Episode 509: AI Won't Wait: Setting Up Your Team for Agentic Development in Business Central

In this episode of Dynamics Corner, Kris and Brad sit down with Milan Milincevic to explore how he's transforming his team of five developers into a fully agentic development operation using Claude Code. Milan shares the practical steps behind building standardized agent configurations and rules — inspired in part by the BC Code Intelligence MCP — so that every team member can produce consistent, production-ready output. The conversation leads into real-world results, including a striking example in which a change order originally estimated at 24 hours was completed in just over 1 hour using agentic development. They also tackle the growing questions in the industry: how should partners rethink billing when AI makes developers dramatically more productive? What happens to junior developers who haven't built deep functional knowledge yet? And what's the next frontier — can AI handle Word layouts and reports? Whether you've already started experimenting with agentic coding or you're still on the sidelines, this episode is packed with practical insights and honest conversation about where BC development is headed. 

Transcript

Cold Open And Guest Setup

SPEAKER_00 0:01

Welcome back everyone to another episode of Dynamics Corner. Who is Claude? Is that your buddy? I'm your co-host, Chris.

SPEAKER_02 0:11

And this is Brad. This episode was recorded on February 19th, 2026. Chris, Chris, Chris. Who is Claude? I actually have a prani. And he calls me King Pran. With us today, we had the opportunity to speak with another great member of the Business Central community. And we had the opportunity to talk about what else? AI development and AI in business. With us today, we had the opportunity to speak with Milan. Good morning. How are you doing?

SPEAKER_01 0:55

Good afternoon, guys. I'm good. I'm just running from another meeting to join this podcast with you guys. Oh, that's good. How are you guys? Excellent. Excellent.

SPEAKER_02 1:05

Been looking forward to speaking with you. Uh hope you didn't have to run too far and too fast.

SPEAKER_01 1:12

No, it was not. It was not too too far, just from a different room to this one. So it's okay. I just had enough time to prepare a bit of water for me, and that's that's Yeah, that's good.

SPEAKER_02 1:23

Well, I have plenty over here. If you need some, I'll send it to you.

SPEAKER_01 1:27

Yeah.

SPEAKER_02 1:27

That would be that would be No, it's uh no, thank you for taking the time to speak with us. But look forward to speaking with you. And one thing I was thinking about this morning prior to speaking with you, it's really crazy how we are living now in 2026 compared to where we were in 2025. Because I realize today it's like you think of development and coding, how much do you actually code today? And what I found myself doing this morning was uh I spent so much time prompting. I was opened up Visual Studio Code because I wanted to type my type my prompt before I put it in to the chat uh uh session for the prompting. And I was having within Visual Studio Code, you know how when you're typing, Copilot will uh uh assist writing with you? It was helping me write my prompt. So now I'm actually having copilot help me help me write my prompts.

SPEAKER_00 2:24

So you're not developing, and then you're not prompting.

SPEAKER_02 2:28

Well, that's what I'm saying. It's like we're changing from uh something to help create the prompt. I used it to create uh agent files and street, you know, whatever you want to call them, instruction files, but now to create the prompt, it's kind of crazy.

SPEAKER_01 2:42

It it is crazy. Um yeah, I mean uh to be fair, I didn't write uh a lot of quote myself this year at all manually, so it's uh it's it's it's crazy. Crazy times where in what we live now. Um yeah, you know what what I say, I'm thinking I I'm reading here the small team of developers, it is five five of us. And um I wanted that all of us just do fully uh jointing development, not uh not manual one. So now we are working a lot of with establishing all of the cloud code um rules and instructions and ideas so that he can it it can take it to to cre to create a good output and uh it's amazing, you know. Uh the more we are doing it, it's more amazing.

Milan’s Background And Community Work

SPEAKER_02 3:45

So it's uh it's so it's so difficult to talk about, and I tell people I don't even talk with people about this anymore because I think everyone will think I'm crazy at at the stuff that's happening. But I I I'm interested in hearing more about your AI development workflow with your team and setting it up to incorporate agentic coding into your workflow and what you're doing. But before you do that, can you tell us a little bit about yourself?

SPEAKER_01 4:12

Yeah, of course. Sorry, we we jump we jumped uh immediately to the topic. Um once again, guys, first before uh I introduce myself a bit, uh thanks for inviting me. I'm really excited to talk with you today. Uh my name is Milan Milentrevich, and uh originally I come from Croatia, but last two years I I live in Norway where I'm uh leading a team of developers for BC. Uh I also try to be very active in the community, and so I was speaking at different sessions, uh creating some small open source projects. Uh previously I was writing some blog posts. To be fair, in a in uh late time I I stopped. Uh I'm becoming more and more lazy, but uh hopefully I will pick it again. Now you you don't need to write anymore, you can uh AI to do that for you as well. You just need to give some ideas.

SPEAKER_02 5:08

Yeah, so there's a lot of content out there. A lot of content.

SPEAKER_01 5:12

Yeah, a lot of content, yeah. And that's the thing. Even before AI, for me, what was um challenge was to find a right topic because the easiest thing was just to use something from learn and then just to example and write for that was not adding any value for me. So I I always try to find either some edge cases or to find something more interesting. So and now it it's too much, yeah. As you said, it's content. So I'm struggling to see whether there is actually some value in providing even more of that.

SPEAKER_02 5:46

That is not easy to provide. No, it is, and and not uh we'll we'll get into the topic, but I think you're right. I mean it's I I I think there's two approaches. One, as you said, there's a lot of content, and you try to find the education with the more challenging cases, but it's it's who sees the content and where they are in their journey. And it's it's I understand the feeling. It's I wouldn't call it lazy, it's it's um it's sort of hesitant because you want to add value and you don't want to spend the time to write something if there's already a lot of information out there. And then coming up with the edge cases is a little bit challenging as well, too, because it's sometimes it's time consuming to put that information together. So it's uh I I know exactly where you uh uh sit with that, and then add to the craziness of having to keep up with all of this technology, and it's uh it's a struggle.

SPEAKER_01 6:38

Um I mean it definitely is, but uh it it is fun, it's fun to live in these in these times, and uh I'm just so eager to see what will come next because uh it's so hard to predict.

SPEAKER_02 6:55

Oh, it's it's changes by the day. You you wake up one day there's open claw, next day I saw last night there's hermit claw, and there's like nano claw, and you you see all these different things that are being created. It's it's uh uh it's a challenge. But I think we'll level out in a sense at one point. Um and I don't mean level out like technology will stop, but I think sort of the mode that people will be going into may sort of slow down uh and and build upon the framework versus everybody sort of creating separate different frameworks. So you have a team of five that you're working with development on there, and you said you're working on the the age act agenc development with them now. Uh how did you how do you even uh organize a group of people for that, right? I'm sure there's a team, you have different members of the team, different levels of experience. Um and then also the way this is, how do you you manage that and and bring that into play?

SPEAKER_01 7:57

So to be fair, we are still doing uh piloting, testing the stuff. Uh so now we are just starting to. It is a few of us that started now, we're we're just building the build all almost finished the with the foundation. Now we have actually planned in the next week to have the workshop with others and to just put everyone to start using everyone from that moment, the full IG development. So we we'll see how it will turn out. Now, when we are working is as individuals on projects, it it works, it turns to be, I mean, amazing. I don't know what else to say. But what we wanted to do is uh before my idea was that before just uh giving the cloud code license to everyone and giving it okay, here it is, just start using it. My idea was to build foundation and process and just to have the proper uh configuration and proper uh setup for for the sub-agents and i agents itself. So basically, then they can just uh that all of us can get the basically same output, I would say it like that. Then that we have standardized planning agents, we have standardized one that will do development and review and testing, writing documentation. So now we are we uh I need to be uh uh honest, I'm quite amazed with the results that that I'm seeing now. And uh then this is coming again to different uh discussion. Uh, how we will uh build this uh work because if we are shifting it, should we shift from early rates to the either fixed price or some subscription? I I saw that you had uh guys that you had uh different podcast on on that topic, and I think that this is one of the biggest questions right now in the community how uh what to do, how to turn this uh instead of our hour-based uh work to uh value-based or what the value is actually. So I don't know.

SPEAKER_00 10:01

I I wrote about that. I wrote about that news uh I posted a newsletter not too long ago. I think it was this week, I had it released on Tuesday. You know, I I thought about if you if you if it takes you 10 hours to do work, uh it typical typically takes you ten hours to do work and you did it in one hour because you use AI and be efficient, you're if you're doing billable like time and material, you are essentially uh giving yourself a 90% pay cut, like revenue pay cut.

SPEAKER_02 10:30

So you're being penalized for being so efficient and using the right tools to there's a big discussion on how you can view that because is it a 90% pay cut or does the efficiency just bring down the the the value or the cost of it because you can you can produce more? I mean, that's a separate we we can jump to that to the the tail end, but that's a whole other topic of how you're looking at it from a pay cut, but harvesting you know, back but doing it manually versus using machine, is it still the same pay or the same value because you can do it faster? I mean, there's this there's a big topic there. Uh but jump back to the the team. So you're setting up uh a standard set of agents, and we can jump towards the end of this. I just want to not lose focus on the agentic development topic. The uh so you're you have your team and you're setting up uh rules basically, or instructions or agents. You know, that like I said, I say this every episode and everyone I talk with. I I forget what to call them because it's skills, agents, tasks, instructions, whatever. So you're coming up with a standard set of agents to assist you with your development, and then you'll have those in a repository, in essence, right? And then anyone on your team will load up uh those agents and use those agents to uh produce uh output. But you had mentioned you have an encoding agent, documentation agent, exactly research agent, I'm assuming. I mean, these are the typical ones I see encoding, documentation, research, testing. How are you coming up with those prompts to cre how are you creating those agent files?

Building A Shared Agent Foundation

SPEAKER_01 12:06

To be fair, and I need to give a huge shootout shout out to Jeremy here because uh he created this MCP. Um I forgot the name of the repo. Uh I think it is BC Code Intelligence or something like that. Yes, it's BC Code Intelligence. Absolutely. Yeah, and uh this is amazing. We just use that as our start uh starting point where we just uh I would say tweak that change this, delete it. We just you use that for to get idea, use some things from there, some things from some other parts. The most of the some things that we change to work for us, then we added some, then we see okay, this cloud uh tends to overcomplicate sometimes, then we have some agents that will do like that will simplify stuff. So I I mean it is just I mean but big, big thanks and big big kudos to to Jeremy for his solution. I think that this this is a really nice one. And I I saw that he I didn't use it, but I saw that he created uh VS Code extension uh as well for it. So I I I guess that uh my advice would be if people or companies didn't start using a genetic development, I I I would just I would recommend them to at least try that one and just to get some idea and just to see what they can get. And then of course then they can uh adjust it for their own needs. But uh from what often I'm seeing is that people still think that the AI is on the level of 2024 or or or maybe early 2025, I don't know, but it it improved a lot, so and uh you would be surprised, but at least here is a lot of still partners and individuals that are not using uh AI that much. And I I think that they will need to adapt fast, I would say.

SPEAKER_02 14:16

And um I would say so uh myself. It's uh it's a challenge. So you have a repository with your team, and then you're using Jeremy's uh BC Code Intelligence MCP.

SPEAKER_01 14:31

Um we are not using it itself, but we just used parts of it, or we we just uh used some of the some of the ideas because we let's say that we adjusted it quickly. We we changed it. I would not say completely, but a lot. We just used a lot of his best practices and and and some ideas from from it, and then we built let's say our own uh plug.

SPEAKER_02 15:03

So you you started with his as the foundation and then you Yeah, yeah. Exactly added additional agents for him. Uh additional agents and tweak the agents to to produce the output that you have. Um yeah. Yeah, uh what do you think the learning curve is for that? For somebody who's hasn't let's say if someone hasn't started working with this and and you had recommended that they take a look at that repository and they have some exposure to it. What do you think it would take for someone to who hasn't started working with this to get to a point where they could use it?

SPEAKER_01 15:41

I would say that uh they would need at least few weeks if they want to have something just for for their own to have something tailored for their own needs. Just to be but if they just want to create something, I mean the they can in a couple of days they can get completely fine output. I mean they can use just clothed code out of the box and they will still be able to get something. So it is uh now it now it depends what what they want as a result. But people want but if they want to just start playing and using it, that I think that the more they are playing, they will get more ideas and uh it will be uh yeah. But what I wanted to say also, sorry, we are now jumping from topic to topic uh we with this. Uh what what was my finding is it is still not that perfect, you know. You you need to have a few iterations with it. So you you need to understand what do you want as an output, and you need to understand how you want to achieve that. And there is the catch. Uh you you will just have multiple iterations, and after those iterations, you will get quite a good quite good solution solution in the end. But the catch is in those iterations where you are trying to guide him in or guide it in in a way. And uh for the job that I see here is at least for now is for those juniors. If you if you have some juniors that they are not completely sure what the output should be, uh to be I I didn't I don't have an answer to that how they how how they will get up to uh to speed uh with uh with everything. But uh I guess I guess this is the different uh topic and we'll see. Uh I I I have uh few juniors in the team, so we'll see how they will adapt. But uh at least from my point of view, it will be a bit harder maybe for them uh to produce good quality.

SPEAKER_02 17:49

It's uh you think it would be more challenging because they don't understand the application or understand and have some maturity with the language?

SPEAKER_01 17:59

Uh I would say I would say both, maybe more the application than the language itself, because then you can AAL's getting quite if you have a good set of these rules and stuff, it it will produce quite clean code and the language itself will be quite okay. But you need to understand uh I would say uh application itself as well.

SPEAKER_00 18:22

The the business the business itself, like uh the the functional side, right? You have to have a good understanding, yeah.

SPEAKER_01 18:30

Exactly. Uh and uh maybe maybe you know, maybe the more uh now I got this idea. Maybe the consultants will be new developers in the future because they they have this understanding of the functional side, and if you have good enough rules and everything, the quality of the code itself will be good. You need to check just uh whether output is right.

SPEAKER_02 18:52

Yeah, that's see that's a different twist because it it does seem with Business Central, it's already a predefined application, it's not something that is greenfield that somebody is just developing. So you do need to have an understanding of the application, and then also an understanding of the framework, technically, I believe, at this point, so that you can produce the output. So uh it's almost like back when we all first started many, many moons ago, where you did everything. You know, if you remember back uh in the early days of Navision, you didn't have a lot of uh talent and everybody did something different. They did you know, I know when I started, I did functional consulting, I did development, I did project management, I did pre-sales, I did every role. Um, so I think we all came from that. And it's almost like where it's going to it's almost like history repeats itself. It it's bringing it all back to you'll have somebody who just needs to orchestrate all of those tasks.

SPEAKER_00 19:52

You went through the struggle. I think, Milan, what you're saying is the junior ones have not really gone through that struggle. I think that's the that I think that's the the challenge is that because we're we're not in that space anymore where we did everything, now now they just go straight and somehow develop, but then use AI to code, but they're not really perhaps they're not coding, they're just not understanding the functional side of things or the business logic of things. That'd be inter yeah, it'd be really interesting. I think that's a challenge for sure.

SPEAKER_01 20:27

Yeah. I completely agree. But you you know, I need to also say, guys, uh I I'm not that old as you are, so when I started, I used some uh more mature versions of the visions.

SPEAKER_02 20:41

Thank you, thank you. No I I I I feel I feel good now. No, I I know, I know.

SPEAKER_01 20:47

Uh that's how I influence people, yeah. So invite me again. It's okay.

SPEAKER_02 20:56

I forget uh I forget how old I am sometimes because all of this new stuff makes me feel young again, uh uh excited and enthused about it. So uh my my body will tell you something different uh when I wake up in the morning, but um I I still feel like a couple but uh to to be fair, I was also the more I started working more and more with it, I felt this enthusiasm.

SPEAKER_01 21:19

I and I was so excited. I I I wasn't felt like for a while like that, and then I was just talking about that all the time, and my wife got sick of me because I was just talking about AI and genetic development and stuff. I mean, it's it's amazing.

Learning Curve And Junior Developer Risk

SPEAKER_02 21:33

But uh it is, it's it's uh it goes back with what I was saying is I stopped talking with people because I think everyone they don't have the either the enthusiasm enthusiasm where they look at me like I'm crazy because I was even trying to explain. I even took it at an old code base and said, do a code review, make suggestions, and put the suggestions into a file. And it comes out and it ranked. It one wrote a summary of what the m application did. I don't even know how, and it was correct. Two, it listed all of the things that it thought should be adjusted and ranked them in high, you know, critical, high, medium, below priority. And it gave them numbers. And then, you know, of course, it says, Oh, would you like me to fix it? And it's just like from just a simple, literally, all I said was, you know, I was just practicing and experimenting, I guess you could say, with different prompting. And I wanted to go as simple as that as um do a code review and propose or suggest changes and save them into a markdown file. And I was blown away.

SPEAKER_00 22:37

I have a quick question just for those that may maybe perhaps um if we could take uh maybe just a few minutes, or perhaps for people that may be interested in uh using Claude or AI for development for business central, uh especially for functional consultants, maybe they want to jump into that. Uh when you are talking about Claude and having a conversation with it, are you can you explain to me how someone would do that to get started? Do they do that on VS Code or they do it like a chat GPT? You know, someone had asked me that the other day, so I would just from for this podcast I would like to know.

SPEAKER_01 23:16

So there are different these uh solutions. You have clothed code, you have codex from from GPT, and what we are using is cloud code, and basically this one is a CLI tool. So what we are doing is we are chatting with AI, but from terminal. So you have like terminal, and then you are getting we are chatting with it, you are you are getting the response there, you can use some commands uh different, and you are installing different like plugins and stuff. Um and uh I I I am developer myself, you know, for for for a long time, and I I was not a big you know fan. It's not that I'm not a big fan, but I didn't use so much terminal. You have you know you have these hardcore guys that just enjoy terminal. I I was not fan of those, but now using uh cloud coding terminal just feels natural. I don't know how to explain. It is just uh so easy to use it there, it just perfectly makes sense there. So yeah, and then you are writing there, and it the thing automatically creates the files, output uh like documentation, whatever you want. It even compiles, run, test, publish. So it is the it is like that. So you need to have a VS Code, you need to have a license, and then you reinstall it, you set up, you install some plugins or some third like party solutions that you would like to use to get better uh results.

SPEAKER_02 24:47

Is there a reason why you chose Claude over using like GitHub Copilot within VS Code? I I know that GitHub Copilot within VS Code does have access to some of the anthropic models as well as some of the other models. And then Claude Code is a separate product, I guess you could say, or a separate subscription external, like you said. And I agree with you. Now going back to the command line interface, I almost feel like I'm one of those like hacker type people that you see on TV going and I have multiple terminal windows open. It's just it's it's like history repeats itself because I remember bay way back in the day, again, long before you, when you used to type in MS DOS, right? We used to use MS DOS to to do a lot of stuff. Um and now we're going back to that, it's it's kind of nostalgic and funny at the same time.

SPEAKER_01 25:37

Yeah, yeah, um it is. Uh just to answer the question about why we choose Clot Code. So before we used the GitHub Copilot, uh we had we had a license so you get some uh the normal way to GitHub Co pilot. And then one day uh I talked with a friend of mine like a couple of months ago, and he was so amazed with the uh with Cloud Code. He he just told me uh how they are using how he's using what he set up, what he achieved with it, how he's so happy with it, and then I just give it a try, and I like it, and then I said, okay, this is cool, we should start continuing using that. So there was no actual some uh I don't know scientific uh reason why why I trusted, even though I saw that a lot of people are happy, and then I saw that they are getting different results sometimes, and I see still that I mean both of the options are valid. I it it just I guess what what do you want, how do you want to use it maybe, and uh to be fair, I cannot say for certain because I uh I didn't use so much now like the latest time GitHub Copilot, but I I think that you will get similar output at the end of the day. That it there won't be so much differences, but maybe I'm wrong, but um yeah, no, it's uh it's it's uh it does seem to be a preference.

SPEAKER_02 27:04

I will say from my experience, I use a lot of VS code, uh GitHub Copilot, just because of I guess comfort in um the structure, but I also uh do uh a lot of clawed code. Um I haven't gone fully into using it yet, but it just feels like a different truck. That's I don't know how to explain it. It just seems to be a little more again, it just feels it. I can't even explain it. I can't even factually say again, saying I do about 50-50, right? So I can't even say that it one gives better results than the others. I mean, the different models will give you different results, but it's all a matter of how you prompt it in the context that you have uh within your session um that drives the output versus which framework I think in some cases.

SPEAKER_01 27:55

Yeah, I I I know I I I agree completely. But also it uh it's so natural, it just understands. You you don't even to try too much to give you insolvent, so who prompts it will just understand what you would say. I don't know.

SPEAKER_00 28:08

So for those that are listening, uh essentially they could get started, uh, probably read some documents on uh cloud.com is is where you would start an install. I think it walks you through it walks you through the installation process, whether you're in uh Mac or Windows.

SPEAKER_02 28:25

Yeah, well yeah, Cloud AI is the terminal, and then uh it will give you into the instructions if you want to uh install it. Um command line install, whether you're using Linux, Mac, or PC, Windows PC. Um then you can get into some of the other crazy things that people are creating. That's just unbelievable. Um it's um did you have you had any uh challenges as you've gone along this way with it? Um uh with the with adoption or with uh working with it? Any tips that you may say to somebody that's using it or has used it or hasn't used it, excuse me.

SPEAKER_01 29:08

No, this is a good question. Um I don't know what what what to answer on that one. Um I don't I don't I don't have much tips to others. I I just have tips to that they should start using if not cloud code and uh github or they just should start thinking about it rating development itself and then how they will set up things it depends on their own needs. Uh but uh I didn't I cannot remember that we had so much trouble with uh with any or anything during the way. Uh we built some PowerShell scripts, some simple scripts that we use, and uh then we figure out then it was easier for for Adrian to tell him if he is in that situation, use PowerShell scripts to do that and that instead of uh he trying to figure it some things by himself. But beside that, uh besides that, I don't I mean I don't have much else to say. Okay. I don't know. Look, what is your what is your maybe you can teach me something?

Claude Code CLI Versus Copilot

SPEAKER_02 30:27

Uh what is your uh do you have something for I'm the old guy, so uh well I I I I don't even know if it's it's very difficult to articulate some of this. I would say um learn how to prompt it. It can help you come up with the prompts. Uh if you want to start, you can have it take a look at a repository, and it could even come up with instruction files based on that repository if you wanted to come with it. Uh also um I'd say start small, say focused, and then when you're working on something, come up with a good plan first. Like you said, I think it's it's knowing the output or knowing what your end result is, and then coming up with a plan to go through there. And instead of expecting it to say, create me an application that tracks you know customers and their balances and then sends out invoices, and you know, instead of expecting it to do one big thing with one prompt, I say come up with a good plan and then iterate through those steps. Um and then keep track of it, and then I do reviews periodically as well, and I commit often. Uh cloud code, you can have it automatically commit. VS code, you can have it commit because I've had some instances where it's going really, really, really, really well, and then all of a sudden it went completely out into a different universe. And the easiest way back, some people say, well, just tell it, undo the last changes. Well, that doesn't always seem to work so well from uh from my experience. Again, the models change so frequently. Like now I'm I'm using Opus 4.6. I know Claude Sonnet 4.6 came out the other day, but Opus 4.6 came out last week at the time of this recording, and I haven't stopped using it since because I found that that is just amazing. But um, I would definitely say commit often so that you can roll back if needed because of um uh of that. Um, and then even for an organization, uh, I can come up with some more development tips, but I was gonna say one big thing from an organization, which I'm even trying to talk with everybody about, is it's more than just coding. Uh a lot of people hear about AI and they think it's just coding, but I've seen great documentation come out of it for modifications from a developer, far better than any documentation I've seen a developer, right? And I challenge anyone for that. Uh, and anyone who's been coding knows that. Uh, but even from business processes, I've seen people create great uh swimling diagrams, SVGs through Claude because it can create some diagrams. I've seen uh many uh what I would call uh product requirements documents. You know, you take your transcripts from your recordings from meetings with prospects or customers about things that they need, you can throw it into um Claude and it will create a great document for you. Um also with that, I always throw into that, you know, just be careful with the data that you put in there as well. Uh if I do anything with data, I always sanitize it, to be honest with you. Uh so I do any meetings or stuff like that. I actually go through and I'll do a replace and find for like names and such, and I'll do like person one, person two, just to make sure uh that everything stays safe. But I um uh and some uh yeah, so I think you could take a look at it from an organization from outside of just development, maybe some business process documents, some other types of automation uh that you can have uh and review and summary, and then from the development point of view, I think I covered some of those. I think you just have to practice and play with it. It's very hard to explain. Uh, but I think again, the prompting, uh, have copilot help you come up with prompt, help pro copilot help you come up with I I use the word copilot, but it's just whatever you're using.

SPEAKER_01 34:05

Yeah, yeah, yeah, yeah.

SPEAKER_02 34:06

Uh whatever you're using, have it help you come up with some of those agent files as well and those instruction files, and then you can go in and tweak them. Um, because I've never because I that's where I when I first started working, that's what my biggest struggle is everyone's talking about these cool things called agent files. Um, and this was even before, because I did speak with Jeremy as he was doing the BC code intelligence. It was even before he did that because it's it's like we it's like a blank piece of paper that say you put in a title, you put in a description. Okay, well, what do I put? But now if you have the the the uh whichever model you're using or framework, create the file for you, then you can see it and then you can adjust it and tweak it. Um it's crazy, it's just crazy.

SPEAKER_01 34:49

It's crazy, but I I I I would say that uh you you made one really good, important one that we also do, and which is the key point is that you need to commit often small things, and you should do it the much as possible because then it is easier to just roll back roll back if things go south.

SPEAKER_02 35:09

Yes, exactly. Exactly. Um and again, a commit doesn't mean that you uh some people have a misunderstanding that a commit, like oh, I'm committing, I'm saving it, it's done, it's final. No, a commit is just a checkpoint, right? It's not final until you merge it into the main branch, and you then you create your artifacts for publication. Then it's just like anything you're doing. You can commit, you can save. Uh, but when you're dealing with multiple files, the undo isn't just a control Z. You know, the undo may require you to do quite a bit, so having the option to yeah.

SPEAKER_01 35:45

No, um I I completely agree. Uh I wanted to say something, but then I forgot.

SPEAKER_02 35:53

So I that's the story of my life. The story of my life. Uh hopefully it comes back to you. Hopefully it comes back to you. Uh, and you'll see.

SPEAKER_01 36:01

Uh yeah, I wanted to answer you. Uh, did you did you try cursor? Uh or because I I I use it for a while, I just play with it, and then it didn't feel so so natural for me, and then I went back to uh VS Code. Maybe I didn't give it enough opportunity, but uh I don't know. I I know that some of the people are big fans of it.

SPEAKER_02 36:22

I haven't used Cursor, and my first hesitation, to be honest, was because there was a charge. Uh, and then others were starting to use it, and I've talked with them about using it, and they'd said, Oh, cursor. And I think you have to it depends on where you were in time as well. Because I think cursor at one point was you know, it was a fork of Visual Studio Code. So they were adding features and functionality that made the development experience better, uh, you know, and I'm better in some opinions than Visual Studio Code. But then if you pay attention, if you wait a few minutes, those features showed up in Visual Studio Code, right? Do you see where like it's it's they were ahead of Visual Studio Code, then all of a sudden the features would come back, and now to me it looks like the gap isn't so big or isn't so long. So if you wait long enough, and then um I said to someone last night, I said, Why do I want to learn another IDE? Uh I know it's the same, but it's it's I'm trying to just stay focused with it. Uh uh do what I do well in Visual Studio Code, but then also now I'm starting to do a lot of command line. So that that's why I'm saying, so I have the little mix. I didn't see the value for me uh in moving forward because I don't see it having anything better than anything better that I can say, oh wow, this is worth me to move. I mean, someone else may have a different opinion, but in my opinion, I just stick with uh command line and VS Code. And we'll probably all be at VS Code. I mean, excuse me, we'll probably all be command line again soon, back like when we first started. Um now this would be before your time. I heard somebody make a reference to it. I was listening to a podcast yesterday. They the VMS VAX. Do you know what a VMS VAX was?

SPEAKER_01 38:07

No, I don't know. No, I need to Google. Chris, do you know what a VAX was?

SPEAKER_00 38:13

VMX, VMX VAX.

SPEAKER_02 38:15

VMS VAX, yeah.

SPEAKER_00 38:18

It sounds familiar, but I don't know exactly what it was what it is. But I've heard one.

SPEAKER_02 38:23

When I first started programming, I was learning assembly and Pascal, and we worked on a VMS VAX.

SPEAKER_01 38:30

A VMS VAX was 1977.

SPEAKER_00 38:36

He's gonna age himself in this podcast. You can't hide in your look now, Brad. You can't hide in your look.

SPEAKER_02 38:47

Listen, just because it was invented in 1977 doesn't mean I'm old because the PC was invented in 19 what 80? So you're still using PCs today. But no, when I was in school, we had a VMS VAX and it was like a uh a character, it was a CLI mainframe computer. So you had a terminal connected to a huge mainframe computer, it was green screened, then you had the same thing with the printer. So we used to code on a VAX. And it was like a central mainframe computer, and I it's almost like history is repeating itself because now we're going back to now we're doing again use clawed code command line. Right? Where's the processing power? Right now they use the processing power on your computer. But do you ever think and they also do you ever think that maybe now all this AI stuff again will just all be sent out to someplace for that pro centralized processing? Uh I know some does you know, I'm not going to say what gets sent out from each of these models are not sent out. That's not the topic of discussion. But it's almost like we're going back to that point of like centralizing processing. Yeah. So you don't remember a VMS VAX then, huh?

SPEAKER_01 39:53

No, I don't remember, but yeah, no, I uh I see your analogy how and I understand now. But uh Yeah, I mean this is um I I I'm not sure. I I am so I'm uh the the more I uh me into it the more I'm I when I say confused, but I I'm not sure what to uh what to expect in uh six months, one year, two years, three years, so uh the things are like I think everyone's a lot in that pa space right now, like you know, you want to do so much and then it'll change next week, you know, something different or something cool in the future.

SPEAKER_02 40:33

It's it's uh I'm at two days. You guys can be at six months, three years, two years. I'm at like what will change in two days.

Prompting Habits And Safety With Data

SPEAKER_00 40:41

I think at this point it's like you're trying, you know, uh for me, and I've shared this in the last episode, which hasn't been released as of this recording, um you know, being able to think about outside, even thinking about outside business central. Uh you know, and and I shared it with with the last episode where I wanted to create a website from scratch, hosted on Vercel, you know, and and it built something for me in a couple hours, you know, two hours. And and I'm done with that. You know, what am I gonna do next? What what would like to do next? And uh so you're you're s I start to think about things that I've been wanting to do from a personal perspective, a personal project versus just working on business central and uh you know there's it removed that barrier for me because like then I I'm allowed to do other things that I don't necessarily know the the the code language itself, but I I know what kind of result I want.

SPEAKER_02 41:42

Yeah. Right. But it's that software. It's software, it's it's key for software, it's the the philosophical discussion to software become disposable, but it's outside of software that I'm really starting to and I think it was because of I don't want to say because of open claw, but because maybe I was trying to think of other ways in my life it can help me with with organizing and creating plans and and putting things together for me outside of code um with documentation and such. I mean, coding is a is a great thing, but coding is such a small, I don't want to say a small portion, it's it's only one portion of a business um that you need to take into consideration. I mean data analysis, data consolidation, documentation, um, workflow analysis and stuff. You can do a lot of things with it.

SPEAKER_01 42:36

Yeah, no, I I agree definitely. We just had luck or unlock that we are the first in the that are affected with this, but I just wanted to also ask Chris uh did you start using AI in general more and more in uh your because you are a functional guy, right? Did you start using it more in uh your daily work? And do you see that you are also getting more and more efficient? And uh because I think that this will just slightly start, or maybe it even started moving from okay, developers will be much more efficient, and okay, there are a lot of questions there. How we will uh what we'll do and how we will just earn money and uh what do we take extra time. But uh other you know, I think that in order for us to be able to succeed, functional side consultants need to be able to scale as well with us. So you also need to be faster and to do do things more efficiently, I would say, because or or automize a lot of stuff. Anyway, yeah, the question is how it's going with you, yeah. Where you are currently with AI.

SPEAKER_00 43:40

You're right, like you use it, you know. I Brad had uh kind of brought this up. You you use it sort of beyond just development, you know, and and being able to streamline some of your processes, you know, create templates for you, which usually takes you a lot of time. Documentation is a big, big problem in the past, right? So uh, you know you use a a component of AI for those things uh even even even improving your email writing you know English is not my first language so you know trying to translate that to something that makes it easy for someone to understand that certainly helped you know quite a bit it's it again it removes some of those barriers uh uh in terms of like functional documents uh you could have a template and have a build based upon maybe what a developer had built for you and and kind of translate that as well. So it it really is uh making a lot of you know your day to day more efficient efficient. Now to to answer to the other part of your question is you know what am I doing after that uh you know you do you do more work. You know it goes it goes to that um and I wrote about this too because I felt that when I was starting to use AI is that because you're so efficient on other things you know it does it does it create more stress for you because now you have this extra time that means you're expected to do more work. Because it's saving you 80% of time uh does that mean you have to do 80% more work than what you've already did so you're kind of penalized or does it give you more um uh more time for personal projects right so it really it really depends on that so kinda g it kind of gets into that philosophical thing but uh going back to your question yeah I mean stuff like uh even creating files for you that you typically do not think about creating is is making suggestions.

SPEAKER_02 45:38

So you get the a full kind of a full stack uh solution that you're providing to your clients it's a wild world it is it is and and what to do with the time is uh I I think it'll all level out at some point. I mean I think we're in we'll have to go through a transition period for people to adopt it because I think I think we're still in a world where our our our universe and I'm talking about the the three of us here is a world where a lot of people are talking about and using AI because it's it's in our space but I think there are others I talked with someone even uh another business central partner and I I was talking to them about some AI stuff and they hadn't even used it yet. So I think we have I I think the uh it's almost like you it's it's you're exposed to what your surroundings are, right? So we're exposed with a bunch of people who are into this but I still think there's a large portion of the population that doesn't understand what it is. And I think once that portion of the population starts adopting it and I'm not just saying and again I'm not saying it's only related to business central there's a lot of other technical people using it but I'm saying of the nine billion people on the planet I think you're st we're still a small fraction of the a number of people using it. But once it gets widely adopted into other areas that's when I think we'll have that transition of what do you do with your time.

SPEAKER_00 47:06

Yeah you made it you made a good point there Brad about like what what is you know how what other industries could utilize AI because you're right we're our in our universe we you know we use it on a daily basis now but there's so many industries out there that haven't really touched it or where would it fit uh within their industry or where would it fit within their space right so uh and that's a big component like for example um uh teachers teachers use uh AI to some degree but they're still limited in books and things like that but they could certainly use AI to build maybe a curriculum or a uh a structure what their day looks like or the week looks like and and so that I think there's so many different places that you can utilize AI for sure. I think I think uh uh medicine they're slowly starting to pick that up uh a lot of doctors are still kind of like apprehensive because it may get it may get you know they may read it wrong or something like that. Um but again there's so many different industries that could that can utilize it.

Real Business Central Extension Win

SPEAKER_01 48:19

And you you guys said correctly and I also often forget that we are in um I would not say bubble but often when you are the the discussing this you are discussing this with uh similar people which have similar um say uh uh not affiliation but um well they have similar interest similar drive similar exposure interest yeah I was meeting yeah exactly and then they're they're also so into it and then you think everyone else is like that but then you turn it turns out that you don't need to go far you can just find even BC partners as you said that are not uh using so then you still see I I want to go back to another we were talking about the development for not to jump back to it too much.

SPEAKER_02 49:04

Um in your in in your experience with development with it um because BC has an individual framework do you find yourself having it create more new or or successful with new extensions or new features that are standalone versus extending the core business central functionality. And what I mean by that is maybe you want to create a new feature that's a standalone feature for tracking certificates. I'm just making this up okay um or maybe now let's say let's not have our certificate management process let's say that business central has the existing sales order process and now I need to add functionality to the sales order process. Have you had what's your experience with being able to extend business central um and when I say extend with tables code you know events with code units pages and such so my current in my current role or company where we are working we are basically partner which is implementing system to customers.

SPEAKER_01 50:13

So we are just uh checking their their needs and then adjusting the system for them. So basically that means mostly extending the functionality or adding some features about that. In my previous role I worked for some ISV in in in retail where we were just basically adding a lot of new features.

SPEAKER_02 50:35

Now it is quite I would say now we have a lot of integr currently I'm working mostly with we have a lot of integrations we are having a lot of extending of the functionality here and then we we are adding a new new feature function new like functionalities from scratch but it this is not our main I would say so AI has been successful in being able to pick up on the extensions I'm I'm just curious about the ability to use cloud to extend the framework versus build from scratch.

SPEAKER_01 51:09

So I I I will give you one example now actually I have a good example but because uh I was doing this with uh we are doing preparing this uh internet development framework so others can use it and what I was actually curious is I found a bunch of change orders from from before that we implement to our customers and I just wanted to now develop them with uh AI with uh cloud call just to see how much faster I am actually and then there was one just the first one that I found it was run it was estimated was 24 hours just it was uh something with deferrals posting some some posting things differently with uh with deferrals and I was uh a bit uh and to be fair this 24 hours developer who was working on it he spent all of the hours and he was having some challenges with the right um events to subscribe to get the correct solution and that was fine customer was happy with the solution and now a couple months after when we have this genetic development I I was so lazy I didn't want to use the trade whole change order I I just took a like few sentences from the change order I pasted to our um agent he created it created some plan I checked the plan I didn't want to bother so much I just said okay I just continue developing the solution in a couple of months a couple of minutes I I got something and then I saw the code looks kind of okay I don't know I just put it on uh on sandbox I tested and I see okay it compiles I mean it's published there but it is um it's not working and then I came back and I said this is not working this uh gl account should be have the uh this GL entry should have this GL account uh you are not using it from the setup blah blah blah and I had like two or three iterations like that in four iterations I got like really clean code which is perfectly done solution production ready and I spent like one hour maybe a bit more than one hour with completely testing it's amazing uh I mean it's it's amazing but I uh it's it's crazy but I needed to I I knew what I wanted I I knew what I wanted to have as output and I knew that uh I had GL entries I I tested I posted I was posting uh sales orders uh tracking like whether these are entries correct there and basically you need to know what what do you want as an output and then once you have this it is like amazing.

SPEAKER_02 53:39

You you there's another tip that I had from what you had said. And um I do typically say this and I say this to everyone that don't quit after the first result. And you you highlighted it perfectly there that sometimes you have to reiterate as you go through uh the coding don't think that the first is going to be perfect. And if it's not it's okay just go back and and tweak it like you had mentioned. Another question I have not done this and I haven't asked anybody this before and I haven't seen anybody talk about it. Have you had it write any reports uh that's a good one I really honestly haven't talked with anybody about it and I'm gonna start asking everybody because everyone's talking about code uh how about are you talking about word layout?

Report Layout Automation And Wrap Up

SPEAKER_01 54:28

Any reports even our DLC any reports any reports so I I I I I can say something um about that because I had this in my we have a backlog of ideas and um just for the next week our idea is to start testing with uh word layouts and I don't know what we'll get but we have this in our backlog that we want to do but friend of mine uh who was so amazed about this cloud hold who who I got this idea from in the start they are still using a lot of RDLC and RDLC is nothing else than uh XML at the uh at the end uh so they built a small like solution across of it so they they build a small like page and set up where you can like some generic solution where you can tweak how you want to use the report and he said that it creates perfect RDLC reports because it is just X uh it is just XML and it knows the structure it knows how to to change the stuff that the output of these RDLC are great. I didn't try it on those and I don't we are using just only word layouts and I I just want to try it with Word.

SPEAKER_02 55:40

Word is also consists I think of XML and some other things so I I guess that it should be possible but I don't uh I I you can I I can let you know if you can yeah let us know let us know when you do that because I'm curious and you'll probably see me start to ask everybody about that because I there's uh I I talk often about this stuff but that is one topic that I haven't seen come up so if anyone has worked with word layouts or or IDLC layouts with reports let me know I'm curious to know the results because uh no I I'm curious as well why this is on my backlog because uh we have a lot of not a lot but every customer wants to have some specific uh reports now and they have a lot of these great ideas what do you what they want to change and I have a junior I would say a a bit more than junior developer uh spending a lot of time uh adjusting these layouts and and stuff working there so now I just want to see how whether we can automate this a lot so I I I I'm also curious so I'm also curious what other people will say when you ask them but I I hope that uh there will be a good probability that we'll have a good result.

SPEAKER_00 56:49

I hope so I hope so because uh that's that's uh an important part of it so no well sir thank you very much for taking the time to speak with I could talk with you about this stuff for days I know that uh it's it's getting late over there uh for you yeah it's 5 p.m so now almost I will finish uh I need to answer some emails answer some messages and then I will start uh packing uh for day uh but uh and I for I know for you guys it's your day day is just starting so I guess it will be a busy day for me it's it's always a busy day yeah no it's lunchtime for me so I I'm going to have to I okay it is oh I see uh I mean once again thanks a lot for having me it's I am um I really enjoyed this conversation and uh as you said we could discuss this for uh background no thank you again I appreciate you taking the time to spend with us uh because again you know once you spend any time you spend with us you're not you're not spending doing something else and you don't get the time back we do appreciate it that's that's okay I have uh I have agents doing the development that's what it is the agents you can before you come to talk to uh with us you can uh have your agents are you going to directions North America no unfortunately not uh there is a bunch of uh conferences here in uh in Europe in April and May so uh I just decided to to go to these I think I I I hope I will be there at summit so that'd be great to see you again that would be uh excellent uh in Nashville so hopefully we'll be able to see you in October if um we don't get a chance to get over to Europe before then but again thank you for your time we appreciate it and uh look forward to talking with you again soon thank you have a nice day bye ciao ciao bye thank you Chris for your time for another episode of In the Dynamics Corner Chair and thank you to our guests for participating thank you Brad for your time it is a wonderful episode of Dynamics Corner Chair I would also like to thank our guests for join 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-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 and my Twitter handle is Matalino16 and see you can see those links down below in the show notes again thank you everyone thank you and take care