of agents
- 20:46- to help you kinda get there. So it's
- 20:49- a path forward to observe,
- 20:52- have some measurable governance,
- 20:55- some
- 20:56- sense of
- 20:57- security with a agentic use in your organization.
- 21:01- But even then, I think there's a lot
- 21:03- of room to grow and a lot of
- 21:05- space to
- 21:06- kinda pick things up because as much visibility
- 21:09- as things like Agent three sixty five give
- 21:11- you today, they don't give you a full
- 21:12- three sixty view yet. And I think some
- 21:14- of it is gonna be how do people
- 21:18- start
- 21:19- trying to get that view, and are people
- 21:22- going to adopt agent three sixty five? Because
- 21:24- to your point, it's going to pull in
- 21:26- a lot of things. Like, I go to
- 21:28- my agent three sixty five today. Out of
- 21:30- the box, I haven't done anything special, and
- 21:32- I have 200
- 21:34- to your point about exponential growth, I have
- 21:36- 295
- 21:37- agents
- 21:38- in my organization.
- 21:40- These are not And your organization,
- 21:41- your tenant is tiny as far as, like,
- 21:44- the number of users.
- 21:45- It's not it's not an it's not like
- 21:48- a thousand person org. Mix of things. Like,
- 21:51- I've done some stuff with Dynamics three sixty
- 21:53- five. So I have a bunch of Dynamics
- 21:54- three sixty five agents.
- 21:56- And then I do have a few custom
- 21:57- agents. I have the Microsoft agents, like the
- 22:00- learn agent
- 22:01- and
- 22:03- the planner agent
- 22:05- and the SharePoint admin agent.
- 22:08- And then it does show and this is
- 22:10- an interesting thing that it shows us. These
- 22:12- are just agents that are registered with the
- 22:14- service. It shows things like Databricks
- 22:17- and
- 22:18- Mural and
- 22:21- Canva
- 22:22- and Mural,
- 22:23- where it's agents that are available, it doesn't
- 22:26- necessarily
- 22:27- mean they're connected,
- 22:29- but it means they're registered with the service.
- 22:32- But in order for agents to continue being
- 22:35- registered as people maybe do create ones that
- 22:37- they're using,
- 22:38- there's still
- 22:39- SDKs and different things that custom agents,
- 22:43- depending on where they're created,
- 22:45- there are certain things that have to be
- 22:47- done with those agents
- 22:48- to get them to show up in agent
- 22:50- three sixty five. If someone goes off and
- 22:52- creates a certain agent somewhere and doesn't tie
- 22:55- create the hooks to tie it into agent
- 22:57- three sixty five, it may not show up.
- 22:59- There are some things that are still in
- 23:01- preview, like registry
- 23:03- sync, which does allow you to start
- 23:07- connecting supported
- 23:08- third party AI platforms and syncing
- 23:11- third party agents
- 23:13- in. So, like, if you have agents out
- 23:15- on Amazon Bedrock,
- 23:17- Google Vertex,
- 23:19- Salesforce, Agentforce,
- 23:20- Databricks Genie,
- 23:22- you can set those up to start syncing
- 23:24- those agents into agent three sixty five two.
- 23:26- But there's there is still also a level
- 23:29- of adoption of here's the tool, here's,
- 23:33- not a framework, but a platform we've given
- 23:35- you to manage agents. There's still some of
- 23:37- that onus on organizations
- 23:39- to make sure all those agents are actually
- 23:41- getting in there. It's not the kinda go
- 23:43- click a bot and and set it and
- 23:46- forget it
- 23:47- kind of thing that you need to kinda
- 23:49- get in and figure out. Because it's I
- 23:52- think it was hard enough to manage the
- 23:54- tenant and the sprawl and everything else that
- 23:58- came along with it, and now it's just
- 24:01- that much more.
- 24:07- This episode is brought to you by Trusted
- 24:09- Tech. If you're managing Microsoft three sixty five,
- 24:12- you already know licensing isn't simple. E three
- 24:16- versus e five, the brand new e seven
- 24:18- licenses,
- 24:19- Copilot security bundles, and with price changes coming
- 24:22- in July 2026,
- 24:23- the stakes are getting higher. Trusted tech helps
- 24:26- IT leaders make sense of their Microsoft three
- 24:29- sixty five environment with a free licensing consultation.
- 24:32- Their engineers analyze what you're licensed for and
- 24:35- what you're actually using and where you can
- 24:37- optimize, whether that means consolidating security tools, preparing
- 24:41- for copilot, or eliminating wasted
- 24:44- spend. Plus, they offer proactive and reactive support,
- 24:48- which has been awarded yet another solution partner
- 24:51- designation from Microsoft
- 24:53- for support services.
- 24:54- If you want a clear data backed plan
- 24:57- that has helped over seven five hundred subscribing
- 25:00- customers save up to 20% on Microsoft three
- 25:03- sixty five before your next renewal, visit trustedtech.team/mscloudprom365
- 25:12- and schedule
- 25:25- basics
- 25:26- and maybe take a step back and say,
- 25:27- like, hey. Like, what can you start to
- 25:30- do with agent management
- 25:33- in the Microsoft
- 25:35- three sixty five admin center? So when you
- 25:38- go in for
- 25:40- kind of this govern
- 25:41- governance aspect of
- 25:44- agent three sixty five, which I don't know.
- 25:47- It's a weird tongue twister. I keep wrapped
- 25:48- in my head and getting wrapped on it.
- 25:50- So for me, the first thing I do
- 25:52- when I'm going into
- 25:53- a
- 25:54- client or even in my own environment is,
- 25:57- I would say, where I think most people
- 25:59- should start, where I usually start, is actually
- 26:02- looking at
- 26:04- all agents and looking at that agent registry.
- 26:06- And I will say if you don't have
- 26:08- agent three sixty five licensed, I've compared this
- 26:11- across a couple platforms.
- 26:13- You get some of this. You don't get
- 26:15- the full
- 26:16- you don't get the full report. So if
- 26:18- I go into agent three sixty five and
- 26:20- go look at all agents, it gives me
- 26:23- the name of the agent
- 26:24- if it's available. So am I actually making
- 26:26- this agent available to my users? The platform
- 26:29- that it came from, so is it Copilot
- 26:31- Studio?
- 26:33- Is it third parties are just blank? Is
- 26:36- it Foundry?
- 26:37- Is it some of those other ones? But
- 26:39- then I do start seeing interesting things. And
- 26:42- this is the part that's always fascinating to
- 26:44- me is I can see the risk of
- 26:46- the agent based on some criteria that Microsoft
- 26:49- has established.
- 26:50- I can see the active users, and I
- 26:52- can see the total sessions.
- 26:54- So one of those very interesting things with
- 26:56- me is I just go sort this agent
- 27:00- registry
- 27:01- by active users or by total sessions. So
- 27:04- I can go in and see that
- 27:06- I have
- 27:07- 50 people
- 27:08- using
- 27:09- this HubSpot
- 27:10- agent that are active users. And across those
- 27:13- 50 users, maybe I have 500
- 27:16- total sessions. It starts giving me just some
- 27:19- of that before I even do anything else,
- 27:21- it gives me some of that big picture
- 27:24- of
- 27:25- what agents
- 27:26- have people started
- 27:28- adding. Because
- 27:30- you can just go into
- 27:32- Copilot
- 27:33- and start adding agents. Maybe you have some
- 27:36- added, and you wanna know are people actually
- 27:37- using them. It just starts to give you,
- 27:40- again, that picture
- 27:41- of here's what's out there. Here's what people
- 27:44- are using. I had a very different kinda
- 27:47- take when I had stood up a tenant
- 27:49- and was going through and looking at some
- 27:51- of this. So I kind of approached it
- 27:54- from the lens
- 27:55- of,
- 27:56- I I I guess, maybe, like, even
- 27:58- an internal,
- 27:59- like, rubric for
- 28:01- where does the greatest risk lie. So it
- 28:04- wasn't so much about usage. It was more
- 28:06- around
- 28:07- what are the categories of these things. So
- 28:10- if you see, like, teams pop up in
- 28:12- the list, do I need to worry about
- 28:14- teams right now, or do I need to
- 28:17- go worry about
- 28:19- the third party
- 28:20- connector kind kind of thing? So if there
- 28:23- was stuff in there, let's say, that was,
- 28:24- like, published by your organization,
- 28:27- the the those might be a little bit
- 28:28- safer. Right? Like, you've probably published them out
- 28:30- there for a reason.
- 28:31- They're,
- 28:32- like, predictable to you, so you see them
- 28:35- and you know what they are.
- 28:37- There's the whole category
- 28:39- of
- 28:40- custom agents. So somebody in your organization
- 28:44- created a custom agent
- 28:46- in things like Copilot Studio, which there's a
- 28:49- weird new backdoor in the Copilot Studio.
- 28:52- It, like, even if you haven't licensed users
- 28:54- for it after they GA'd co work. So
- 28:57- so maybe there's some things that you wanna
- 28:58- look out for there, or you're just interested
- 29:00- in, like, what are the parts of my
- 29:02- organization that maybe have a little bit more
- 29:03- maturity
- 29:05- and are going and running with these things?
- 29:07- My my threat matrix for things that were,
- 29:09- like, Microsoft agents, so whether it was Dynamics
- 29:12- or
- 29:13- internal component kind of thing. Hey. If it
- 29:15- was developed by Microsoft and it was integrated
- 29:17- with Microsoft, like, that might not be my
- 29:19- my primary concern.
- 29:21- I'd rather bubble up things like the external
- 29:24- agents, so stuff that was created by either
- 29:26- a third party
- 29:28- was maybe created by, like, a developer that
- 29:30- a random business unit of yours
- 29:32- went and spun up because, you know, they
- 29:34- were able to or you didn't have the
- 29:36- existing
- 29:37- governance constructs in place in your
- 29:39- in in your tenant already. So and then
- 29:41- from there, dial into usage and see. So,
- 29:44- like, if I said, hey.
- 29:46- The external
- 29:47- or the, yeah, external partner agents and maybe
- 29:51- custom agents created, like, within my orgs. Those
- 29:54- are my two biggest, like, factors.
- 29:56- Then I'll go sort by usage and see
- 29:58- if you
- 29:59- can kinda suss out who's doing what there.
- 30:02- And I guess maybe mine isn't necessarily usage,
- 30:04- but it's looking it's looking for similar types
- 30:07- of stuff as you based on usage. That's
- 30:09- how I would freight. That's what I'm looking
- 30:11- more at as again, if co work shows
- 30:13- up at the top of my list of
- 30:14- usage, big deal. If Asana shows up and
- 30:17- I'm like, I don't even know anybody was
- 30:19- using Asana in the organization, and now there's
- 30:21- a bunch of sessions with the Asana agent,
- 30:23- it's why or
- 30:26- something that's risky. The other thing I think
- 30:28- that's fascinating, and I don't know how much
- 30:30- you played with this, is once you find
- 30:32- one of those agents, clicking on it and
- 30:35- looking at all of the details
- 30:37- inside
- 30:39- of each one of those agents. I don't
- 30:41- to me, I think I don't know how
- 30:43- much it's a governance topic as much as
- 30:46- it's a information as you're maybe assessing
- 30:50- the risk of it, or you're assessing
- 30:53- why is this agent there, what are people
- 30:55- doing with it, or if you need to
- 30:57- from the governance perspective,
- 30:58- if you deem need to go in and
- 31:00- start limiting
- 31:01- who can use that agent. Again, it's that
- 31:03- weird mix of now I'm not worried about
- 31:05- just users. I'm worried about APIs
- 31:08- and what they consume. And so now you're
- 31:11- back to the identity, the data access,
- 31:14- the that that whole spread of of things,
- 31:17- and it's just it's a lot of, like,
- 31:19- cognitive load. So you can kinda see what
- 31:21- Microsoft's doing by trying to make it easy
- 31:23- to see. So, hey. Cool. I got a
- 31:25- single pane of glass.
- 31:27- I've seen tenants where it's like, that single
- 31:29- pane of glass is overwhelming. Like like, it's
- 31:32- not helpful. It's more like, oh my gosh.
- 31:34- I'm gonna have a heart attack just having
- 31:35- to, like,
- 31:36- consume this page and trying to figure out
- 31:40- what do
- 31:41- I need to do here,
- 31:44- especially when you're not
- 31:46- maybe
- 31:47- you you haven't read your way all the
- 31:48- way through the documentation or you haven't, like,
- 31:50- fully started to consume these offerings. I think
- 31:52- Microsoft is trying to, like, move the needle.
- 31:55- I don't know if you've had a chance
- 31:56- to play with the the shadow AI component
- 31:59- and some of the stuff that comes out
- 32:01- of Frontier.
- 32:03- So you've got this if you're in the
- 32:05- Frontier
- 32:06- program,
- 32:08- okay, because, hey,
- 32:10- you're on the latest cutting edge of AI
- 32:11- innovation here. It also opens up the shadow
- 32:15- AI piece in the admin center there, and
- 32:17- then that starts to get you
- 32:19- information
- 32:20- about,
- 32:21- like, agentic harnesses in use within your organization.
- 32:24- So let's say you had somebody who went
- 32:26- out and installed, like, OpenClaw, things like that.
- 32:29- I don't know if it can detect, like,
- 32:30- some of the other harnesses, like Hermes
- 32:32- or or things like that, but it's very
- 32:34- much like preview capability, but helps you start
- 32:36- to rationalize
- 32:37- it just a little bit so you're not
- 32:40- having to, like, do everything
- 32:42- your yourself and have some of that overwhelming
- 32:45- stuff. Because if you are in an established
- 32:47- tenant and you were licensed already for Copilot,
- 32:49- like, you had an e seven. Right? Like,
- 32:50- we we've talked about e sevens in the
- 32:52- past. So great. You upgraded those. All of
- 32:54- a sudden, all your users are licensed for
- 32:55- this stuff, and they're just out there, like,
- 32:57- doing whatever. And they've been running that way
- 32:59- for three months, six months, a year if
- 33:01- you've never been in there, like, it's gonna
- 33:03- be a little bit of an overwhelming experience.
- 33:06- I was hoping for more in Shadow AI.
- 33:08- Maybe more It's pretty bare bones today. It's
- 33:10- pretty mine has open claw in it, which
- 33:12- I can start detecting. And then it has
- 33:14- a coming soon. And And under coming soon,
- 33:16- I have Ollama in Po desktop from Quora.
- 33:20- So those are the only I have one
- 33:22- it can detect too that are coming. I'm
- 33:24- surprised
- 33:25- there's not more in there. But, again, we'll
- 33:28- put our disclaimer on there of Frontier as
- 33:31- up and coming. Think of Frontier as, like,
- 33:33- beta preview type stuff. Hopefully, we see more
- 33:36- coming in there soon. Or, like, where does
- 33:38- Scout show? Because Scout's a harness, but it's
- 33:40- not Shadow AI.
- 33:42- But there's no other, like, dedicated
- 33:44- pillar category,
- 33:45- yeah, for
- 33:47- for harnesses
- 33:48- or or things like that. So because Scout
- 33:50- doesn't even show yeah. Scout shows up nowhere
- 33:53- right now in here. It's a weird world.
- 33:54- So maybe we should talk a little bit,
- 33:56- in the couple minutes we have left, about
- 33:59- kind of visibility into agents
- 34:01- and what you can see about them. Because
- 34:03- we kinda talked about high level. You can
- 34:05- get in. You can see what agents are
- 34:07- in use, the categories,
- 34:09- things like that. But then when you actually
- 34:11- have an agent that's out there, you can
- 34:13- get a little bit of high level detail
- 34:15- about what the agent is. You get insights
- 34:18- into the users,
- 34:19- potentially data and tools, permissions
- 34:22- for that agent, activity, things like that. So
- 34:26- so maybe we hop through a couple of
- 34:27- those just for folks who can get a
- 34:29- sense of what they can dig into. To
- 34:31- me, this
- 34:32- is super helpful when you're trying to dig
- 34:35- into it. Like, the details initially
- 34:38- gives you things like when it was created.
- 34:40- Has it been updated? Agents get updated over
- 34:42- time as you build them. Who published them?
- 34:45- Owners. What channels are they available in? So
- 34:48- I'll do Asana, for example. I use Asana.
- 34:51- So it's available for me in Copilot. It's
- 34:53- available in Outlook. It's available in Teams or
- 34:55- the different channels. So where can people access
- 34:58- it? Details really gives you a good overview.
- 35:00- It even gives you, like, if there's
- 35:02- instructions
- 35:03- with that agent,
- 35:05- if it's tied to an environment, this is
- 35:06- gonna be more applicable to, like, a Copilot
- 35:09- Studio agent, which environment is it in, just
- 35:12- a good general overview
- 35:14- of that agent.
- 35:16- Users
- 35:17- is more controlling it. So we talked about,
- 35:19- like, again, from an application perspective, what does
- 35:21- the agent have access to? Users is who
- 35:24- is it installed for and who is it
- 35:26- available to or who is it shared with.
- 35:30- So you can like, if you're a developer
- 35:31- and you've created it, you can just share
- 35:34- a an agent with somebody. But then available
- 35:36- to is who can actually install Asana? Should
- 35:39- it be
- 35:41- shown to all your users or are there
- 35:42- just specific groups that should be allowed to
- 35:45- install Asana?
- 35:46- And then installed for is do you actually
- 35:49- wanna pre install it? All my users are
- 35:50- gonna get Asana by default, So it's a
- 35:53- little bit for me, it relates very closely
- 35:55- to Intune when you have, like, required versus
- 35:57- available. Installed is this is required. Everybody gets
- 36:01- it. Available is they don't have to have
- 36:03- it, but if they want it, they have
- 36:05- the ability to go add it to Copilot.
- 36:07- And that's back to the, like, analogy about
- 36:09- applications. Right? So so age agents are applications,
- 36:12- but they're applications
- 36:13- with, like, just a broader set of things
- 36:16- that that you need to go look at.
- 36:18- The other interesting thing
- 36:20- about, like, the that that analogy of apps
- 36:22- to agents
- 36:24- to carry it forward
- 36:25- is you've also got options around, like, agent
- 36:28- availability and installation.
- 36:30- So there's a difference between
- 36:32- making an agent available
- 36:34- and letting users
- 36:35- discover it versus
- 36:38- pre installing or pushing an agent or agents
- 36:42- out to a security group or a set
- 36:45- of folks that that are out there. So
- 36:47- there's this kinda like, I made it available
- 36:49- in visibility play, but users have to go
- 36:52- self install, and if they install, great, they
- 36:54- needed it, versus
- 36:56- I've made this agent
- 36:58- or this agent set of agents, whatever it
- 37:00- is. It's actually, like, part of the user's
- 37:03- job now. Right? Where, like, you you push
- 37:04- it out like you would push out
- 37:07- an application through Intune
- 37:09- or things like that.
- 37:12- You're out of time. Copilot needs to be
- 37:15- deployed ASAP, but is your tenant really ready
- 37:18- for it? Years of data, permissions, and users
- 37:22- lurk in its shadows, ready to be exposed
- 37:24- by AI.
- 37:25- Sharegate Protect sees it all, so you can
- 37:28- find the exposure risks, fix them fast, and
- 37:31- deploy AI with confidence.
- 37:33- Microsoft
- 37:34- three sixty five governance.
- 37:36- We got this. Learn more at sharegate.com/governance.
- 37:45- And then on the other side, again, going
- 37:46- into applications, I'm gonna jump over the permissions
- 37:49- tab. After you've looked at users or you
- 37:51- wanna do that, one of the other things
- 37:53- you have is permissions that are very much
- 37:55- like, if you're looking at those enterprise apps,
- 37:58- where it gives you and, again, Asana as
- 38:00- an example has
- 38:02- channel read basic
- 38:04- all to read channel names and channel descriptions
- 38:07- in Teams. This is where you need to
- 38:08- know, like, your graph permissions because all of
- 38:10- a sudden, you're seeing permissions. Scopes? I'm amazed
- 38:13- they haven't, yeah, like, done a better job
- 38:15- with abstracting the scopes. It's still, like, group
- 38:17- dot read dot all, teams activity dot send.
- 38:20- So So it's all this random stuff. And
- 38:21- it makes sense when you see it, but
- 38:23- does everybody actually know, like, what What that
- 38:26- means? Yeah. Like, what is user dot read
- 38:28- dot all, like, versus
- 38:31- another read dot all scope or permission
- 38:34- set that's out there kinda thing.
- 38:36- So so you end up in this world
- 38:38- of, like, there there's probably
- 38:40- I
- 38:41- my experience has been, like, there's common
- 38:44- permissions or, like, they're broadly applicable. Like, there's
- 38:46- a bunch of stuff out there that has
- 38:48- things like user dot read dot all. So
- 38:50- is, like, is that okay in your organization
- 38:52- that this thing can read your entire orgs?
- 38:54- This thing can? Like, yeah. Maybe. Maybe it's
- 38:56- a requirement of it. Maybe it's not a
- 38:57- requirement of
- 38:58- it,
- 39:00- what whatever it happens to be. And then
- 39:01- you've also got the whole delegated permissions
- 39:04- aspect of sometimes these agents act autonomously,
- 39:08- so the agent is the agent, and sometimes
- 39:10- they act on behalf of user, and sometimes
- 39:13- they do both at the same time.
- 39:15- Yes. And that's one, like, I do there's
- 39:19- a few things I like in here. This
- 39:20- does show delegated versus application there. The other
- 39:23- thing I like here, and I don't think
- 39:25- it shows it when you're dealing with enterprise
- 39:27- apps or app registrations in Entra, is the
- 39:30- last column does show privilege level. So if
- 39:32- I'm looking at channel.readbasic.all,
- 39:35- like, to your point, what does that mean?
- 39:37- Should I be worried about it? Here it
- 39:39- says that's a low privilege level. Okay. So
- 39:42- if it's a low privilege level, I maybe
- 39:44- don't know what it means. It does give
- 39:46- it details. It says it reads channel names
- 39:48- and channel descriptions on behalf of the signed
- 39:50- in user. But then it says privilege low
- 39:52- versus
- 39:53- chat dot read write, that's gonna be a
- 39:55- medium privilege.
- 39:57- So as you're looking through this too, you
- 39:59- can get you still wanna understand them, but
- 40:02- you can get a nice
- 40:03- scan of,
- 40:04- are there a bunch of high privilege ones
- 40:06- in here, or is it primarily just low
- 40:09- privilege?
- 40:10- What are those different privilege levels of those
- 40:12- graph permissions? There's another weird tab. We should
- 40:14- talk about it while we have a couple
- 40:16- minutes left. Agent certification
- 40:18- or certification.
- 40:19- I'm waiting for mine to load. Agent certification.
- 40:22- Publisher
- 40:23- attested.
- 40:24- This is
- 40:25- yeah. This is
- 40:26- one I have some clients that are very
- 40:28- concerned about. Yes.
- 40:30- And it's super important to them
- 40:32- is this gives a whole bunch of details
- 40:36- on
- 40:37- how is this agent certified according to
- 40:40- different
- 40:42- compliance
- 40:43- regulations?
- 40:44- So I have a sign up, so I'm
- 40:45- gonna keep going with it. Is
- 40:47- one you have, is it publisher attested? So
- 40:50- for Asana, is this attested
- 40:52- by
- 40:53- a self
- 40:54- assessment reported
- 40:56- by Asana? So so to be very clear,
- 40:58- self assessment
- 40:59- by the publisher. Publisher. So this is Asana
- 41:03- telling Microsoft.
- 41:04- Even for Microsoft, you'll see Microsoft certified
- 41:08- versus,
- 41:09- yes,
- 41:10- an Asana or somebody else or SAP is
- 41:12- saying, yes, self attestation.
- 41:14- This is an area that I would love
- 41:16- to see more transparency
- 41:18- about
- 41:19- what's actually checked in the self attestation flow.
- 41:21- Like, how does Microsoft verify that? Because there's
- 41:25- callouts in the docs when you go and
- 41:26- and read through this. So let let me
- 41:29- pull it up so I get it right.
- 41:30- So Microsoft three sixty five certification
- 41:32- reflects an independent
- 41:34- security
- 41:35- and compliance assessment that's valid for twelve months
- 41:38- and helps confirm that the agent meets Microsoft
- 41:41- standards for handling tenant data. Great.
- 41:43- What's in that independent security and compliance
- 41:47- assessment? And what is, like like so and
- 41:50- it's self attested for twelve months. What happens
- 41:53- if Asana pushes an update and now they've
- 41:56- broken their compliance all the way? Like like,
- 41:59- did you come back and review that update?
- 42:01- Like, it's a
- 42:03- that one's a little bit of a weird
- 42:06- situation.
- 42:07- And I get where Microsoft is. Like, there's
- 42:09- no, like, broad, like, agent security consortium or
- 42:12- something like that where you can say, here's
- 42:14- the open standards. Here's the way this stuff
- 42:16- goes. But, hopefully, the industry does coalesce there
- 42:19- over time. Like, you'd think they'd have to
- 42:21- for regulated industry and all that. How would
- 42:23- they even go through
- 42:26- certifying it? Because it says right in here
- 42:28- when it's publisher tested that Microsoft makes no
- 42:30- guarantees regarding accuracy.
- 42:32- I mean, I suppose Microsoft could require certain
- 42:35- documentation to be uploaded, but even
- 42:37- when I go in it's
- 42:40- if I'm looking at something, an agent, how
- 42:42- would I even go in and verify that
- 42:45- it meets
- 42:46- HIPAA, that it's
- 42:48- HITRUST, that it's SOC one or SOC two.
- 42:51- Other than just getting
- 42:53- documentation,
- 42:54- like, to get SOC two, I know that
- 42:56- when you have to go through a whole
- 42:57- bunch of stuff, you get those you get
- 42:59- the certification.
- 43:00- I don't know if Microsoft makes them upload
- 43:02- it, but you do get things like HIPAA,
- 43:04- where based on everything I've understood about HIPAA
- 43:07- is
- 43:09- you don't there's no stamp of approval of
- 43:12- once you do a, b, c, d, e,
- 43:13- f, g, you're HIPAA compliant. It's a lot
- 43:16- of focused on you need to make your
- 43:19- best effort to protect this type of documentation,
- 43:22- and one person's best effort might differ from
- 43:25- another person's
- 43:26- best effort. And it's really hard to actually
- 43:30- stamp something. Like, even Microsoft three sixty five.
- 43:33- Microsoft three sixty five is HIPAA compliant.
- 43:35- If you go in and do the right
- 43:36- stuff,
- 43:37- is a sign of the same way when
- 43:39- they say they're HIPAA compliant or there are
- 43:40- other certain configurations you need to make. It
- 43:43- it's to your point, it's kinda
- 43:46- vague. It's some of these can be hard.
- 43:48- Some of them are a little bit more
- 43:49- straightforward.
- 43:50- I think I said, I hope it's one
- 43:51- of those areas where, like, the industry coalesces
- 43:54- over time. Because today, it's
- 43:56- it's a little bit of, like, a gut
- 43:57- feeling of, alright. So I know who built
- 44:00- the agents. Good. I was able to identify
- 44:02- that. And, like, how much do I trust
- 44:05- that person
- 44:07- or that publisher?
- 44:08- And a little bit of light, like, what's
- 44:10- it been validated
- 44:11- against kind of thing. It's I think it's
- 44:13- no different probably than or or maybe analogous
- 44:17- to,
- 44:18- like, app stores. Right? Like, Google Play
- 44:20- or Apple running the
- 44:22- iOS or Mac app stores. Right? Like, hey.
- 44:25- Customer signed it. They came in. They pushed
- 44:27- it out, but that doesn't always
- 44:29- prevent malicious apps or other things
- 44:33- that that are coming through. That might be
- 44:35- a fun day, right, when the very first,
- 44:36- like, malicious, like, broad malicious agent comes through
- 44:39- And somebody's like, oh, hey. I gotta, like,
- 44:40- yank this agent out of every tenant, kinda
- 44:42- like publishers do with app stores and
- 44:45- and things like that. Conditional access for agents,
- 44:48- risky agent access policies and conditional access, Scott.
- 44:52- All sorts of things. So, yeah, so I
- 44:54- I think this is, like, just to wrap
- 44:56- up, like, this is an area for folks
- 44:57- to
- 44:58- go and look at, especially if you're in
- 45:01- licensed environments and maybe you haven't had the
- 45:03- time to dig into it. Like, you should
- 45:04- have access, I believe, to all this if
- 45:06- you're on e sevens and kind kinda latest
- 45:09- and greatest there. There are ways to license
- 45:12- and get in. So if your tenants are
- 45:13- configured for things like Copilot,
- 45:16- for Cowork,
- 45:17- and these things like that, it's worth at
- 45:20- least browsing through
- 45:22- and seeing what's out there if you haven't
- 45:24- been in there. And, hopefully, no, like, big,
- 45:26- like, red flags or overreactive
- 45:28- decisions
- 45:29- to take, but having some of that visibility
- 45:31- and starting to
- 45:33- kinda build some of that confidence because all
- 45:35- like you said, you're going from managing, like,
- 45:36- these individual constructs to managing a whole bunch
- 45:39- of bundled things at once.
- 45:41- It's a lot more to
- 45:43- wrap your head around
- 45:45- and stay on top of, and then you're
- 45:47- very much trusting in vendors
- 45:49- and folks like Microsoft
- 45:51- to kinda give you the tools with the
- 45:53- insights
- 45:54- and visibility that you need. Yep. And really
- 45:56- starting to think about what does agent governance
- 45:58- look like at your organization. What is
- 46:01- how are you gonna start not just looking
- 46:03- through this, but how are you gonna start
- 46:05- managing those agents?
- 46:06- What are people allowed to do? I feel
- 46:07- like it's also one of those topics that
- 46:09- a lot of people have even put much
- 46:11- thought into yet is what is agent governance
- 46:14- look like for our organization?
- 46:15- It's a rapidly
- 46:17- evolving thing. And probably even more important,
- 46:20- and I'm sure we'll wrap back around on
- 46:22- this topic sometime in the future,
- 46:24- but as all this stuff's transitioning
- 46:26- to
- 46:28- consumption based billing versus kinda like the broader
- 46:32- the broader unlimited usage.
- 46:34- Now there's also the aspect of, hey. That
- 46:37- thing that you deployed to your users six
- 46:39- months ago that, I don't know, let's say
- 46:41- it costs $20 in usage per user per
- 46:44- month, like GitHub Copilot, for example,
- 46:47- all of a
- 46:48- sudden could be costing,
- 46:50- like, thousands,
- 46:51- tens of thousands. I've seen some really wonky
- 46:54- bills and and threads flying around on the
- 46:56- Reddits of the world and and things like
- 46:58- that where some of this stuff is going
- 47:00- off the rails for folks. So there there's
- 47:02- also a little bit of, like, a cost
- 47:04- management play for organizations
- 47:06- as to, like, who's using what, when are
- 47:08- they using it, things
- 47:10- in the world of certainly more in the
- 47:11- world of consumption billing than it kinda was
- 47:14- in the past a couple of months building
- 47:16- up to this. For sure. Cost management is
- 47:18- one that'll be another topic for another day.
- 47:21- I started playing with it, though, because
- 47:23- there's a bunch of stuff that's going to
- 47:25- change. We'll have to fit that oh, by
- 47:28- the time we come out with a new
- 47:30- episode, it will have already been implemented. There's
- 47:31- a bunch of stuff that changes on July
- 47:33- 1 in terms of availability if you don't
- 47:35- have some of that set up. The world
- 47:37- continues to change. Another topic, another day, Scott.
- 47:39- But for now Another topic, another day. All
- 47:41- good. I'm gonna go to a graduation, open
- 47:43- house, and hang out with my family up
- 47:45- in Michigan. Alright. Good luck. Have fun on
- 47:47- the Peninsula, Ben. Alright. Thanks. I will do
- 47:50- that, and enjoy your weekend, and we'll talk
- 47:51- to you soon.
- 47:54- If you enjoyed the podcast, go leave us
- 47:56- a five star rating in iTunes. It helps
- 47:59- to get the word out so more IT
- 48:00- pros can learn about Office three sixty five
- 48:03- and Azure.
- 48:04- If you have any questions you want us
- 48:05- to address on the show, or feedback about
- 48:08- the show, feel free to reach out via
- 48:10- our website, Twitter, or Facebook.
- 48:12- Thanks again for listening, and have a great
- 48:14- day.