Latest / Microsoft Cloud IT Pro Podcast / Episode 432: AI-Era Consulting Billing for Microsoft Cloud Delivery
Transcript
- 0:03- Welcome to episode 432
- 0:05- of the Microsoft Cloud IT Pro podcast recorded
- 0:09- live on 07/13/2026.
- 0:12- This is a show about Microsoft three sixty
- 0:14- five and Azure from the perspective of IT
- 0:16- pros and end users, where we discuss a
- 0:18- topic or recent news and how it relates
- 0:21- to you. In this episode, Ben and Scott
- 0:23- discuss, deliberate, and pontificate on how consulting is
- 0:27- changing
- 0:28- as AI continues to grow. Part of that
- 0:31- is AI doesn't eliminate Microsoft three sixty five
- 0:34- and Azure consulting, but it does change how
- 0:36- you do it and what you may do
- 0:38- for your clients.
- 0:40- Enjoy the show.
- 0:43- Do you want to introduce our topic? All
- 0:45- right. James Fletcher (3five thirty three): There was
- 0:46- an article that I saw come out a
- 0:49- couple weeks ago and I had sent it
- 0:51- to you.
- 0:52- And
- 0:53- the
- 0:55- thought process
- 0:56- and
- 0:58- premise of that article, it came from The
- 1:00- Wall Street Journal,
- 1:01- and it's all about
- 1:04- the shift in billing models
- 1:07- for consultants
- 1:08- in the age of AI.
- 1:10- So as AI continues to mature,
- 1:13- how do
- 1:15- MSPs,
- 1:17- folks like you,
- 1:18- small consultancies,
- 1:20- even larger consultancies,
- 1:22- how do you think about changing your business
- 1:25- model
- 1:27- in the world of AI? So I thought
- 1:29- that would be an interesting one to cover
- 1:31- because I think it's been probably hitting you
- 1:33- a little bit in your business from what
- 1:35- we discussed, so it's kind of close and
- 1:38- near and dear. But I think there's a
- 1:39- bunch of different ways that we can talk
- 1:41- about and talk through
- 1:44- how folks
- 1:45- can think about shifting their businesses
- 1:48- and their business models around M365
- 1:52- and Azure
- 1:53- in particular,
- 1:55- and what needs to potentially
- 1:58- change in a delivery
- 2:01- along the way. So I guess one of
- 2:03- the things that I said, I saw this
- 2:04- article, it kind of resonated with me having
- 2:06- been a former consultant, kinda being close
- 2:09- to the work that you do and what
- 2:11- was going on. It's like, oh, this
- 2:13- does
- 2:14- make sense. The the funny thing to me
- 2:16- as I was reading it was it kinda
- 2:18- sounded like part of the world that I
- 2:20- used to do when I was in consulting.
- 2:22- So I used to work for a couple
- 2:25- of different places where
- 2:28- it was a mix of delivery models from
- 2:30- hourly billing
- 2:31- to fixed bid
- 2:33- or more, like, structured
- 2:35- statement of work. Like, hey.
- 2:38- Let's see.
- 2:39- We'll come in and install
- 2:41- an on premises SharePoint environment for you kind
- 2:44- of thing, and we think that's gonna take
- 2:46- this long with this many people for this
- 2:48- size of environment. We'll configure it in this
- 2:50- way, and that's gonna cost you
- 2:53- n
- 2:53- dollars
- 2:55- all rolled up. So if we come in
- 2:57- under hours, great. We make a little extra
- 2:59- money. If we come in over hours, well,
- 3:01- that's on us. We eat it kind of
- 3:03- thing. So I'm pretty used to that world
- 3:06- of fixed bid SAU's, which
- 3:09- can be tough to do and kind of
- 3:12- hit or miss in the cloud because
- 3:14- if you're doing something like, let's say M365
- 3:17- governance,
- 3:18- well, like what does that really mean? And
- 3:19- what are we going to do? And what
- 3:20- are we going to dial into? Is it
- 3:22- better to be hourly
- 3:24- and sell a fixed number of hours? Can
- 3:26- you get customers to do that? Or
- 3:29- if not, how do you start to think
- 3:30- about some of these things and pivot them
- 3:33- to more
- 3:35- outcome driven deliveries
- 3:36- and being able to clearly articulate those outcomes
- 3:40- for
- 3:41- customers?
- 3:43- And I think it is changing for customers.
- 3:45- Right? Like, they they're looking at these things.
- 3:48- Maybe they think they can vibe code them
- 3:50- quicker.
- 3:51- Maybe,
- 3:52- hey. If I've already deployed Copilot and Microsoft
- 3:54- is
- 3:55- pushing
- 3:56- the Copilot train at me,
- 3:59- Why can't I just ask Copilot how to
- 4:00- configure my environment? Isn't it supposed to tell
- 4:02- me how to do those things?
- 4:04- One would argue that it is supposed to
- 4:06- tell you, but it is telling you the
- 4:07- right thing
- 4:09- and and what to do along the way.
- 4:12- Maybe, but also
- 4:14- maybe not.
- 4:15- So a TLDR.
- 4:17- Hourly billing, I think, is a thing that's
- 4:19- changing. It's something that folks need to think
- 4:21- about,
- 4:22- especially if you're out there consulting. We said
- 4:25- whether you're working for a company, delivering consulting
- 4:28- services, maybe you're an MSP, things like that,
- 4:32- and you have a
- 4:33- consulting bucket of hours, that world's probably changing.
- 4:37- But my hypothesis, or I guess my argument
- 4:40- would be that
- 4:42- the AI thing doesn't eliminate the need for
- 4:46- Microsoft three sixty five or Azure consulting,
- 4:50- but it does change
- 4:51- what customers think they're buying.
- 4:53- So what customers are buying
- 4:56- more and more, and arguably they've been buying
- 4:58- this the whole time, although it's been abstracted
- 5:00- away in different ways, is outcomes. Like, they
- 5:02- wanna get to a certain place
- 5:06- with a certain thing. So in that old
- 5:08- world of
- 5:10- just walking in and saying, Well, pay me
- 5:13- hourly and we'll see how far we get,
- 5:15- kind of thing,
- 5:16- they were buying your expert hours.
- 5:19- They were buying your expertise and everything along
- 5:21- the way. I think in that new world
- 5:23- of we wanna get to outcomes
- 5:25- or we wanna get to
- 5:27- measurable
- 5:28- success criteria,
- 5:30- that's gonna be things like,
- 5:33- let's go ahead and get us ready for
- 5:35- m three sixty five Copilot, but not just
- 5:37- get us ready in these specific dimensions and
- 5:40- dial those in.
- 5:42- Could you do things
- 5:44- like
- 5:46- rather than maybe working in, like, the well
- 5:47- architected framework
- 5:49- in
- 5:50- Azure
- 5:51- land, maybe you start doing things like wrapping
- 5:53- up landing zone deployments as fixed packages so
- 5:56- that you get that
- 5:58- fixed set of outcomes for deploying a governed,
- 6:01- a well governed
- 6:03- Azure landing zone implementation
- 6:05- for workload
- 6:06- x, y, z kind of thing.
- 6:08- I think the other space that's kind of
- 6:10- interesting, you and I were chatting about this
- 6:11- as we're kind of prepping, is stuff like
- 6:14- secure scores and your defender for cloud posture
- 6:18- and all those kinds of things. Could you
- 6:20- offer things that say, hey.
- 6:22- Take your secure score from x to y,
- 6:26- from this number to this number? Or not
- 6:28- from just x to y, what if we
- 6:29- did, like, x y z and you had
- 6:31- a good better best offering thing that you
- 6:34- could
- 6:35- put out there?
- 6:37- So anyhoo, we
- 6:39- should talk through this space
- 6:41- and maybe
- 6:42- why M365
- 6:43- and Azure are
- 6:47- good test beds or good places to go
- 6:49- do, like, outcome driven deliveries?
- 6:52- And what are some potential, like, practices or
- 6:55- deliveries that could shape up around
- 6:59- those technologies. Josh Young (zero fifty seven:thirty seven):
- 7:00- That's interesting. And I think the other aspect
- 7:02- that we kind of talked about and how
- 7:04- you touched on it is
- 7:07- how AI can
- 7:08- assist in these and why
- 7:11- AI has
- 7:13- caused people to think about this more because
- 7:15- I would agree from my perspective and some
- 7:17- of what I've seen and how I
- 7:19- execute different projects now, AI has absolutely
- 7:23- changed
- 7:24- how I do it, which I think also
- 7:26- plays an effect on some of
- 7:29- how you start scoping out these different projects.
- 7:32- So this is absolutely
- 7:33- a fascinating
- 7:35- conversation to dig through. Robert Leonard (zero fifty
- 7:37- three:thirty seven): Let's get into it. Where do
- 7:38- you want to start? I mean, and I
- 7:40- know it's near and dear to your heart,
- 7:42- being a consultant and being in this space.
- 7:44- Yeah. It is. And we have a
- 7:47- lot of
- 7:48- notes here and oh, like, this whole topic
- 7:51- is just it's a tough one to unwrap.
- 7:53- I don't
- 7:55- I don't know. I don't know where to
- 7:57- start on this, Scott. Maybe we should ask
- 7:58- AI where to start on this.
- 8:01- Maybe it is. It's the fact that I'm
- 8:03- thinking Why don't we do this? I'm having
- 8:04- the same thoughts you are about how do
- 8:06- I change my offerings. Like, I'll be honest,
- 8:08- and I haven't had conversations with all my
- 8:10- clients yet, so if you're one of my
- 8:11- clients,
- 8:12- there might be a conversation coming about how
- 8:14- I start structuring
- 8:17- my new packages and how I'm going to
- 8:19- adjust
- 8:20- what I offer.
- 8:22- And I would say this came somewhat before
- 8:24- AI, but I think AI has accelerated how
- 8:26- I'm thinking about, like you said, offering more
- 8:29- of these
- 8:30- fixed
- 8:31- these outcome based deliveries is what I'm looking
- 8:34- at.
- 8:35- In the past,
- 8:36- I have absolutely done everything from a T
- 8:38- and M perspective, just straight up time and
- 8:41- materials. And
- 8:42- this isn't a secret. If you go to
- 8:44- my website, you can start seeing it. How
- 8:46- I'm adjusting where it's going to be
- 8:49- projects where it's very much a fixed
- 8:52- I'm not going to say it's necessarily going
- 8:54- to be fixed price, but it's very much
- 8:55- a fixed project. Very well defined statement of
- 8:58- work, this is where we are, this is
- 9:00- where we want to get to, here's
- 9:03- what we're going to deliver,
- 9:05- or doing
- 9:06- a little bit more of shifting to
- 9:10- what I would say more of a fractional
- 9:12- M365
- 9:13- architecture type stuff that's going to
- 9:17- lead to a little bit more of here's
- 9:19- what I can
- 9:21- Here's the value that I can deliver, not
- 9:23- necessarily just call me up when you need
- 9:25- something.
- 9:26- There can be a part of that in
- 9:28- there, and I know you and I have
- 9:29- even talked about how you kind of package
- 9:31- some of that up. Jason Brett (3nine forty
- 9:31- seven): I guess we both agree that it's
- 9:33- changing
- 9:34- and that it's out there. So why don't
- 9:36- we maybe talk through
- 9:38- and start with some of the options
- 9:42- around the different
- 9:44- billing models
- 9:46- and how those shape up. Right? So I
- 9:48- think we both agree that
- 9:51- the world is changing a little bit there.
- 9:53- So you've got a bunch of different pricing
- 9:54- models that you can approach these things with.
- 9:56- There's fixed fee pricing.
- 9:59- There's the outcome based stuff. I think there's
- 10:02- also subscription pricing, and there's an argument to
- 10:05- be made for hybrid. Like, you're still gonna
- 10:07- have a world of hourly billing and
- 10:09- and combining that together.
- 10:12- And then within those segments, like within a
- 10:15- fixed fee, an outcome based, subscription hybrid, wherever
- 10:18- it happens to be,
- 10:20- you've got kind
- 10:21- of the on the ground view or I
- 10:24- guess the reality of what delivery looks like
- 10:27- there.
- 10:28- So that's a bunch of stuff that folks
- 10:30- have already had. Like, when you think about,
- 10:31- like, the MSP space, repeatable IP,
- 10:35- automation,
- 10:36- things like that,
- 10:39- all the, I guess, client prerequisites,
- 10:42- like assessments,
- 10:44- things like that happen.
- 10:45- But there is that new nuance of AI
- 10:48- assisted delivery
- 10:51- that sits
- 10:52- within there. So
- 10:54- I think maybe if we talk through some
- 10:56- of the
- 10:58- delivery models,
- 11:00- those pricing models and how they shape up,
- 11:04- it should form out and kind of bring
- 11:06- it together. Jay Jacobs (3nine fifty seven): If
- 11:07- we start with the AI assisted, and this
- 11:09- is
- 11:10- where I've struggled with some of the pricing
- 11:13- and how you start thinking through these projects,
- 11:16- because
- 11:17- I absolutely
- 11:18- use AI in a lot of my projects.
- 11:20- And there's a couple aspects to this, right?
- 11:23- AI has absolutely made my
- 11:26- deliveries
- 11:27- faster.
- 11:28- And
- 11:30- this is where
- 11:31- some of that comes in where I was
- 11:34- doing fewer hourly, and
- 11:37- AI
- 11:38- enables me, empowers me to do some of
- 11:40- that quicker. AI is very good at some
- 11:42- of the data analysis, some of the security
- 11:45- assessment, some of that stuff that I used
- 11:47- to do,
- 11:48- which I'm still providing the same amount of
- 11:50- value. But as I use
- 11:53- AI, a portion it allows me to do
- 11:55- a portion of it faster, but it also
- 11:57- changes. Like, I absolutely still have to review
- 12:00- some of the AI outcomes. AI still hallucinates.
- 12:03- Surprise. It doesn't get everything right all the
- 12:05- time.
- 12:07- There's still a level of, right, there's still
- 12:09- a level of expertise, like I can use
- 12:11- AI, but part of the ability I have
- 12:14- to use AI when it comes to these
- 12:16- deliveries is the fact that I also
- 12:20- know
- 12:20- what I'm looking for, so so I know
- 12:23- what to ask AI about. I can go
- 12:25- export a whole bunch of data from Microsoft
- 12:27- three sixty five. I can get conditional access
- 12:28- policies, I can get user account security groups,
- 12:31- I can get DLP policies,
- 12:35- but all of that stuff
- 12:36- still has to work together, and you still
- 12:38- have to know what you're looking for and
- 12:40- what to ask AI about. So I think
- 12:42- there's
- 12:43- that level of expertise that as much as
- 12:45- AI can assist with those deliveries,
- 12:48- you still have to have
- 12:50- a certain level of expertise to know how
- 12:52- everything fits together, to know what you're looking
- 12:54- for, to know
- 12:56- where AI is hallucinating,
- 12:58- and to know what data to tell AI
- 13:00- to look at. So I think there's that
- 13:02- aspect of it when it comes to these
- 13:04- AI assisted deliveries. The other thing that comes
- 13:07- into play, guess what if I ask AI
- 13:10- to go look over, like, hundreds of CSV
- 13:13- files with thousands and thousands of lines of
- 13:15- data?
- 13:16- I may get faster, but guess what? AI
- 13:18- is expensive.
- 13:20- The amount of
- 13:22- there's also that aspect to, well, AI may
- 13:24- help me do it faster. It is also
- 13:26- more expensive
- 13:28- for me to do it faster because AI
- 13:30- is not cheap. So this is where I
- 13:32- start thinking through
- 13:34- how do you start setting up these deliveries
- 13:37- and thinking about consulting when it does come
- 13:39- to something like a Microsoft three sixty five
- 13:42- security audit or just an audit of the
- 13:45- overall configuration.
- 13:48- And I would say,
- 13:49- to my earlier point, certain things got faster,
- 13:51- but I've also noticed the projects aren't
- 13:54- because of the new things I have to
- 13:56- pay attention to or some of the things
- 13:58- I'm able to incorporate.
- 14:00- Some of them actually aren't as much faster
- 14:03- as I thought they would be when I
- 14:04- start looking at hours. Ryan Vaughn (3zero thirty
- 14:06- seven): It's interesting. So you've got this,
- 14:08- I think, explosion
- 14:10- of artifacts
- 14:11- that are all of a sudden available to
- 14:13- you. So
- 14:15- assessments become
- 14:16- maybe not quicker to draw out the telemetry,
- 14:19- but certainly to analyze the telemetry for a
- 14:21- given
- 14:22- surface.
- 14:23- So things like assessments go faster.
- 14:26- Your whole
- 14:28- library, I guess, of scripts
- 14:29- and runbooks and things like that begins to
- 14:32- expand
- 14:33- and explode.
- 14:35- I've noticed in
- 14:37- my world even in like the internal things
- 14:39- that I do,
- 14:40- the number of,
- 14:42- I guess,
- 14:43- artifacts that I'm carrying around now for
- 14:46- Bicep modules,
- 14:47- ARM templates,
- 14:50- Azure policies, things like that, it's just started
- 14:53- to kind of like balloon and get a
- 14:54- little bit out of control. Like I'm having
- 14:56- to need to go organize all this stuff
- 14:58- in well system
- 15:00- it
- 15:01- soon.
- 15:03- It's also the world of
- 15:06- Copilot
- 15:08- and prompt packs. So the early stuff that's
- 15:11- been going on with getting good at prompt
- 15:13- engineering, I think that's a set of reusable
- 15:16- IP that everybody's building out on their own.
- 15:18- And you can not only use that to
- 15:20- accelerate your deliveries, but maybe it's also an
- 15:22- offering that you kind of give to
- 15:25- your customers
- 15:27- that are out there.
- 15:28- The other interesting thing that I've seen is
- 15:31- not just the assessment side of things, but
- 15:35- selling the story about what you're trying to
- 15:38- achieve and what the outcomes are. So things
- 15:40- like
- 15:41- more
- 15:42- like executive framed
- 15:44- readouts
- 15:45- or presentations
- 15:47- become
- 15:48- kind of
- 15:49- part of that delivery window, right, and pushing
- 15:52- things up there. So
- 15:54- it's not necessarily a new world, but it
- 15:57- is an additive world. So
- 16:02- if you're in a place where you've been
- 16:03- doing this while and you're a little bit
- 16:04- mature, you've probably been pitching all along that,
- 16:09- let's take,
- 16:10- like, Azure, for example.
- 16:12- You you make this
- 16:14- shift from
- 16:15- we have people who know the portal. Great.
- 16:18- You know how to do click ops. Awesome.
- 16:20- How does that really help me? To something
- 16:22- like,
- 16:23- we have a delivery machine that can actually
- 16:25- reliably improve your environment, and here's what that
- 16:28- machine is. Here's what it does. Here's how
- 16:30- long it's gonna take us
- 16:33- to to get there.
- 16:35- So you've probably had a bunch of these
- 16:36- things all along. So, like, I imagine in
- 16:39- your world, you have, like, a whole
- 16:41- repository
- 16:42- of scripts and things that sit out there.
- 16:45- But now what you have to do is,
- 16:47- you've already built the IP, you have the
- 16:49- IP, you have those things, but now what
- 16:51- you do is you have to get to
- 16:53- the place where you can sell the outcome
- 16:56- and be able to measure the timeline ruthlessly.
- 16:59- If you go and sell
- 17:00- or you've put together an offering that is
- 17:02- effectively like, hey, this is going to take
- 17:04- me an hour to deliver and it ends
- 17:05- up taking me three hours every time, well,
- 17:09- that's no good.
- 17:10- You're not getting
- 17:11- to where you need to be. So I
- 17:13- think as folks are out there and they're
- 17:15- kind of evaluating
- 17:16- all the things
- 17:18- that exist and all the things that you've
- 17:19- built up and what's going on,
- 17:21- you really start to take a step back
- 17:23- and think about, like,
- 17:25- what's the baseline? What are the metrics that
- 17:27- matter here? Like, what is success criteria?
- 17:31- And what's the criteria that
- 17:33- I'm going to sell to my customers
- 17:36- and what are my customers going to take
- 17:38- as evidence
- 17:40- of success
- 17:42- along the way?
- 17:44- And then I think the other thing you
- 17:45- get to sell
- 17:47- is LLMs are good at some things or
- 17:50- great at some things,
- 17:51- but I found especially in M365
- 17:54- land, they love to hallucinate.
- 17:58- Asking an LLM to fire off a WorkIQ
- 18:00- query is sometimes like, Whew, where did you
- 18:03- come up with that thing?
- 18:05- So I think making sure that there's more
- 18:08- human in the loop and expertise around
- 18:10- hallucinated reports
- 18:12- or doing those gut checks, right? So you're
- 18:15- still an expert in your space. You're going
- 18:17- to know, and it's probably not going to
- 18:19- pass the sniff test for you if somebody
- 18:21- goes and runs
- 18:23- a random query with an LLM
- 18:25- against something versus
- 18:27- you just know the way this telemetry comes
- 18:29- together and what can be presented,
- 18:32- out of it kind of thing.
- 18:35- So
- 18:36- you get a couple different
- 18:38- gut checks in there, effectively, like the advisory
- 18:41- stuff,
- 18:42- which is still important. Right? And, like, that's
- 18:45- not
- 18:46- not an inconsequential
- 18:48- thing.
- 18:50- If you're managing file storage today, you've probably
- 18:53- felt the pressure. Refresh cycles are harder to
- 18:55- plan, costs and complexity are increasing, especially as
- 18:59- file workloads scale across regions.
- 19:01- That's why Microsoft customers trust Nasuni. As an
- 19:05- Azure certified software,
- 19:07- Nasuni provides a global file system designed for
- 19:09- enterprise scale environments, keeping file data secure, governed,
- 19:13- and accessible across Azure environments without added infrastructure
- 19:17- complexity.
- 19:18- It's the foundation to keep your unstructured data
- 19:21- stable even as your infrastructure
- 19:24- evolves. To see how much you could save
- 19:26- while keeping your initiatives on track, visit nasuni.com/tco.
- 19:32- That's nasuni.com/tco.
- 19:38- That's another
- 19:39- area that I've kinda found
- 19:42- in talking to different people and where people
- 19:45- are still coming to me for questions is
- 19:47- what you just said, some of the advisory
- 19:49- and the architecture type stuff. And that's where
- 19:54- a lot of that expertise
- 19:55- and what people still need is, I have
- 19:58- found AI is really good at looking at
- 19:59- data. I can go take a whole bunch
- 20:01- of data from Microsoft three sixty five, and
- 20:03- it can tell
- 20:04- me where are
- 20:06- the shortcomings? Where do I have security gaps?
- 20:09- Where do I have potential misconfigurations?
- 20:11- Okay, great. I've identified
- 20:13- all of these things now,
- 20:15- leveraging AI
- 20:16- of what needs to be done. How do
- 20:18- you go actually interface with the users now?
- 20:21- How do you go start
- 20:23- taking everything
- 20:24- that maybe you've identified
- 20:26- and
- 20:27- architecting out a solution to solve all of
- 20:29- that, where you're not going to
- 20:32- impact the business? Where you're figuring out how
- 20:35- do I start piloting this, how do I
- 20:37- make these tweaks, how do I make these
- 20:39- changes, what does this overall architecture look like
- 20:43- that maybe
- 20:44- solve some of this? And I think that's
- 20:46- where there's still a lot of value to
- 20:47- when you are doing the Microsoft three sixty
- 20:49- five
- 20:50- consulting, where you are still doing the Azure
- 20:52- consulting is
- 20:54- it's getting a little bit more into some
- 20:55- of those I don't know if you'd call
- 20:57- them soft skills or the more architecture type
- 21:00- skills, the design skills
- 21:03- that
- 21:04- AI can give you ideas, but AI also
- 21:07- hasn't been in there. Part of the advantage
- 21:09- I've had is customers ask me, Well, how
- 21:11- did you do this with another client? What
- 21:13- worked? What didn't work? From
- 21:15- a working with humans perspective, from a things
- 21:19- you don't necessarily think about in terms of
- 21:22- a deployment plan or where users are going
- 21:24- to get tripped up, all of that type
- 21:26- of stuff,
- 21:28- there's still a lot
- 21:30- of value
- 21:31- in that in the AI world, and I
- 21:32- think where a lot of consultants
- 21:35- start focusing on this AI world where it's
- 21:37- good at producing results, it's good at looking
- 21:39- at data, but it's not always good at
- 21:41- knowing
- 21:42- how is this going to impact users and
- 21:44- how do I work with users
- 21:46- to
- 21:48- react to architect, whatever that may be, to
- 21:51- whatever I discovered and
- 21:54- investigated
- 21:55- using AI as a support tool, as another
- 21:59- tool in my toolbox? Jason Brett (3nine thirty
- 22:00- seven): The expertise of consultants doesn't go away
- 22:03- in this world, would be my
- 22:05- assertion,
- 22:07- I think that's what you're saying too is
- 22:09- there's
- 22:10- a certain degree of the tools that exist
- 22:13- today can tell you
- 22:15- what.
- 22:17- They can tell you how, they can't always
- 22:19- tell you why,
- 22:21- and they can't always tell you how in
- 22:23- context of the overall constraints of an environment.
- 22:26- So if I think about something
- 22:29- like Secure Score,
- 22:31- in isolation
- 22:32- sounds like a great thing. Like, hey, I
- 22:34- wanna go from
- 22:36- what's the max for a secure score these
- 22:37- days, 800, something like that?
- 22:40- Like, I don't know. You sit down and
- 22:42- you look at your secure secure secure score
- 22:44- and you go, oh, I'm 450, and I
- 22:46- wanna go to 800. Well, maybe going to
- 22:48- 800 isn't the best idea because you're gonna
- 22:50- do a lot of things that that maybe
- 22:52- aren't applicable to your environment.
- 22:54- But I guarantee when you go out and
- 22:56- look at your score, it's gonna keep pushing
- 22:58- you insane.
- 23:00- Do more.
- 23:01- Same thing with, like, co pilot readiness. I
- 23:04- think that would be another kind of big
- 23:05- area in this kind of current
- 23:08- phase of M365,
- 23:10- things like this. It's like, you can go
- 23:12- get a Copilot readiness report. Is that readiness
- 23:15- report actually going to get you to the
- 23:17- spot where you can configure your licensing properly,
- 23:20- where you can go and configure
- 23:22- your, not just your environment configuration, but also
- 23:26- your SharePoint sites and everything else that are
- 23:28- out there
- 23:29- in a way that are gonna, like, deliver
- 23:31- the right amount of value for your business?
- 23:35- I don't know.
- 23:37- Probably not. Right? That's still where you're paying
- 23:39- somebody else
- 23:41- for their
- 23:43- expertise
- 23:44- all along the way. So
- 23:46- for something like Secure Score, you don't wanna
- 23:48- have a customer walk in and just think
- 23:50- like, oh, hey. There has to be this
- 23:52- wild movement.
- 23:53- Like, I I need to go from
- 23:56- $4.50 to 800.
- 23:57- No, you don't. Like, you have to go
- 23:59- from maybe a different
- 24:01- a different x to y,
- 24:03- but, like, here's what you're getting out of
- 24:05- that. Like, here's the reduction in risk. Here's
- 24:08- the improvement in your posture
- 24:11- within your organization,
- 24:15- those kinds of things.
- 24:17- So I think that's all
- 24:19- good stuff and it doesn't take away from
- 24:22- the expertise that folks have
- 24:24- built up over the years.
- 24:26- There is that adage, right? When you have
- 24:28- somebody, like a trades person come into your
- 24:30- house, so like, Hey, you get a plumber
- 24:31- and it takes a plumber ten minutes to
- 24:34- fix the leak under your sink, but you
- 24:35- have to pay them for an hour of
- 24:37- their time.
- 24:38- Well, yeah, you're not just paying them for
- 24:40- the hour to pay them for the hour,
- 24:41- you're paying them for their ten, twenty plus
- 24:44- years of experience, and that's how they're able
- 24:46- to get it done in ten minutes. So
- 24:48- that part's not going away either in this
- 24:51- world. Jack Mallers (zero fifty seven:forty seven): I
- 24:52- had that the other day. I tried to
- 24:53- fix my own Well, I did fix my
- 24:54- own shower, Scott. I fixed a leak in
- 24:56- my shower, but it took me eight or
- 24:58- nine hours.
- 25:00- And had I paid a plumber, it probably
- 25:01- would have taken him two because he would
- 25:03- have known all the tricks and this is
- 25:06- how you get this particular part out of
- 25:09- the shower to stop the leak, and I
- 25:11- ended up, like, three trips to Home Depot
- 25:12- and a trip to a friend's house for
- 25:14- tools and all of this. So, yeah, I
- 25:16- think that absolutely still applies to this world
- 25:19- as well. And another one of these, going
- 25:21- back to the Secure Score, I'm working with
- 25:23- a client right now, and they're like, We
- 25:25- need to implement CIS benchmarks across our entire
- 25:28- environment. CIS has a great set of here's
- 25:31- everything you need to do. It's six fifty
- 25:33- seven lines in Excel spreadsheet. Hey, I did
- 25:36- a great job at looking through the PDF
- 25:38- and creating the Excel spreadsheet. It gives me
- 25:41- did a great job ahead of it. Let's
- 25:42- look at all the categories. How can we
- 25:44- maybe lump these settings together?
- 25:48- How can we best
- 25:50- deploy different settings in a logical way so
- 25:53- settings are grouped together logically?
- 25:56- But it sure can't tell me which users
- 25:58- are using which functionality, or how is turning
- 26:01- on this particular setting going to affect users
- 26:03- and finance versus IT,
- 26:06- how are we gonna go through and push
- 26:08- all these 650
- 26:10- settings out to 6,000
- 26:12- users without disrupting the business,
- 26:15- that's where a lot of that value still
- 26:17- comes in. My computer is beeping at me
- 26:19- with the words. It's beeping.
- 26:21- We've talked about this for years too. Right?
- 26:23- Like, when we went from on prem to
- 26:25- cloud, how what you do and maybe how
- 26:27- you do things changes as technology evolves. I
- 26:30- think the age of AI consulting or
- 26:32- AI assisted consulting, we're kind of at that
- 26:34- point now where And we talked about this
- 26:37- on an episode, I can't remember which one,
- 26:38- where a lot of the same stuff, it's
- 26:40- not like jobs as Microsoft three sixty five
- 26:43- consultants, Azure consultants is going to go away,
- 26:45- but you do need to start thinking through
- 26:48- how does your job as a consultant change,
- 26:50- and where are you going to bring
- 26:52- value
- 26:54- now? Because now you do have new tools,
- 26:56- you have new ways you can bring value,
- 26:59- maybe you can save time doing one thing,
- 27:01- but you're going to spend more time bringing
- 27:03- value in another way. There's that continued
- 27:06- evolution in this consulting world. I've been doing
- 27:08- it for twenty five years now, and I
- 27:10- can tell you what I did twenty five
- 27:12- years ago, if I kept doing that today,
- 27:14- probably wouldn't have any work to do.
- 27:16- It's you're always thinking
- 27:19- how do these changing tools, the changing landscape
- 27:23- change what I do and where can I
- 27:25- continue to provide value and
- 27:27- leverage the changing in technology to my benefit?
- 27:30- And I think we're absolutely in that spot
- 27:32- with AI too. Brad Thomas (3three fifty seven):
- 27:33- How much do you think about
- 27:35- in the accelerated delivery world? So you can
- 27:38- create these artifacts
- 27:39- quicker, right? So you can go create,
- 27:42- well, something can go and create
- 27:45- a shell script or a Python thing that's
- 27:47- gonna read your PDF, convert it to Excel,
- 27:49- blah, blah, blah, blah, blah. Do all the
- 27:50- things it needs to do, and it's gonna
- 27:52- push it out there.
- 27:54- Great, you can go faster, but still you
- 27:56- need that human in the loop thing. So
- 27:59- there's now maybe a need for increased review
- 28:03- or more diligence around review.
- 28:06- And yeah, I think
- 28:10- my assertion would be you have to put
- 28:12- that diligence in because
- 28:14- it's
- 28:15- your reputation on the line because now you're
- 28:17- selling outcomes.
- 28:19- You're not just selling hours. It's not like,
- 28:21- oh, hey, Buy ten hours from me, and
- 28:23- if you need more, that's great. If we
- 28:25- don't use
- 28:26- them, whatever.
- 28:28- You're kinda like, in that example of taking,
- 28:30- like, CIS benchmarks and bringing them through to
- 28:32- an environment, like, you're selling an outcome of,
- 28:34- like, CIS
- 28:37- implementation
- 28:39- across these categories or across
- 28:42- 50% of these in this category, 25% over
- 28:45- here, or 100%
- 28:47- of everything that's out there kind of thing.
- 28:49- So So that's another one that I think
- 28:50- is interesting to me is, like, the whole
- 28:53- risk reward thing or the ratio for, like,
- 28:56- reputational
- 28:57- harm. Like, should you just generate something fast
- 29:00- and it goes out there and it doesn't
- 29:02- align to
- 29:05- either the outcomes that customers were purchasing
- 29:09- or it's just like wildly wrong. Right? So
- 29:11- the customer thinks they're getting an outcome and
- 29:13- then they walk away happy and you're like,
- 29:14- oh, great. I delivered it. But it turns
- 29:16- out that nobody actually
- 29:18- got what they wanted or the customer didn't
- 29:21- achieve what they wanted in that world. James
- 29:23- Lavish (zero 50 three:thirty seven): And it doesn't
- 29:24- align. And this is where I've
- 29:26- absolutely spent more diligence.
- 29:29- To your point, I can create an assessment
- 29:31- document, but I do spend more time reading
- 29:34- through it because maybe AI generated a lot
- 29:37- of it. And
- 29:38- I don't want it to have things in
- 29:39- there that are blatantly wrong. To your point
- 29:42- earlier, I'm going to read through it and
- 29:43- make sure I don't have any
- 29:46- To make sure that
- 29:47- my gut instinct about what should be there
- 29:49- and what I know should be there and
- 29:51- what I've seen in the past is correct,
- 29:53- but also taking
- 29:55- some of that and aligning it to what
- 29:57- I know
- 29:58- the customer needs. Because leading up to that,
- 30:00- I'll have phone calls with customers. Oh, we
- 30:02- didn't do this for this and this reason,
- 30:05- or
- 30:06- we tried this and it didn't work for
- 30:08- x, y, and z reason. There's a lot
- 30:10- of intricacies
- 30:12- that
- 30:13- I
- 30:14- may know, that I've written down based on
- 30:16- conversations with clients that
- 30:18- may not necessarily
- 30:19- align with some of these outcomes. So I'm
- 30:21- being very diligent in going through here to
- 30:24- make sure that
- 30:26- maybe what got generated
- 30:28- aligns based on all the other conversations I've
- 30:31- had with the client and all the other,
- 30:33- the history I have with that client, what
- 30:35- I know about their environment, what I know
- 30:37- about their users.
- 30:39- And that's where I think some of the,
- 30:42- where you spend your time shifts
- 30:45- Because you are doing more of that gut
- 30:47- checking, that alignment
- 30:49- after the fact now, as opposed to when
- 30:51- I would maybe run PowerShell scripts
- 30:53- and write the entire document by hand. I
- 30:56- know what's in the document because I wrote
- 30:57- it all, and I'm doing that as I'm
- 31:00- writing it, now I'm just doing it afterwards.
- 31:02- Matthew Piepenburg (3seven 50: It's a shift there.
- 31:05- I wonder
- 31:07- how much you're running into this. So maybe
- 31:09- one last question
- 31:10- we go through to pick your brain here.
- 31:14- So
- 31:15- there's been this thing
- 31:17- I think where
- 31:18- particularly in
- 31:20- the fixed delivery space MSPs especially, you develop
- 31:23- your set of
- 31:25- IP. So you end up with this bank
- 31:27- of scripts,
- 31:28- things like that.
- 31:30- I never ran into customers really asking for,
- 31:33- like, hey, what's the thing you're running? Or
- 31:35- what do you actually run under the hood
- 31:36- to do that assessment? They always just wanted
- 31:38- the assessment at the end. It's like, again,
- 31:40- get me to the outcome
- 31:42- that you told me we're going to get
- 31:44- it to. I wonder, like, are customers kinda
- 31:47- pushing back at all, or do you find
- 31:48- them coming in and
- 31:51- really asking for some kind of
- 31:55- visibility into, like, the the things that you
- 31:58- are running or what you're creating and what's
- 32:00- out there. Because I like, there is a
- 32:01- certain, like, kinda risk matrix as well to
- 32:04- think about, like, oh, hey. I hired this
- 32:05- person to come into my environment. They say
- 32:07- they're gonna run a a PowerShell script or
- 32:09- they wanna install this application on my server
- 32:11- to
- 32:13- be able to do this assessment, or
- 32:15- they're asking for a user account with this
- 32:17- level of permissions so they can go against
- 32:19- the Graph API and get what they need
- 32:20- to be. Like, do you find them asking
- 32:22- for those artifacts for evaluation
- 32:26- or
- 32:27- anything like that? I haven't seen that
- 32:30- change a whole lot in terms of
- 32:33- what they're asking for.
- 32:35- Before this, over the last few years, I
- 32:37- will have customers ask for a script. They're
- 32:39- like, do you have a PowerShell script or
- 32:41- you generated this. Do you have the script
- 32:43- that you used to export all of that?
- 32:46- I haven't seen that changing a whole lot.
- 32:49- Maybe now they're asking
- 32:51- I had one I was talking to recently
- 32:54- that was like, Do you have the skills,
- 32:57- the MD files, the skills that you used
- 32:59- in AI
- 33:00- to
- 33:01- help you generate this?
- 33:04- And that's one of those, to your point,
- 33:05- I think that starts turning into more intellectual
- 33:08- property. I will say in the past, if
- 33:10- I have a client I'm working with
- 33:12- and I generate a certain report form and
- 33:14- they want some of that, this may be
- 33:16- good, may be bad, I tend to share
- 33:18- it with them. They don't need to come
- 33:19- back to me and pay me every three
- 33:20- months to generate the same thing.
- 33:23- I would rather provide value in other ways,
- 33:26- help them with new things rather than just
- 33:29- generating the same outcomes for them over and
- 33:31- over again. Part of that is my personality,
- 33:33- I get bored with that. I don't need
- 33:35- to do the same thing every month if
- 33:37- they want to do it themselves.
- 33:39- I'd rather have them come to me with
- 33:40- their new challenges, their new struggles, and help
- 33:43- them with the new stuff.
- 33:44- But that is one area in terms of
- 33:47- asking for some of the artifacts and stuff
- 33:49- behind the scenes. What they may ask for
- 33:51- has changed, but the level of them asking
- 33:54- or the amount of times they ask, I
- 33:56- haven't seen them asking for it
- 33:58- more than maybe they did in the past.
- 34:00- Josh Young (zero fifty seven:forty seven): Gotcha. I
- 34:01- wonder if that shifts more and more over
- 34:02- time, or if customers are just going out
- 34:05- and creating their own stuff. Right? So
- 34:07- it goes both ways. Like, you can kinda
- 34:09- iterate faster on the front end and then,
- 34:11- therefore, go do these deliveries.
- 34:14- I would imagine that there's also a subset
- 34:16- of customers I can say, like, oh, hey.
- 34:18- Ben gave me this great report on my
- 34:20- environment. It gave me a set of action
- 34:22- items to to go and do. What if
- 34:24- I just, like, backport that report and say,
- 34:26- like, hey. How could I generate this like,
- 34:28- throw that PDF or that Word doc at
- 34:30- an LOM and say, how could I generate
- 34:32- this report myself? Like, go write me a
- 34:33- bunch of scripts, right,
- 34:35- or PowerShell bash, you know, whatever it happens
- 34:37- to be
- 34:39- to spin things up. It will be interesting
- 34:42- to see
- 34:43- how that changes is even what,
- 34:45- I think what IP changes.
- 34:49- This is not me. I had this conversation,
- 34:51- though, recently with
- 34:53- one of my contractors
- 34:54- where there were all these There's a lot
- 34:56- of websites out there where
- 34:58- people were providing a lot of value for
- 35:02- writing PowerShell script. It's like, Here's a PowerShell
- 35:04- script to go
- 35:05- do ABC. Go export all my mailboxes.
- 35:08- Go
- 35:10- bulk make this change on users in Entra.
- 35:13- It's like, That's changed. That's- Ryan Vaughn (3three
- 35:15- thirty seven): Maybe like what Tony Redmond's done
- 35:17- over the years with the Office three sixty
- 35:19- five for IT Pros,
- 35:22- the book, but he's got that whole GitHub
- 35:24- repo, and him and his team of folks
- 35:25- have that GitHub repo out there with tons
- 35:27- of sample scripts. They're often updated for latest
- 35:30- iterations of the graph, things like that
- 35:33- as he's going through and
- 35:35- running into things. So, yeah, those resources exist
- 35:38- and you can go It's super easy these
- 35:40- days.
- 35:41- You don't have to know how Git works.
- 35:43- You can But is there gonna be demand
- 35:45- for those resources
- 35:47- going forward? I think the demand for those
- 35:48- has changed because people are so much quicker
- 35:51- to say,
- 35:52- Claude, chat GPT, go write me a script
- 35:54- for this.
- 35:56- It's interesting to see
- 35:58- or will be interesting to see- Jason Brett
- 36:00- (zero fifty three:thirty seven): I think it's going
- 36:01- to swing back. Joshua Zappia (zero fifty three:thirty
- 36:02- seven): I guess what resources are going to
- 36:03- be in demand
- 36:04- as time continues to progress and as AI
- 36:06- can do certain things. Joshua Zappia (zero fifty
- 36:07- three:thirty seven): That's the whole human in the
- 36:09- loop thing,
- 36:11- and nothing's going to account for
- 36:14- expertise and taste, right, to be able to
- 36:16- put these things together. So
- 36:18- can an LLM go and write you a
- 36:20- PowerShell script? Yes.
- 36:22- Can it interrogate the Graph API for you?
- 36:25- Sure. Is it going to do it the
- 36:27- right way or is it going to bring
- 36:28- back the right set of data?
- 36:31- Sometimes
- 36:33- I think you don't know.
- 36:34- I run into this regularly with
- 36:37- in my world where folks are going and
- 36:41- they're creating these solutions. Like, they look really
- 36:44- pretty and could be a dashboard
- 36:47- or, like, an assessment engine, things like that.
- 36:49- And then you start to dig into some
- 36:50- of the details and you're like,
- 36:52- this feels like it's kinda like not right.
- 36:56- Potentially,
- 36:56- wildly wrong. And you ask them like, well,
- 36:59- how did you get here? Like, what was
- 37:01- the input? Like, can you go tell me,
- 37:04- maybe it's like
- 37:05- a a Kusto query kinda thing. Can you
- 37:07- go tell me the Kusto query? I didn't
- 37:09- write that Kusto query. The LOM wrote it
- 37:10- for me. Well, how do you know if
- 37:11- it's good? How do you know if that's
- 37:13- right? How do you know if it was
- 37:14- the right data source? Well, I don't know.
- 37:15- It just told me. And they kinda take
- 37:17- that as fact. And then you say or
- 37:19- you do things like, oh, hey. Great.
- 37:22- I put the query together, but then
- 37:24- the magical LLM thing decided to wrap it
- 37:26- in a Python script. Okay. Cool. Do you
- 37:28- know Python? Nah. I don't know any Python.
- 37:31- Do you know what it's doing? Nah. I
- 37:32- don't know what it's doing. It's running the
- 37:34- script. It's a script that runs a script
- 37:36- that runs a script.
- 37:38- We might want to think about that one,
- 37:40- right,
- 37:41- and come back along the way. So I
- 37:43- do think there's value in that stuff. If
- 37:45- I think about the world of
- 37:48- hopefully the stuff that the Tony Redmans of
- 37:50- the world are doing, doesn't go away
- 37:53- because I think those are the
- 37:55- more opinionated,
- 37:57- arguably more tasteful solutions
- 38:00- rather than something that's just been spun up
- 38:04- by an LLM or a set of agents,
- 38:07- things like that. Can they get the job
- 38:09- done? Sure. Are they going to do it
- 38:11- as efficiently, as tastefully,
- 38:13- but with the same level of, I think,
- 38:15- care, like, as you're going, like, step by
- 38:17- step through the process?
- 38:19- Probably not.
- 38:21- So, you know, if somebody like Tony can
- 38:23- write a script 10 x faster because he's
- 38:25- got access to an LOM,
- 38:28- that that's pretty cool. Right? Like, hopefully that
- 38:29- leads to, like, like, more out outputs there.
- 38:33- And then folks like us, like the mere
- 38:35- mortals, we can ground our learning and round
- 38:39- our stuff in that. Like, I think you'd
- 38:41- have a way better experience
- 38:43- probably going out to something like their GitHub
- 38:45- repo for that I think they just renamed
- 38:47- the book actually,
- 38:48- Microsoft three sixty five for IT Pros, finally.
- 38:52- But go out to that Git repo and
- 38:53- clone that repo and say, Hey, can you
- 38:56- build me a script that does this, or
- 38:58- can you extend this script? Rather than start
- 39:00- from scratch kind of thing. It's interesting to
- 39:02- see how all this is changing. It keeps
- 39:04- going along the way.
- 39:06- And then like you said, make sure you
- 39:07- charge enough money to cover your AI usage.
- 39:10- Let me tell you, the amount I pay
- 39:11- for AI is continuing to go up partly
- 39:14- due to using it more, partly due to
- 39:15- people changing pricing on me or
- 39:18- stuff coming out of preview. And Pricing will
- 39:20- continue to change on you, by the way.
- 39:22- So
- 39:25- hold on to your checkbook or your wallet.
- 39:27- And as you hold on to it, just
- 39:28- prepare to open it and have the money
- 39:30- fly out. But in the meantime, I have
- 39:31- Fable five for next week now. They did
- 39:33- Fable five for next week.
- 39:35- Yeah.
- 39:36- So, yeah, by the time this comes out,
- 39:38- Fable five may or may not
- 39:40- may or may not still be there.
- 39:42- So CheckTPT
- 39:44- sorry, OpenAI. Codecs
- 39:46- also got rid of the five hour window
- 39:48- limit. And I don't know if you've been
- 39:50- playing around with
- 39:52- 5.6.
- 39:53- So you've got Luna, Terra and SOL,
- 39:55- good, better, best kind of
- 39:57- effort and reasoning, things like that.
- 40:00- 5.6 SOL is pretty good.
- 40:04- Yeah. So
- 40:05- just you have the option and you're looking
- 40:07- out there and you're like, Hey, I'm stuck.
- 40:08- And it's going to be interesting over the
- 40:10- next couple of months, I think,
- 40:13- while the frontier
- 40:16- model trainers
- 40:17- and
- 40:18- AI delivery companies like
- 40:21- Anthropic
- 40:22- and OpenAI, things like that, have to now
- 40:24- compete with local model weights
- 40:27- that are coming out from
- 40:29- China and other places,
- 40:32- and kind of where they sit and how
- 40:34- that stuff comes together. So
- 40:36- we'll see. Can Anthropic keep up and get
- 40:38- rid of the five hour window or the
- 40:40- five hour usage reset window? I don't know.
- 40:42- Be kind of nice.
- 40:44- I didn't I'd enjoy it.
- 40:46- But, you know, I I I see OpenAI
- 40:48- doing some things that make me go like,
- 40:50- oh, hey,
- 40:51- maybe I should go back and
- 40:53- be in that ecosystem. Or it keeps you
- 40:55- kind of like you keep your toe in
- 40:57- the water for both of them, which makes
- 40:59- it really Matthew Gauthier (3seven 30: I don't
- 40:59- keep my toe in the water for both
- 41:01- anymore. I can't. I can't. I'm still there
- 41:03- and I'm still trying.
- 41:06- I'm a Claude fan boy, but like I
- 41:08- said, they're doing some
- 41:11- interesting stuff over in
- 41:14- OpenAI land. Yeah. As long as you avoid
- 41:17- their whole super app crap that they just
- 41:19- came up with last week, That thing's horrible.
- 41:21- Stay away from that. That wrap us up.
- 41:23- Anything else? I think that covers it. We've
- 41:25- been going for, like, forty five minutes now.
- 41:26- I think that covers it. Hopefully, it gives
- 41:28- some folks some
- 41:30- things to noodle on. If you're out there
- 41:33- in the consulting or delivery space and you
- 41:36- have any
- 41:37- opinions or feedback,
- 41:39- go to our website. It's contact form. Let
- 41:41- us know what's going on, or comment on
- 41:44- the podcast on your
- 41:46- platform of choice. Absolutely. And until next time,
- 41:49- Scott. Until next time.
- 41:53- If you enjoyed the podcast, go leave us
- 41:55- a five star rating in iTunes. It helps
- 41:57- to get the word out so more IT
- 41:59- pros can learn about Office three sixty five
- 42:01- and Azure.
- 42:02- If you have any questions you want us
- 42:04- to address on the show or feedback about
- 42:06- the show, feel free to reach out via
- 42:08- our website, Twitter, or Facebook.
- 42:11- Thanks again for listening, and have a great
- 42:13- day.