Episode 521: 50 Developers, Zero Playbook: The Leadership Conversation Nobody's Having

Why AI Costs Feel Unclear

SPEAKER_02 0:05

Welcome everyone to another episode of Dynamics Corner. It's an exciting episode because I don't know the cost of AI and the value of AI in our world of development. I'm your co-host Chris.

SPEAKER_01 0:19

And this is Brad. This episode is recorded on June 11th, 2026. Chris, Chris, Chris. The cost of AI. That's a good question. We had a lot of conversation about that today with Matt Traxinga and Tanya Breckomess.

SPEAKER_00 1:05

Hello there.

SPEAKER_03 1:06

Hello.

SPEAKER_01 1:07

Nice to see you both. Nice to see you both. We're here we are in the summer. We made it halfway through the year.

SPEAKER_04 1:17

Already.

SPEAKER_03 1:18

Yeah. Yeah. We haven't had any 100 degree days yet, so I'm happy.

SPEAKER_01 1:23

Oh. We've had 100 degree days. I don't know if it's been 100 degrees temperature-wise, the feels like um Hell's Kitchen is probably what it feels like. Uh but it gets better. Where does the time go? Like where like if you think about it, it's alright. I was thinking about that this morning because I was looking, it's it's June. So we are halfway through the year. Summer's beginning. A lot of the conferences, which is uh some of the things we're going to talk about today, are coming to an end, as they say, right? You go through that whole period of April through June with his conferences all over. But it seems to me it's going by so fast, rather quickly. And I say what are you doing with all your time?

SPEAKER_00 2:13

I don't know. I keep asking myself, like, what what am I doing? Because time is just flying by.

SPEAKER_01 2:21

It is. It goes that way every year, though.

SPEAKER_03 2:24

I mean, it gets faster than myself the same thing.

SPEAKER_01 2:27

Yeah. I wish I knew how to manage that or handle that, or is it just how everybody always feels? Or is it just the way that we work? It's a good day.

unknown 2:42

Yeah.

SPEAKER_01 2:42

But does it feel like Groundhog Day sometimes? It does.

SPEAKER_02 2:47

Oh, yeah.

SPEAKER_01 2:52

I get it.

SPEAKER_03 2:52

It's a lot even worse now as fast as things are changing. So it makes it feel even faster.

SPEAKER_01 2:59

The speed that things are changing, I swear things are changing by the minute, but the reality is it's not changing by the minute. I mean, I well, okay, I'm not going to get all philosophical on this because things do change by the minute. But with technology, it feels like things are changing by the minute. And I thought back to it and I realized there's so much coming out that it feels like it's changing by the minute because I can't keep up with the volume. So every step I take, I see something new. If you follow what I'm saying. So uh and even if you look at with the Visual Studio Code updates, GitHub Copilot updates, Claude update, like every day there's like a list of 50 things that are new. And I don't even know. I don't even know how to keep up with it, to be honest with you.

SPEAKER_02 3:47

Talk about going how fast things are going.

Keeping Up With Rapid Change

SPEAKER_02 3:50

I know recently uh a lot of us got got our uh sessions approved, and you had submitted that perhaps several months ago or something like that, or a month or two ago, you get approved. By the time that conference comes around, like do you have to like adjust so much of what you're trying to present? Like uh like do you leave last minute? Because like you said, it everything changes by the minute or even by the day in in our world, then you gotta adjust to that something new. Uh so uh you're almost like you're learning the same time as who you're trying to present to. It's a challenge.

SPEAKER_01 4:30

It is a challenge, and it's speaking of that, the conference is one of the things we wanted to talk about, uh, as I had mentioned. It is a challenge coming up with sessions so far in advance because you want to talk about things that are relevant at the time or relevant, but the relevancy of some of these topics aren't that long, if you follow what I'm saying. I mean, it is good information, it's good to pick topics, you need to have all this information, but sometimes what you want to talk about. I recently experienced that with directions, but what we had planned to talk about in November was almost obsolete or outdated by the time April came, so we had to revamp it. So, but how do you put together a session that can stand the test of that time?

SPEAKER_03 5:12

Yeah, I I I can't always do it, but I I I like to try to tie my sessions into some type of foundational principle, you know, that and then show how that's changing nowadays, you know, and it's it all still applies, but you know, I think if you kind of ground it in in that type of thing, then it's a little easier. But if you're focused purely on what's new, then yeah, you're gonna have to redo that session two weeks before.

SPEAKER_01 5:41

Yes, see the foundation that's what I say. I mean, the foundational the foundational information is good and it's it's necessary, but also the what's new at time is necessary. So you think of Business Central with the waves that are coming out, sometimes the participants at the conference, again with the timing, you have directions in April, you have here in the United States, directions in April, you have Community Summit in October, you have DynamicsCon in May. They're about the time they're about the time of when the waves are released, therefore uh I would have the expectation that some individuals want to see what's new and how to apply what's new so that they can move forward. So uh enough with enough of all that, I guess, but because we'll get into what we wanted to talk about

Conference Talks That Age Too Fast

SPEAKER_01 6:30

today. Before we do that, you mind telling us about yourself, Matt?

SPEAKER_03 6:33

Uh yeah, uh Matt Traxinger. Uh I have been doing business central development for the past two decades or so. Um for the past 15. I've been with Archer Point by Cherry Becker. Uh I'm essentially the dev lead there. I focus on dev processes and tools and education and all of that good stuff. Uh former MVP, uh, I guess MVP alum from I think 2011 to 2016. So it's been a while.

SPEAKER_01 7:00

Yeah, that's me. Excellent. Tanya, cat lady.

SPEAKER_00 7:05

Yes, I am Tanya Brickomeski, otherwise known as the BC Cat Lady. Um, my blog is the bcdevnotebook.com. And on there, my cats are featured. Um, so I like to bring my own personal brand of my cats to the content that I share with the community. I've been a part of the community for a decade now, and um just like to share kind of what I'm working through or kind of the issues that we've been dealing with, and um just love being a part of the community and being able to share my knowledge.

SPEAKER_01 7:43

Excellent, great. And we know you have that. I've experienced the cat content and the presentations too. So it's a lot of fun. Um speaking of conferences and the topics and what type of information is presented at the topics, excuse me, at these conferences you had I had come across and see that you had presented at this point um a topic on development and management. Can you talk a little bit about that?

SPEAKER_03 8:14

So, I mean, the the session we did, I I wouldn't say it was a presentation on you know development management, so to speak. It was really more of a um just an open forum for questions, you know, that people wanted to ask. I I think a lot of times at these sessions, I we all go through it, we submit our topics and we wait for months and then you know some of them get accepted and some of them get get rejected. And over the years I've noticed that the ones that don't quite fit development and don't quite fit leadership get rejected more often than than the other ones that that purely fit into those those types of tracks. And so I I've been trying to figure out where where we can make space for those types of conversations. And so we we ended up doing one at directions uh a few months ago.

SPEAKER_01 9:07

And with that, you're talking about the uh sessions that uh don't fit into management or don't fit into you know neither management nor development. What type of sessions are you referring to?

SPEAKER_03 9:24

Um so things like you know, in the leadership tracks, you tend to see they're more focused on sales or marketing or things like that. You know, they have a whole marketing pre-day at a lot of these conferences nowadays. Um with the with the developers, it's purely focused on coding or you know, heavily focused on on AI or you know, whatever the topic of the day is, but more so, you know, how do you get your teams to adopt these new tools, these new processes? How do you get them thinking in a different way? Because I most of us have been doing this for a long time, and we're we're pretty upset in our ways of of how we do things, and so I I think those are the challenges that no one really has a place to talk about, at least you know, at least when it comes to the these conference sessions.

SPEAKER_02 10:16

Yeah, that's a good point. I think there's uh less conversations and governance when you have a big team, and how does that look like from a leadership perspective? How do you get everyone to come together, learn something new as it changes? Yeah,

Making Space For Dev Leadership Talks

SPEAKER_02 10:30

so that is a good uh those are good sessions to have because you're right, Matt. There's it usually there's a clear delineation between like leadership, it's all about growing like growth and revenue and all that stuff, and then development, it's very technical, but if you look at it from a uh you know high level, you kind of need to have both. You kind of need to have that um blend of like how do you work with your team with all these changes, because whatever changes you make, it's going to affect your, I guess, bottom line uh in that sense. So I think that's uh an important session to have. You do need more of those. Um, I think we're starting to see those blend where there's more QA, more conversation-like, because if you look at the conferences that we put together or when you go um attend, um you tend to get more feedback or more interactions when it's more QA, because you're not only you're learning yourself, even though you're essentially presenting, but for me, I'm also learning to see how others are doing it. So that conversation is very, very important, less so more of a lecture, more of a story. And I think Brad and I were talking about that. Like, how do you present? I mean present as a story along with getting some feedback.

SPEAKER_03 11:46

Yeah, yeah. And I I I honestly think it's more important than ever from the dev side. Because I mean, we have like about 50 developers, we allow our customers to you know write their own code and contribute to their apps alongside of us. So, you know, how do we enforce standards when we have other developers here that don't even work for us, you know, do it doing that? So it gets complicated very quickly. So you know, I I think it's important to have those conversations, not just you know, with partners, but you know, also with users at the same time. How do we all work together and you know actually become a team together?

SPEAKER_01 12:24

Yes, it is important. I think it's uh I've been seeing more and more on change management over the past couple years, and that type of topic I would put that topic I would put into the change management role as you're talking about, Chris, but even uh Matt, as you're talking about with the standards of the implementation.

SPEAKER_00 12:42

I think it's also like when you you know, if you're presenting a technical topic at a conference, like it's very grounded in that content and it's like definitive, like you can say this is what you have to do, as opposed to these types of presentations. It's more of your personal interpretation of it, like how are we doing it? And there's that vulnerability side of it where you're sharing the struggles that your company is going through, or the the things that you know we don't have this figured out. Like, I want to hear from Matt, like how are you guys implementing this, or how are you handling these tough conversations with your customers? Because, you know, like we don't necessarily know how to handle some of those situations either.

SPEAKER_01 13:35

And I think even more so with the rate of change, I uh it it becomes more important. And as Chris started speaking about, I I think there's two parts. One, I think sometimes the conference may dictate the content, because some con some conferences, I don't know if it's on purpose, so design just could be my perception, uh maybe geared towards towards a certain audience. I think the conferences need to have that because there are so many conferences, because here in the United States, as I had mentioned, uh the ones, the popular ones that uh everyone's familiar with with directions is for partners, dynamics, uh DynamicsCon in May is for primarily focused on um users of Business Central, and then Community Summit, well users of Microsoft Suite, I should say, right? Because it's more than just Business Central, it's just our my universe is business central, so I focus on that. And then in October, you have primarily a user-based conference for Community Summit, and then you have all the meetups and everything in between. Is there a shift of the type of topics that you would have for the partner type conference? And then, like I said, said you can go into Europe, they have all the the same thing directions, BC Tech Days, NIA, they have all of those uh uh as

Partner Vs User Conferences Content

SPEAKER_01 14:50

well. Does the should there be different content for different types of conferences? Because then there's the challenge of how does one choose which to attend if uh there isn't some I don't want to say consistency, but some sort of general theme for the conference.

SPEAKER_03 15:10

Yeah, and it's it's been interesting because I I I guess I'm on I'm on the programming committee for a community summit this year, so I've been you know approving the dev sessions, and my my co-track leader she put it uh you know, the development track to me in a way that, you know, I don't know if I agree with her or not, but end user developers for Business Central are usually accidental developers. And I and I don't know if sessions should all be geared towards that. I I I versus a part of the conference where they're gonna be deep and technical, because that's all you know those types of developers do that work for partner resources and user developer who might be developing for VC might only be a small part of their job. So I I don't know what the balance is in providing something for everybody or whether the conferences should really become more focused.

SPEAKER_01 16:10

Yeah.

SPEAKER_02 16:10

Well, I do yeah, that's uh that's that's tough.

SPEAKER_01 16:13

I do agree with you, by the way. I think is what I heard properly. I do believe some organizations that are what we would people would consider as like users of the application, some of their development team is on par, if not better, than some of the developers that may work at partners. I don't think I don't think the status matters because I know people who've worked at partners that move to end users, and I uh that was we call end users, and I know people from end users that move to partners. And as I had mentioned, um I I think I look at it as everybody's just working with the product and it's it's your job to do business-central development. Now, do you have people as you had mentioned at some end users that have multiple roles in development may not be their full-time focus and they do smaller tasks and then work with the partner or somebody else to do the more complicated? Sure. But you have the same thing, I think, on the partner side, because you have developers that may uh do certain types of tasks and move to the other ones as well. So it's yes, I don't think where it is matters. I think um what you get with being in the partner sometimes is just exposure uh to different things that work and don't work, but the skill level um is uh in my opinion is no different uh today.

SPEAKER_02 17:21

That is that is very tough, I think. Um Matt, you had mentioned you're part of the uh programming committee. I am as well, and it was pretty tough looking at some of the sessions because you'll see that it's a blend of I mean, I I think I was on uh I'm an uh uh administration, but administration and business central can be very broad uh nowadays because uh maybe so much less because there's some tools you can use now to manage business central. And so you know, you could get down to a telemetry component as administration, which is now blended to kind of development. I mean, how do you read it from a business perspective? So you see that it's kind of like it's so broad, so really, really difficult. Uh maybe perhaps in the future uh the I mean the community could change and maybe not categorize it in that way, categorize it from a business perspective rather than a specific technical track, uh, which is majority of those are technical track. Um maybe you get what one leadership and one marketing, uh, which is uh you know entirely separate as well. But yeah, I mean we maybe in the future conferences, maybe uh perhaps they can uh change a little bit of that so you do get a blend of you know what does it mean for your business from a client perspective, and then uh and then you got your leadership track, you know, things like that. So I don't know I I what the answer would be. Uh but I think there should be a little bit of shift or change uh with the times because now AI blends that. Like being administrative, do you really need to administer? Could use could I just pop in an agent to do that and connect to Business Central and then administer it for me? Uh maybe. Right.

SPEAKER_03 19:20

I think all of the conferences on the whole are gonna have to decide whether they want to change in the coming years because if I can get that content from AI or from Microsoft Learn or a blog or whatever, then it's extremely hard for me to justify spending that that money and that opportunity cost to send somebody there.

SPEAKER_01 19:44

And that is a whole candle worms. I've asked that question in previous episodes. I asked I've asked that everywhere, and that's what I said. I realized uh after one of the con I don't know, I think it was maybe you have to after directions. I I think I I came to the realization there as well that the conferences need to be less what I call academic and more functional, because with what I think your topic is more how do I you know the the new features are great. This ability to do this is great. You can pick whatever it is, application-wise, technical, anything. How do I apply it? How do I apply this? How like as Tanya used to mention, what's your story for using it? What experience do you have? Things that I could look at, right? How do I, you know, the sales order agent is great. I love seeing how it's working. Can you show me how to set it up? Can you talk to me when I would use it, when I wouldn't use it? See, those types of things I think would be more driven towards the space of a conference attendee because I can look online and get all the content anyway. It is, it's a challenge. Like you said, it's a challenge to determine what to say. So you had that session. What was some of the feedback that you had in the group? I'm curious now to hear if it was an open forum discussion, what the discussions were

AI Adoption Strategy Inside Teams

SPEAKER_01 21:00

like.

SPEAKER_03 21:01

Chana, you were there with me.

SPEAKER_01 21:02

How do you think it went?

SPEAKER_00 21:04

Um, obviously it it shifted towards AI um because that's the the hot topic right now. Um so we were never guessed right. Um so I think we were intending to talk more about, you know, like how do you manage your team or how do you manage some of these like edge topics? But obviously it shifted towards AI. Like what does AI adoption strategy look like? How do you get people on board? How do you, you know, are you trying to take your entire team of developers and push them all at the same rate? Or do you maybe focus on, you know, your people who are really excited about AI and they're gonna kind of help drive the adoption. They're gonna be the ones looking into, you know, building out custom agents or orchestration. Um, so I think it it was not what we expected, but at the same time, in the back of our minds, I think everyone.

SPEAKER_02 22:00

on the panel said like okay yep we could see that ai was going to become the topic of conversation yeah it's less it's less of the technical that's a technical I know like in the past where Tanya you'd probably uh your sessions would talk about like this is how I would develop and this is you know this is the the functions I would use there's I I feel like there's less of that it's more like well AI can just handle that but this is how I use this is what I use to go ahead and accomplish accomplish what I was trying to accomplish but you're right the conversation has changed in in some of these sessions.

SPEAKER_00 22:39

Yeah and even you know going back to Brad's last comment about what like the the content at conferences um I was talking with Microsoft and they said you know like they put together all the launch event videos so when they get to the conferences they're kind of expecting maybe you've already watched those so you can have those deeper conversations and ask those questions. So I think it goes back to there's so much content. No one has a strategy for consuming all of the content online and coming up with those ideas. So I think a lot of people go to conferences to try to figure out what's new. What do I need to learn more about how do I implement these strategies? And so I think we could maybe shift some of the conversations then to okay how do I actually use all of this in a a real world context or ask have more of those open form dialogues. I really feel like those are the sessions that people like gravitate towards the the more of the panel sessions where you hear more perspectives in a short amount of time I tend to agree there are a few comments the from a couple people who said it was the best session they attended something that that made me happy.

SPEAKER_01 23:57

Yes that's a good thing to hear yes I think those sessions are are good. I like the the oftentimes the quick tips ones are usually standing room only as well but I think it it serves a couple purposes one because everybody who's going will have a question that they would like to have answered. And how you get that question answered can be many ways. Either you go to sessions and ho and uh with the hopes that whomever's presenting is going to answer that question through the content that they have. You may be able to talk to other individuals through networking at the conference or being able to ask the questions in the sessions. And I've agree with you I find that a lot of times those sessions where you have that audience participation and interaction in the speaking people tend to favor those a bit more because I think they all feel like they got something out of it that was more personal versus just watching somebody lecture for an hour. Again it's everybody's different everybody has different things so it's curious. But AI took over it's it's amazing how it takes over and the challenge with that it's even if you talk about how do you manage a team I think the larger team in my opinion is one of your opinions is the larger team difficult to manage within the world of AI or is it easier?

SPEAKER_03 25:19

I think it's harder to be honest. It's because I I'm a person who likes consistency and for everybody to do things the exact same way you know every single time and getting 50 developers and 50 internal developers and 50 client developers all working together to do things the same way every time is awfully hard I think that is difficult and a challenge without AI.

SPEAKER_01 25:52

Yeah exactly you know it I think that is significantly easier when you have a very small team when you can I I don't want to say I I'm not a big fan of micromanagement at all, but at least observe the majority of what is going on you know I I think you have less friction too in the conversations because I I I know it's a challenge in my opinion with AI again and I'll I'll stick with the within our space with the artificial intelligence for development documentation project management type tasks. But the stuff the features and functions that are added are added so rapidly it is almost challenging with a larger team to disseminate that information because by the time you have something in place I feel like something new is already here. And do you still continue down the road of what you had put in place or do you always try to incorporate some of the newer and I quote more efficient features and functions that they added if you're working on an older code base do you keep it on the old standards so to speak or do you change it when new things come in and have it be more complicated and inconsistent and I I don't know that there's a good answer to that. Yeah. Tanya what about you on that? How do you feel with uh the adoption for AI within an organization?

SPEAKER_00 27:23

I think and the dissemination of standards it's definitely a challenge. Um you know we're trying to provide like a framework I guess for how do you you know work with AI how you know because everybody's at these different points in their AI journey and I think that's one of the hardest things is it you know that um adoption of or like that understanding of oh I need to be using AI and there's still people out there who haven't quite like caught on to that fact yet that oh I need to be you know building my own custom agents. I need to be figuring out how can I do what I'm currently doing with AI. And so when you have you know 30 or 50 people that you're trying to all get on the on the same track, it's very hard to get them all adopted like at the same time. And so one of the things that like I picked up at directions is identifying more of our like as power users or people who are really interested in AI adoption and kind of letting them go and figure out things and then share it with the group. And so we're doing a lot more of that where okay I I built this custom agent it does X, Y, and Z. Oh, I need to instead of having a massive code unit with all of our event subscribers that's consuming now all of my contacts window if I had a smaller code unit, you know, I'm using less tokens. So especially as GitHub has changed to charging for like the credits and the whole building structure like it's another shift of how are we using AI? How do you use it effectively? And so I think in the past it's been working on you know working one way and you kind of work that way for a couple of years you kind of get into that rhythm and we're constantly shifting right now and there's no like standard you don't feel like you're really grounded in your standards because your standards are always kind of changing as AI is evolving at the same time.

SPEAKER_02 29:40

I do have a quick comment about uh um I I think you have a that's certainly a challenge Tanya that you have to deal with and in a the adoption of AI. I do believe that you know in the past we've talked about slowing down a little bit you know how do you get someone to um sort of understand the the the value of AI. A lot of times we tend to chase you know how can I apply this AI tool into a new project you know perhaps something that you could give value back to your clients or your your customers of looking back at maybe a a PTE that you built for them last year and say hey let's let's try to improve this now the benefit of that is that you know how that was built and you know how you wrote it perhaps at that time that bringing AI into that will give you uh maybe some improvements and efficiencies that could help your your client uh in the long term so I think to me that's a good way to um have an adoption of how you would use AI into an existing one that you know how you built it versus trying to build something and then applying AI on top of that. But again that's a suggestion that I would have because you know um on my little side projects uh you know not necessarily related to uh business central I went back to some of the things I've worked on in the past and say okay I wonder what AI would do to this thing that I know how it functions and see if it'll find improvements. And that allowed me to learn and quickly adopt of

Standards, Tokens, And Team Consistency

SPEAKER_02 31:20

how AI could work for me and perhaps something we could do in the business central world because it changes all the time. And uh but again it goes back to like is are your clients willing to pay for that and are you willing to put effort and or time aside to do that because you're chasing for that new revenue right so that's uh it's a weird balance let's lots of balancing to do one of the I just want to go back to 1980 and 1990 things were much easier.

SPEAKER_00 31:49

One of the things that we've been looking at is not necessarily changing like our day-to-day processes like the how do you write code I mean obviously we're using autocomplete or we're leveraging agents um but we've been focusing a lot more on all of those like edge cases that we just haven't had the time or like the money to do that. So I think Matt, you we've talked about test driven development and where does that come into play. And now you know if you can just tell AI to start writing the test code units for you and all those test procedures and you know you can kind of bake some of that into your time. You know we've been focusing a lot more on performance and using like the business central performance toolkit. So having AI help write some of that code so we can do more of that performance testing, make sure that upgrades are more stable um we're focusing more on these topics that we must like in the past we haven't really had time for or we haven't focused that energy on before.

SPEAKER_03 32:58

Yeah I I I I do think it's important that we we don't try to do things faster but we try to do more in the same amount of time.

SPEAKER_01 33:08

Absolutely otherwise we're gonna be in trouble from a from a billing standpoint as partners um you know but I like that see that is a good way to put it it's not you're trying to do things faster you're trying to do more in the same amount of time yeah we're we're trying to deliver the absolute best possible system we can cost significantly more than we could before in the Tanya's point as you had talked about again you maybe I don't want to get down to the whole testing I'm I'm a I will say I'm a big fan of tests because tests have saved me personally. If there weren't some tests written uh we would have had some issues with some um extensions that were created but this is giving you the opportunity to do more within that time not to say that you didn't deliver quality code before it's just makes it easier to write quality code in a sense and and quality is you know is a is it's uh one of those words define quality right it's uh is it good code is it well written code you know is it performant code is it code that does what it's supposed to do right so how do you measure quality but I like the approach of saying you do more so now you do more testing you do more documentation. I mean that's some of the stuff that I've seen you know we're talking about coding a lot is uh the documentation that you can generate with AI. You may have to adjust some of it which is fine but I'd rather have it generate documentation for the the extensions or documentation for the procedures than have to make some adjustments then have to write it all from my mind or you know from scratch as people say but as you're saying now you can provide all of those as part of the deliverable within the same amount of time that it would take to develop it. I'm not saying it's a one for one but like I like that approach of looking at it that way. It's not getting it done faster it's providing more value in a shorter period of time.

SPEAKER_02 35:05

Whereas it would have taken it made me chuckle we said more documentation because I don't think they ever existed most of the time. So uh you add more yeah you're right you add more value.

SPEAKER_01 35:18

There's nothing wrong with the number field specifies the number value or you know specifies the number right that come on that's documentation.

SPEAKER_02 35:28

We've all had that documentation since the beginning of time yeah so uh so I'm curious for the group for the three of you are you know focusing on development do you when you're using AI is it primarily uh I don't know if you have percentage like is 80% of your usage of AI uh more on the development coding side or do you or do you um or do you also do documentations and test scripts and things like that that are uh um are just not just development side of things because you know I've spoken to somebody and they said you know I only use AI for development and that's all they use it for and they don't really do more more beyond than that uh they would say like 99% of the time that's what they use it for.

SPEAKER_03 36:12

Do you use AI copilot Claude whatever uh whatever particular model you use uh more than just development I know you guys brought up documentation and creating test scripts what else do you guys use it for from from your side of things I I use it for everything if if anything I use it for development the least I've I you know urgently I guess I find if you don't and it's not I mean it's it's this way even without AI if you don't spend the time doing a good design you're gonna get bad code as an end result. So I I find I spend even more time using it to do the design of the solution and then you know the dev is 10 or 20% of the total time spent but yeah that that's me but yeah I use it for tests and documentation and design and the the whole thing. Nice.

SPEAKER_00 37:04

Yeah how about you're kind of in the same boat where we're spending a lot more time with AI doing the planning stage so figuring out a solid design and how are we going to implement it um reviewing other solutions that are already in place and how will this interact with all of those making sure you know we're hitting all of the edge cases. So I think that's one thing where we've really leveraged AI is now we have access to all of the designs. So it's not just you know the person who worked on this one specific task you now kind of have that knowledge because AI can look at all of the different designs at the same time and say hey like this is happening over here do we need to worry about this how is this going to be impacting it? So on especially like our larger projects we've found that it's been super beneficial to kind of catch more of those edge cases. And then you know whether it's writing like test scripts for users or that they then feed into like the test scripts within like our development as well um being able to definitively like define what those test scenarios are. We've been focusing on that as well because that kind of sets the framework for how is this going to work and then the solution in the middle doesn't necessarily matter as much is it you know AL development is it something in the power platform could be something totally different. Like I don't I don't care what the solution is I care what the problem is up front in that like we have this end result.

SPEAKER_03 38:46

The middle like you could throw whatever you want in there if you really wanted yeah and I I find myself like I'll I'll do the initial design with Claude and then I'll let chat GPT be my code reviewer or my my design reviewer rather and then I'll let Gemini design review that that second one and then you get those kind of layers of of quality built in until they kind of converge on a solution and then go from there.

SPEAKER_01 39:13

So you use a multi-model multi platform harness approach for your designing and it sounds like I I can't make myself trust any one of them they all sound believable. They do with everything that I I get out um they could make they could make a catastrophe sound like uh a pleasant experience uh with the way some of stuff comes out so it's so I'm hearing or what I heard is that you are spending a lot of time with AI doing the background planning preparation not

Using AI For Design, Tests, Docs

SPEAKER_01 39:50

so much for the coding and maybe that's where you have someone do traditional development so they could develop it according to the plan. Do you have to use the AI tools to do the development as well?

SPEAKER_03 40:02

Yeah boundary has to feed those designs into copilot or cloud code and let it build it and it it gets it 80% of the way there and I'll fix the fields that it didn't you know know were were incorrectly named you know it didn't know the posting setup was P U R C H dot instead of and it's spelled out purchase you know things like that. But for the most part it it gets it pretty correct and yeah.

SPEAKER_01 40:30

How times have changed how times you know changed.

SPEAKER_02 40:36

From a business perspective and leadership perspective Matt you had you know kind of alluded to this in the beginning is how you work with the team and the technology and how do you adopt Tony you had mentioned how you adopt you had also mentioned tokenization right like how do you manage tokenization uh with using AI you know how do you uh incorporate that in the back end of things of like hey it may cost you so many tokens uh just to accomplish something that's something that has never been uh sort of a cost uh you know in the past you kind of just uh it's more the time spent of building something now time spent building something of course you bring more value but then you also have now this cost yeah that's another one right like those things are just gonna I'm thinking of the conversation some conversations I'm going to have is you uh to follow what Chris is saying it was a question that I was thinking of it that you're talking about this you both do a lot of estimation for projects and development and tasks and working with the team how do you estimate it now that is something we are still figuring out we're still estimating the same way I think we are estimating the same way that we have been in the past because you don't necessarily know who's going to be working on the task.

SPEAKER_01 42:10

But if you have someone who can do it in five minutes there could be a cost to them doing it in those five minutes as well.

SPEAKER_03 42:18

Yeah the cost may quickly out outrun what you're billing for those five minutes.

SPEAKER_01 42:24

Yeah that's that's what I was getting at now and it it's more visible with the GitHub copilot pricing change that it went from you know you had the monthly pay plan and the monthly plan you know gave you a certain amount of of use and then now you have the actual credit use. So how do you factor that into the process? It's just a question I'm throwing out there. And I'm asking the question because I think there's a perception and a reality of some things and some people and some things that you see where people say oh I have all this cool fancy process I can do all this stuff what are they actually doing with it and then how are you getting what are you returning on the value that you're producing and how are you getting that out of there because I looked at some usage of uh credit usage. I always have to use the new term now because in GitHub Copilot it's credits. That's AI credits, right? It's not tokens. So I look at how many credits we use on something that I did and then I'm saying well how can we even do the math to see what it will cost to do whereas before I could say it would take me an hour to do it. You know you can sort of equate what the cost would be.

SPEAKER_02 43:43

Yeah that's an it's a that's an interesting balance right because you know as uh Tony I I think you had mentioned this the the copilot is the the the value of the tokens or credits is certainly uh I guess less because it's more expensive. Uh and and and what does that look like when you are now building solution? I mean, yeah, you could somehow figure that out in the back end and still present like the total cost. But as a business, you know, especially if you are a small partner out there, that's something you have to consider. Um, and it's something you have to consider when you're having maybe perhaps you're working with a contractor as well. I mean, that's a cost to them. And, you know, do you provide that tool uh for your contractor? Because you want them to have good quality, right? So you you you want them to give you good quality results. You know, do you provide those tools? So as a business, that's something you have to consider. Do you just use Copilot or do you bring Claude or Gemini or whatever model uh at their disposal? Uh again, that's go goes back, Matt, you're saying you know, the the the leadership component and in the business owner uh component. So it is an interesting balance because I think from my perspective, that will become perhaps an issue down the road. Right now we're not feeling that, but um, I have a feeling that that becomes part of a business challenge of how to manage credits of the tools.

SPEAKER_01 45:18

We have a lot of challenges from this.

SPEAKER_03 45:20

Yeah, I mean, so so far for us the cost hasn't been problematic. Like we have very few people who are exceeding their monthly allotment of tokens or credits. I see these stories of people who accidentally spent whatever, five thousand dollars in a day or well, whatever that was. Oh, you can do almost five million dollars in a week or something. And I I don't understand how they did that.

SPEAKER_02 45:47

Well, yeah, I I think I think uh some announcement the other day, uh Brad and I were talking about this. They, you know, Claude released Fable and then Mythos. It's like, you know, I think someone had tested that and like, okay, it's gonna cost you this much dollars. It's like, ooh, that's scary. That's scary because if that's available to a team member of uh a team member and then they go ahead and apply it. Um I'm sure you have spending limits.

SPEAKER_01 46:15

Well, yeah you have to binge it with spending limits. And it goes, I think, with just education to just how to utilize it and manage the context. But I think it's going to come from when to utilize that. But it does come with a cost, and it's not just for partners. I mean, you have it's any business, any organization, anybody using it, it does have a cost with it. It doesn't make it free, it doesn't make something cheaper, it doesn't make something you know necessarily I mean, I think you can do it f some things faster, but I like Matt's approach of you doing more in the same amount of time. So you you're getting something you you're delivering more in the same amount of time as you would before, uh, which adds value to it. So I I think it's a challenge for everybody, and I like the candidness. I think everyone's trying to figure it out because I think it's moving every day. Um and I don't know where it all plays out. Like I said, I almost want to go back to 1980, 1990 when when actually 1990, 2000s when we had one update every couple years, and that's all we had to worry about. And uh we would just do our development and our implementation.

SPEAKER_00 47:22

I think I think um it's learning how to use like AI responsibly and understanding like your credit usage. And so, you know, like we might have a more experienced developer who they can write a majority of the code and they need to then use maybe

Estimating Work When Credits Cost Money

SPEAKER_00 47:40

a more expensive model to get the last mile done of the development, you know, like this really complex piece that, hey, I want to use Claude to help figure this out. Whereas a junior developer, they don't necessarily know all of our best standards, or maybe they they don't know, you know, the best way to approach things. So they might get stuck in the churn cycle and so they might burn tokens faster. So we're trying to figure out, you know, how do you use tokens effectively? Um, and especially in a larger organization. So we have developers working not just in business central, but other products as well. So um, how do we manage costs? And so how do you say, like, this person needs more credits because they're working on like this big project, or they they know what they're doing, they're they're using these credits effectively, and you know, we can get something done in a couple hours that used to take us an entire week of development. Um, so I think there's like all these different conversations of, you know, maybe it's sometimes you need things faster, or maybe you need to get like the baseline done, and then it's just doing tweaks. Um, maybe it's we've been using it for code reviews as well. Um, so we wrote like a code review agent, which is great, but then do you need to use that on every single commit? Um, so you know, when do you bring AI into the conversation? So you're not just burning through all your credits in the first week. That just made me chuckle a little bit.

SPEAKER_02 49:18

It's like it's like having uh living in a neighborhood, you run out of sugar. I'm gonna knock on Tanya's door. Can I have a little bit more sugar? Can I have sugar? Because then I ran out. Can I borrow yours?

SPEAKER_00 49:32

Yeah, yeah, exactly. And so um I think especially we we use GitHub. So um, you know, how do you manage the credits across a team? And so right now there's kind of like this learning period, this grace period where we're trying to figure out usage and how are people using it, how do we use it effectively? So we've now have like all these um like best practices, you know, like if you're you you know building something out, like can you use autocomplete to help you fill out all your tool tips? Do you really need an agent to write those? Like, or could you sit and hit tab for two minutes to fill everything out? Like, what's the trade-off?

SPEAKER_01 50:13

Yeah. Yeah. Interesting balance. It's this whole thing has me thinking a lot about it. It's it's uh it's it's it's new for everybody. And um it's a challenge that we have, and it dominates every conversation now. I I really would like to go one day. I have to figure out maybe I'll just go out somewhere by myself in into the woods or something and bring nothing with me. And I could go a full day without hearing anything about AI.

SPEAKER_02 50:43

Um But do you do you try to balance all the different models though? Like, for example, like you know, uh so I've been uh I was working on um on Claude, like if you use Haiku over the other models like Opus and you know Sonnet uh for different specific tasks. Uh do you guys also manage those as well? Like GPT 5.5 versus other versions on specific uh tasks just to save a little bit of token or credit?

SPEAKER_00 51:13

Yeah, yeah. We've been um trying to use like codecs or things like that, like the cheaper models for um, you know, especially like documentation or code reviews or things like that. And um, you know, while we we've say like, oh, this one works a lot better, like is there is it that much better that you need to pay for the the more credit usage? Um, you know, could you use the the cheaper models and still get you know 80% of the way there, and then you as the developer just have to go in and fill in the missing pieces.

SPEAKER_03 51:54

I'm not doing experiments for which one is better in certain use cases, and you know, I I stick to the cheaper you know options. Um, the the 1xs are the 0.5x token usages.

SPEAKER_01 52:09

Um how do you define better, by the way? Yeah, exactly. I mean that's hard, yeah.

SPEAKER_03 52:14

I I I think the one the simple ones are doing a fine job, personally. And you know, I think more power to the people who have time to experiment and benchmark and decide those things. But that that's not me.

SPEAKER_01 52:28

You know, I yeah, I'm with you. I uh I and I do experiment to see what I get in some, but I've I've toned back on that now uh significantly because I I don't know if I'm in the capacity to be able to say this one's better than that, to be honest with you. And and I'm talking across all domains because I use it for everything, not just development. But I I I want to finish up with one question, and it goes with what you were saying with the experimentation of the models. AI and these large language models have been around for a long time, and from what I understand, they're trained on domain data, right? There's a lot of data that they're trained on. They're consuming data, whether it be public domain data, um, there could be some private data or whatever. We're now a few years in and you're seeing some of these newer models coming out, and you almost wonder what's the difference between these models? Is the difference with the models the training data? Is the difference how it's processing? And then I hear reports again, it's it's all what you hear. Like what's the truth? I I don't even know anymore sometimes. I have to I don't even know how to validate the truth sometimes, uh, with all of the information that gets presented. Are we going is are we at the point now where these models are learning on data that was created by the previous AI models? Meaning a lot of people are generating so much content online, you know, some people call it AI slop, some people call it whatever, right? It's a lot of AI generated data. Are we at the point where we ever get to the point where the AI is learning on data that was created by AI from the data that it knew before? And is that impacting the performance of those models?

When AI Trains On AI Content

SPEAKER_03 54:13

I think it's interesting that it's still getting better anyway.

SPEAKER_01 54:17

Yeah. I start to think about that. Because it's you know, if it's learning from the written data and it's writing the data, here we are years later, a couple years later, three years later. How many I don't even know how many years it is that the talkbook goes by so fast. Is it three, four years? How much of that data that it's trained on was created by it, and with a lot of this misinformation out there and a lot of this hype. Oh, this is the best thing ever. Oh, this model's the best for this. How does it discern between fact and fiction as well?

SPEAKER_02 54:51

I mean, that's I think Matt's right. Like I think we've been we've been in that space for so long, even before AI came around. You know, when when you get a scientific journal gets released where you have proven evidence, someone takes that and write it from their perspective, and then someone looks at their editorial and write it from their perspective, and so it becomes a um you be you know, you create a bunch of content that has can references each other, and you don't know, hey, there is actually a scientific journal that was written by Harvard, which you know that the original source of all of that. Big difference. That's no different now.

SPEAKER_01 55:26

There's a big difference. I can write 40,000 of those articles today before that is true. So the the it's it's it just sped up the volume at which things are produced now. I I'll tell you, it's even I even talked with someone yesterday. I don't I used to love to blog and used to do all this stuff, and I'm almost at the point, I don't even find excitement in it anymore because there's just so much stuff out there.

SPEAKER_03 55:50

Like it's I I I will say I I think there's an interesting parallel to before AI. You know, we all learned by looking at Microsoft's code in-based objects to some extent. Correct. And most of us would probably say that that's not the greatest code that they've ever seen. If we're being honest, you know, no disrespect to Microsoft or whoever it was they bought the product from and their coding abilities. But I think we would all say that. And yeah, we still produce code, we would say it's better, probably better than what we've seen in the base objects of the past. And it's good, I think it's just a kind of a different flavor of that to an extent on a much bigger scale, of course. But I'm not sure it's that different.

SPEAKER_01 56:44

No, it's good. It's a great perspective. It's a great perspective way to look at it. Uh with that, I appreciate you both taking the time to speak with this afternoon. Um, I I enjoyed the conversation. My mind is uh every time I have these conversations, my mind wanders into

Where To Connect And Closing

SPEAKER_01 57:00

uh some things to talk about, some things to do afterwards. Uh if anyone would like to reach out and contact you to learn more about what you do, talk to you a little bit more about AI, talk to you about um you know sessions or any other uh things, business central, what's the best way to get in contact? Jeez, tongue-tied. Contact with you, Matt.

SPEAKER_03 57:17

Yeah, you can always uh get to me on LinkedIn or uh I guess reach out via email, mtraxinger at archerpoint.com.

SPEAKER_00 57:25

Yeah, I would say the best way would be LinkedIn, just connect on there and continue the conversations. I think that's where this conversation even came from was LinkedIn. We were all commenting on each other's posts. So uh I think that's where everybody's kind of collaborating or connecting right now.

SPEAKER_01 57:46

No, it is. It's a great avenue for that. Well, again, thank you both for your time. I look forward to seeing you both soon. I'm sure I'll see you. I think I'll see you both definitely uh, if not before, in October at Community Summit. So I look forward to talking with you both then. Uh thank you again. Speak with you soon. Chao ciao. Thank you so much.

SPEAKER_00 58:01

Thank you.

SPEAKER_01 58:02

Take care ofe. Thank you, Chris, for your time for another episode of In the Dynamics Corner Chair. And thank you to our guests for participating.

SPEAKER_02 58:10

Thank you, Brad, for your time. It is a wonderful episode of Dynamics Corner Chair. I would also like to thank our guests for joining us. Thank you for all of our listeners tuning in as well. You can find Brad at developerlife.com. That is D V L P R L I F E dot com. And you can interact with them via Twitter, D V L P R L I F E. You can also find me at mattalino.io, m-a-t a l i no.io, and my Twitter handle is mattalino16. And see you can see those links down below in the show notes. Again, thank you everyone. Thank you, and take care.