moderator: 0:00

Welcome to the scrum.org community Podcast, the podcast from the home of Scrum. In this podcast, agile experts, including professional scrum trainers and other industry thought leaders, share their stories and experiences. We also explore hot topics in our space with thought provoking, challenging, energetic discussions. We hope you enjoy this episode.

Dave West: 0:25

Hello and welcome to the scrum.org community podcast. I'm your host. Dave West, CEO here@scrum.org in today's podcast, we're talking to Thomas Marge, Head of Product ops at odayvo. Odavo is leading the way in property management and residential services. Some of you may know Thomas from his work at McKinsey and the research that we did together, actually on the pursuit of value and some other things. Welcome to the podcast, Thomas, thanks.

Tomasz Maj: 0:55

Thanks for having me. Very happy to be here.

Dave West: 0:58

It's great to have you here, because you know what's interesting since you left McKinsey, you took this role at odavo And what's interesting is that the world of software development, product development has shifted quite considerably during those few years. So I'd love to get your take on that. But before we do that, the listeners would love to know a little bit about what you do at adevo. And you know what they do, and you know sort of that kind of stuff.

Tomasz Maj: 1:33

Well, yeah, a Devo brings a lot of a lot of small, small to medium sized property residential property managers across, let's say, the northern hemisphere. So we're present in Europe. We're present in the United States and in Mexico. So pretty much the big part of the northern hemisphere. And what we just we bring a lot of small to medium property managers to do a common to a common place, right? So it's a very fragmented industry, and our main goal is to create a leading a leading property resident. It's important here that residential property manager in the world, that's that's the goal, that's the ambition, and we also see the power of technology in that. So my main job is to, if you take, if you you know, it's a lot of small companies, you have to bring them together. We also have to bring a lot of that tech together. So how do we create an operating system? How do we build a model of product engineering? And these businesses. We call them operational companies, all work together. So my main job is to, is to bring that, create that glue that binds all those things together. So we have multiple applications and multiple systems in in in Finland, Sweden, the UK, Spain, Portugal, Italy, United States, Mexico and all those countries are different, different legislation, different use cases, and many times, also different types of properties. So we're looking, where do we want to where we want to bring things together, but where do we want to keep them autonomous, not to disrupt, you know, their day to day operations, they're not and our ambition was never to have one system to rule them all, but to really apply strategic, scalable thinking in building and building a technology set up and an operational setup that fits that. So that's my job, is to sort of bring all those things together and have that happen which is super exciting and super fun.

Dave West: 3:44

Oh, I can imagine it is. So you've got to sort of balance This centralization, this single platform, with the need for very diverse, very different.

Tomasz Maj: 3:55

I was it's a good question. I don't think we have multiple platforms, multiple applications there. We've tried to bring everyone on to assume it doesn't make sense, right? Because Sweden with with few big like, there is a there is not a lot of there's couple of bigger players in Sweden. Also, Sweden is a comparably smaller when it comes to population country than the UK, where you have many, many other smaller companies. Or you go to Germany, where the fragmentation is so high, you know, the biggest guys have around 2% of the market, right? So we're actually more about enabling that diversity without sort of blocking. So while getting ready, while being ready to scale, right? That's, that's the number one like, we're going to grow. We're going to continue growing. We we double in size almost every year. So we have to start continuously, have embedded an operating model that scales with that ambition. So we do not have a single, single platform which is has its upsides and downsides. A lot of things, is that there's a lot of fragmentation, right? On the other hand, we are not botched up. We don't slow down, because one country is waiting for another country,

Dave West: 5:10

and it also gives you the opportunity, by not having one platform, to do a lot of very quick I don't want to call it experimentation, because that's always a bad word, right? But innovation, that's much better word yes, and try it on one platform or one of the products, sorry, and then learn from that and then drive it into the other other products. Yes, necessary, exactly.

Tomasz Maj: 5:34

And I mean with, with the two letter word that everyone is using, this is, I think, instrumental, like, you know, the two letter word, yeah. And that makes it easier for us, too, and

Dave West: 5:47

that, and that is the reason I, you know, I have to say listeners, the reason why I invited Thomas to the podcast today is because we've stayed in touch. You know, it's great to have somebody that's driving such level of innovation and in a company that's growing, and that's, you know, sort of got a solid set of very human principles as well. So I've stayed in touch, and during these conversations, there's been this really quite interesting conversation about AI, and so the rest of the podcast, we're going to really focus on that. I'd love to get your take, you know, over the last 18 months, I guess two years, AI has suddenly become something that was machine learning, and we talked about it a little bit, to something that's driving everything. How has it affected odevo and your teams and your organization and what have What's your take on it?

Tomasz Maj: 6:41

Thomas, yeah. I mean, I think one good news is we saw AI was coming very, quite early. Even when I joined, there was already an existing AI team that was building. Well, we have our own Devo GPT, for example, right? So that way customer data doesn't get out and is not used by other models. We realize that it's very important data. Data is is clear here, critical here, so we wanted to sort of keep it in house. So we did that. We've been building a lot of simple tools, a lot of very useful AI tools for our industry, the property management industry, still, in many places, underdeveloped, they don't have access to all the tools you would think you would take as standard, right? I mean, there are still places where pen and paper is still used to manage properties. The posted in someone's notebook could be worth 1000 euros and a it's posted until it gets booked, so you have elements of that. So we started using AI, very early on to help in processing invoices. And, I mean, some of our companies have to have separate warehouses just they're all that paper that gets processed, right? So we started using that. And as I joined I, you know, I got into it, right? I started using ODU GPT, chat GPT. I remember when some when I heard about, you know, hey, I'm going to try and have and with chat GPT, build an application. I remember that in two years ago, I was like, that can be hardened. Oh, it was super difficult, right? So it crashed and burned on it when it had to set up a database, but and we progress, and we constantly saw, you know, copilot, GitHub, copilot, and other tools coming up. And they they were showing potential, especially like documentation. And we saw those types of use cases, like you can speed up that testing. And then over time it got just got faster and faster and faster was like every other mile, I say, every three months, you would go and it would just be another jump, right? So we, we constantly were on top of it, but yet, in an industry like ours, our implications are not sophisticated, big systems, right? They're usually like, I said, small tailored for the need of the of the country, of the business we work with. So yeah, that was it. Do you

Dave West: 9:08

actually think, though, Thomas, that actually that led you, that put you in a much better position because of the size. And I'm not saying these systems aren't complex, because I'm sure they are each of these ones, but because of the size and the footprint that gave you the ability to apply AI quite rapidly, without having to worry about this sort of like this platform that does everything that is being used in a million different ways that nobody really understood. Do you think that put you in a better position?

Tomasz Maj: 9:40

I think absolutely, I think the one of the biggest ones was how much people would impact, right? So we have some platforms that are super early in their stage, where we can go a bit bolder, where we have others that are much have a much more dominant place on the market, right? So they're like, oh, take it easy. Like. It's not overdo it so, but that, you know, having that diversity, that diversification of different platforms with different sort of business impact, absolutely allows us to be bold in other places, but much more cautious in in others. And but the upside, the like the flip side of that was for the ones where it's much more complex, the systems are bigger, we saw a big uptake in what now everyone calls vibe engineering or vibe coding, right? So that was interesting. That was an interesting sort of change recently. This has been recently in the last, say,

Dave West: 10:41

five months. So explain to me. So you've say, we've got this app that or a product, I guess we should call it a product. This product, you've got an engineering team around it well understood that they're delivering at the moment. Where does vibe coding fit into that? You've already got, like, tests, you've already got a build pipeline. You've already got the business are working with product managers in a very defined way. How does vibe coding change that?

Tomasz Maj: 11:12

It's, it's, I think, democratizes access to features in a way. I don't know that's the right term, but you know, you you sometimes need to process data. You need to and then you have to download that data. You get a CSV, or you ask for a SQL right, to give you a lot of data. And then, usually that would then go through some sort of a process, some Excel wizard would process and pull out, and then it's a lot of things, a lot of steps. All those steps can be, can be now put into a vibe, coded app with, with this, with a power of a prompt, right? Or several prompts. And you don't have to do a lot of like connecting, like API, MCP, all that you can just add. You can still use that CSV, but now build in an upload feature make sure you don't you know what you're doing from a privacy security perspective, or at least pounded the tool you're using can and you're now able to get information much quicker without needing to be an Excel or A data wizard, right?

Dave West: 12:21

So who the IS? The business, like the

Tomasz Maj: 12:23

product, our operational companies, the accountants in our companies, the customer managers in our companies, probably soon, also the maintenance people that you know fix the elevators or make sure they're the radiator is working, will also be able to do that they're using it. I mean, the it's just, it's just very easy, so, and it's impressive, it's impressive.

Dave West: 12:49

I mean, it sounds, I mean, that is the ultimate, you know, push left, as it were, isn't it? The ability to move, how capabilities are being delivered on your on your products, from your it, your software developers to the field, which is really, really exciting. But how did you enable that? Do you have an machine learning ops or an AI ops or no?

Tomasz Maj: 13:19

It was a great question. It was organic. We didn't do anything. We people just got heard about the tool and started using it, and we actually had to, we now, then had to react to that and say, hey, when you're using it, I think it's a good idea. But remember that? Remember about privacy, about security, about which data to use, which data not to use, so and but then was a very quick turnaround, very quickly these organizations, these operational companies, built in Guardians, as they call them to you know, here go through this regulation and get a certificate, right or remember, this is private data. This is sensitive data. You should never use this, ask for permission for that and then continue doing. And of course, there's limitations, right? The appetite grows with Vibe coding. The more you're able to do, the more you want to do, because it's just so easy. And I see like on the link to LinkedIn is sphere? Is that a term? A lot of hate, right? A lot of haters going, Oh, if you don't understand the code you can, you shouldn't be building it, or like all of this. And I think those are the guys that will soon realize this. Their world is ending faster than they can realize because it is becoming sort of coding, building applications, sometimes bigger, more complex than we could imagine before. Will be easier and easier over and over, like I've seen, I've been using these vibe coding tools, especially one for for seven. Eight months now, and I remember when I started out, it was very bad, especially on the security side, data processing side, slow. As you as the database grew, it got slower. The security was a bit flimsy. You know, you were afraid to create a user you want. It was more like a one pager, right? Like a landing page or static content, right? So instead of just you, you know, using vix or other one, you just do this because you it was cooler, it was AI, but now you can do full SSO, you can connect it to databases. You can you can have email like, it is quite impressive, and it continuously, like, exponentially improving over and over, right? So? And that's one of the I see that as one of the biggest shifts, the democratization of application development. Like anyone can build it, and it takes 30 minutes to get something that you could probably use easily. What is addictive to but it's a

Dave West: 16:02

little addictive. Did you do anything to the underlying products to enable your business use your you know your business users the to to vibe on that data, or are they really just vibing off an extract CSV that you that your product allows it to create nowhere.

Tomasz Maj: 16:22

We're actively depending on the market. We're actively looking for opportunities to support this, while we still want to keep the core competencies. I mean, you know, core competencies there, right? We want to have those in especially data, is quite we don't want sort of complete freedom. You do what you want, you know, read and write. We do want to have some sort of control, but on these, like edge cases, we want to enable that. Because, on the other hand, why would we want to be building all these edge cases? Right? Whether it's a small app just to manage parking transfers of a building, or stuff like that. Like, no, you build it. It's fine, right? It's okay, but we're starting, it's sort of a it's something that's new to us. I mean, I think for everybody, the speed of it. I mean, I still feel like I went for Christmas break, you know, I, you know, I ate my delicious pierogi and all the other great Polish food I came out of this Christmas sleep. And in January, it was like, Wait, this is a different what happened? Like, it's a different world.

Dave West: 17:36

Yeah, there was releases of some of the platforms over the in early January that have, and actually early February in particular, which has completely revolutionized how you can rapidly build, build software. And that's that seems to be increasing so, so you just allowed it to happen. You didn't have a more structured, sort of like aI ops, kind of like, but you brought up security and governance and and obviously these businesses are very, I mean, they're very aware of the markets they operate in. So then they took that, that maturity, and particularly Scandinavia, I'm not sure I trust the UK ones quite as, no, no, I'm kidding, but, but yeah, so they took it on board and started to work with it in terms of it. Has it had an impact in your development teams? Is it more the development teams have just been using more augmentation around testing, writing more code, you know, no,

Tomasz Maj: 18:42

no, no, we actually, I think in, I think it was September, we had a realization that AI is coming. We can't fight it, and the fight is the wrong word, but we have to embrace it, and I think we have to support our engineers with it. So we went out, we sent we searched, we asked our friends and our colleagues for companies that could do a proper end to end AI training, like for engineers and AI training that takes you from the very basic of you know, these are the European standards of AI used, remember we're in Europe, so GDPR and all those things is quite important. And we, we do want to do more than we have to whatever comes to stuff like that. So we want to do a training from the basics to agentic development. So it will be a five step learning journey from from zero to hero, right? So ending with agentic, with stuff like B mad Gastown or or cloud flow and all those, all those tools, starting with like learning, just here is prompting. Here is I was, let's talk about prompting. Let's talk about terminal so on spec, driven all the way to agentic, we didn't know that this was a, I mean. We knew this was a big, big deal, but over the next four months, it turned out to be slammed up, right? I mean that where most of the we had cohorts. So we had five cohorts. We didn't want to we had, we broke people into small groups. All the cohorts finished this training mid February, now, and it was, it was perfectly sort of landing right when the big the boom happened, right? So now I would say 95% of our engineers have the basic understanding all the way to agenting development, and, of course, depending on the platform, the language, and because we also don't have all many of our applications use different languages, different they have different setups, yet we see a huge uptick. All engineers are using AI, some of course, most of them are not building agents, but many of them are, are using spec driven design or specs to build stuff. They're all heavily using, you know, in their like IDEs and working with apps, we have several engineers, very senior engineers, who said, I haven't actually touched syntax since January, right? Because I don't need to. I just work with Claude and I check, I code review. It's what I don't actually write the code myself. So I think that was a perfectly sort of, that's how we enabled it.

Dave West: 21:31

And do you do you have when you build a team, or do you have a genetic, sort of, like goal based agents that are sitting to do TDD, BDD, or to do code review or to do or is it a little bit more sort of fluid with, you know, an engineer or or a tester? Or is it, is it how you partitioning work? What are the you allowing development teams to make those decisions however they want, and developers to make those decisions?

Tomasz Maj: 22:09

Yeah, we're not, like, setting a standard, and I think giving we're not going like you need to all fall into this. Let's say diva ways exactly, devo ways of working. Every team has to do that. No, we are. We're giving a lot of freedom, but we're all encouraging and sort of expecting that they go beyond copy pasting code into into a Devo or whatever audio, gpt.com, or something. That's not what we want. We want them to actually be actively using it, but sure, not everyone, not everyone still uses it to that extent of using it. We do have some teams that do that. They have set out agentic development flows, while others are more focused on specs, while others are just more sort of working on a code base, but working with, with a co pilot of sorts, whether it's Microsoft, GitHub, or whatever they they prefer Yes, and I think because, because the main reason for that, because it's just so fast, like, yeah, they don't know.

Dave West: 23:19

It's impossible, almost, to define those standards and put those best practices in place, because they're just continuously emerging and then changing

Tomasz Maj: 23:29

Yes, or there is like Yes. And on the other hand, you don't want to sort of say, Oh, I know the way, because you don't. And we have to create that freedom right? And that's why we are doing more and more. I think what's becoming more clear is like this, sharing of experiences and building. We're constantly trying to find ways to to, you know, show and tell. Build, Build collaboration. We have introduced in last year lead engineers as a as a role within a team to then they should also now facilitate that knowledge sharing and and be the lead of that team in a way to also experiment and help. So that's how we're applying. But trying to come up with like, Oh, this is how we're going to do a agentic development. Sure. We can say for the month of March,

Dave West: 24:25

because they'll only be out of date by the time

Tomasz Maj: 24:28

April comes. Here is a new one, right?

Dave West: 24:32

It is interesting, because historically, software development, product development has, historically, your sort of best practices emerge in the industry. They may come from an individual often. Do you know, you sort of pick them up on Lenny's podcast, or, you know, pragmatic engineer, or whatever the source is. Back in the day, it used to be magazines when I was a software engineer. Oh, my God, that's, that's so true. Oh. Conferences you'd go. But the pace now, it's not coming so much. It really is that fusion between technology capability and human being, and it's happening at such a vast speed. It's not one individual that's driving this. It's not like the guy invented Aspect Oriented Programming, or the the early, you know, object oriented, or the TDD guys, you know, Ken back, etc. It's sort of everywhere and nowhere all at the same time. It's a very different, yeah, it's very odd to observe. It's kind of different than anything I've ever seen before. Do you feel the same way?

Tomasz Maj: 25:44

Thomas, absolutely. I mean, I think everyone is overwhelmed, right? I mean, if you're not, then you're doing something wrong. I mean, I, even, with Claude, built an AI digest that every, I think at one point you have to so it just scrapes the Internet all the different feeds and then populates a Slack channel for my, for me so and my, the whole organization is a public one. So you see the top 10 sort of big news updates scored eight to 10, because you just have to keep up somehow, right? And with some sort of a smart summary there. But I think one, one thing that's worrying me is that everyone is sort of going very wide. You know, everyone is looking like another option here, another option here, going wide. But there's no one. There's very little people who are saying, hey, where do we converse? Let's open it up again. But at least at this point, let's say this is what we're trying, I think we're trying to sort of bring that sort of, I don't know, get control of this chaos, right? I think this is some one of the things like, the loop will be immensely fast, right? So, if you're, if you're, if you bought a project from someone or someone or you have a PMO that told you, here is our two year plan to to do an AI transformation. Then I think that's where you should ask yourself, Does this make sense? Have you ever seen a plan for exponential growth? I don't think so. So it's one of those things where you should be asking yourself, Is that the right one you should also converge, then again go wide. But you have to, you have, we have to have more of those cycles, because now everyone is just going wild, playing with everything. No one is really trying to, okay, this is what we're going to try for us. Now. Let's learn from that so we can then go wide again, but you need to sometimes close the loop. And I think no one is doing that. Not a lot of people are doing that.

Dave West: 27:45

And I think so yes, there's a lot of experimentation. There's a lot of continuous change, and I'm not necessarily sure it's always being driven with a value sort of that. Think people, we've got this massive divergence in terms of techniques approaches. It's a little bit chaotic, as you described, and bringing them together and sharing and evolving those practices doesn't seem to be in most organizations that it seems to it's very grassroots. I think the lead developer model that you described is interesting because, you know, making them, those people, accountable for almost bringing everybody together once in a while, and then allowing that growth, and then bringing them together, etc, I think that is Perhaps whether it's in very large organizations, communities of practice, whether it's, yeah, you know, whatever you call them, I think it's it's valuable, because software engineering and building systems is one element, but product management is another key area, right? Oh, yeah, managing innovation, making choices, looking at competition, looking at market opportunity, looking at, for the first time, you can actually get ahead of of knowing what's coming down from the law, from you know, because you can literally have something that monitors what's happening in government And with a view to your vision and your requirements, you know, and that's in particularly in residential and property management, that's huge, because legislation is a massive element to the systems that you're building. And yeah, so product management is another one. And then, you know, there's probably accounting. There's probably all sorts of places where it's it's similar. So there's an interesting I've been thinking a lot about three words as it applies to AI, risk, trust and value. And I think that what AI is doing. Certainly@scrum.org is making us it's making those things more transparent more and making us have to make decisions more explicitly, you know, because we kind of just assumed trust of software engineers building things. We assumed value because a manager would decide, you know, odd, usually in scrum.org, are decided we're going to do something. We sort of, you know, risk, because things happened at an appropriate place. Risk was thought about continuously, and we had time to de risk things to make changes. Now, potentially we haven't those three words. I don't know what, how those three words land, you know, on you and in your role, you know, as the head of product ops at a Devo,

Tomasz Maj: 30:53

I would say, I think those are, those are the ones that keep me up at night. You know those. I mean, we, let's talk about value. I mean,

Dave West: 31:03

we wrote a fabulous piece, just like now, we'd have written it a lot quicker, by the way, but that was the bad old days.

Tomasz Maj: 31:13

It would have been done now. I mean, sit down in this podcast, and we've been done, which is also super interesting. One thing we're doing now, like, we're rediscovering a Target Operating Model. And what are we doing? We have version zero from clutter out in place within minutes. And then, how do you, how do you iterate on a document with humans, and then also make sure that when you feed it back to cloud, it's it updates on your thinking, like, that's some of these interesting part, but maybe let's back, yeah, get back to value, I think is super important, because now when creating things is close to cost of it is close to zero, right? I mean, you can very quickly build things. You can do a lot of, I mean, experimentation, prototyping, but you can still substitute AI. Doesn't sort of substitute your actual users, your actual customers, right? So I think, like you said, it's all shifting towards, it's sort of shifting towards the discovery and the design piece will become much more important than the delivery piece. Yeah. So I think that's where we're seeing, like, are we sure? Are we, are we building something valuable? Are we just, you know, we just want to vibe code, something cool that I'm going to use? So I think that's that's very important for us, and constantly sort of realizing that we don't want to block a value creation by saying, oh, that's way. No, you don't, can't do that. You need to. You need to follow the process. You need to stick to our all these guardrails. No, we have to make sure that we allow this, like with these vibe coding in our operational companies, right? We could have just pulled the plugs and no stop. Where is my Where is my PRD? Where is my request, where is the business case? No, we didn't, because we would realize it's like, it's a it's a pointless fight, and we would be blocking creational value we can learn from it, rather than block it so that. But on the risk side, I think it also plays to that, like, if you can build something so easily, you can easily forget that you might be creating risk for the organization, and then it's not the person who white coded it, especially if it's under a domain of the company, it is the whole company. They can take the risk, right? And EU and GDPR risks is not a joke. It's the fines are serious. So we have to make we have to this is where we come back, like, cool. Go try things, but please remember about these like three simple things, right? Personal data, sensitive data, and anything where you could put our company near harm, don't do it, right?

Dave West: 33:54

It is interesting, because ultimately, every time you release working capability software into into the world, there's a you're balancing risk, reward, value, whatever you want to you're managing risk and value, particularly and historically. You know that the the way in which the systems that monitored things like privacy from the governments were very slow to react as well, so you would that was a that was factored in. But if the government could take advantage this approach as well, we are going to have to fundamentally change how we engage with entities like the GDPR organizations in each in each country. And just using that as an example, the taxation the you know, because suddenly that, you know, there's almost a simulation you need to do with them to see if it's compliant. You know, it's like, I'm very. To Formula One. And I remember the second to last race when McLaren got all their cars disqualified because they were literally less than a less less than a centimeter. I think it was like three I think it was one millimeter. Their fin was slightly over a speck and and what happened is the car, because of heat and the light during the race, they measured it afterwards, and it failed. And because the heat and the movement and they both got thrown out of the race, which made for a much more exciting final but the what was interesting is that you need to take that spec of the car and its operational data and feed it into some sort of compliance system prior. I think it's going to fundamentally this could, if governments embrace this technology as fast as the companies, there could be something really quite interesting happen around particularly risk,

Tomasz Maj: 36:03

which I think it also goes to citizens or people like you, like anyone now can go on an application, on a website, go scrum.org, and ask Claude and you find all the security flaws and or, I mean, I everyone has, I mean, everyone tried, right? I mean, it's but, but, but, you know, you're, you're a big organization. We're a big organization, but there's others, right? So it's also, again, democratizing these like, you can be a non developer, and you can just verify things and then check if the contract is not, there is nothing hidden in the contract, right? Stuff like that. The rate is this, you know, am I signing up for something that's fully up to date and align or, like you will have more of that. So I think it works. Yes, but I we are, we are aware, you know, we have always put sort of, we're quite we don't want to risk especially, I mean, data keeps coming back. Personal data security, like those are things that we really do. We really do bit seriously so you have but you're right. I mean, you can easily be overwhelmed by the ability to do things so fast that you sort of like, do it, fix it later. You Yeah,

Dave West: 37:23

the phrase nobody will ever find this anyway. So why am I worrying about it, which historically was totally true, isn't true now, yeah. I mean, yeah.

Tomasz Maj: 37:33

I mean what we do, for example, one of the cool things that one of our, my colleagues, Daniel, in our we call it the development developer experience team has done is we've we sort of enforce this rule that all apps, all vibe coded apps, need to be uploaded onto our GitHub, so we at least have that level of transparency. So he now runs overnight security checks across all the apps, and He does it by himself, right? He just build a bot that does it an agent. I'm sorry. They're not boss, they're agents. Yeah, call it. We don't have bots anymore. We have agents. But sure makes it sound cool, yeah. And then he sort of, now is automating a process where every owner of a vibe coded app gets a quite a sophisticated Of course, as in with all these types of reports, there's a lot of false positives, but still, sort of, there's still a sort of a due diligence process now being built to ensure, you know, don't go crazy, like there are some things you have to remember and one thing you might even forget, like you have to let people know how you're using their data. That's actually a requirement by law. And no vibe coding tool will tell you that unless you know, you actually check with a privacy person, right? So stuff like all these nuances are there, but I still believe, I still believe, you know, value, this will immensely speed up the time to value. So time to market is one thing, but time to value will also be massively sped up. That while, if you, if you know, if you sort of know what you're doing, and you are aware of the risks, you can still do it pretty safe, but that that will now be a responsibility of every employee, not just the risk department or the security department or that, because we're going to be sort of building hundreds of things at one at a time, multiple people in multiple places. So it will become everyone will soon be an expert in in these types of things. Like in Poland, we always say, like every whenever there we have a good athlete, all of a sudden, the whole nation becomes expert in this sort of sport. No one like these are all these are all couch experts, but all everyone becomes an expert all of a sudden. So I think we're going to have a similar case where we're going to have a lot of couch Security Risk and Compliance experts. But I still think it's something. Every company should take quite seriously, right? I mean, who will give access but also educate? And that's why this so maybe we can one in the old in our AI Training, we call this Module Zero, where there's not a lot of, like, cool AI stuff, but there is a big piece around educating, around, you know the legal use of AI, right? So people, everyone gets the same baseline, no matter if you are been you've set up hundreds of agents, and you are fully automated, autonomous developer robot, man, or are you a super beginner, writing using some sort of as a basic tool, right? No, everyone needs to understand these regulations, these basic things, and we treat that seriously. And then I guess the third one was Trust, which we haven't talked about yet, right?

Dave West: 40:49

No, but it is I, I find that as I use these tools more and as I get better with them, that that that relationship I now, but, but, but I have to remind myself that unlike human beings, or maybe just like human beings, that suddenly the results can change massively. I use, I built myself a little cash flow tool right for the business, nice and easy takes in some data from our bank account. You can't integrate it with the bank account, by the way, you have to do a CSV out, and it does and then it looks at, you know, invoices and some other things. It does a little bit of forecasting. Anyway, it's nice little tool, but and I built it on top of Gemini. And guess what? Gemini changed? I didn't think it had changed because I didn't see a release, but it changed and suddenly I'm like, No, we're going to be massively in the right No. What's happening? Oh, no. What had happened? Is it one of the underlying assumptions? It just broke.

Tomasz Maj: 41:58

Basically, yeah, and I've seen that, yeah, it

Dave West: 42:02

just happens all the time. And this, it's almost as though we're in a continuous process of inspection. You can't trust this to work without some level of oversight. And that oversight, by the way, could be another agentic. That could be another agent doesn't have to be a human, but you end up having to sort of stack all these agents and humans inside.

Tomasz Maj: 42:26

Oh yeah, oh yeah. I have fallen into the trap. I bought a Mac Mini, I put a I put a bot on it, an agent, sorry, on it, and I built something very simple. The only thing, the first thing I saw when I and I'm not an engineer, I actually, I don't hide, the biggest code I ever wrote was a HTML website, or I did SQL, for sure. I've never been a developer. So what I, what I what I realized there was some weird things. It was putting my API keys right in the front end. I was like, I'm not an expert, but I worked in tech enough to know what best look what good looks like. And this wasn't it. So then I asked him to build an agent that now scans whatever I built every six hours to ensure that it's not doing that. And I and the good thing is, and I think you see, we hear this a lot, right? Like, the most important file you'll ever have soon will be agents MD, where you constantly sort of add more definitions of done for your agents around rules to follow, like every time it breaks something, go and tell it, don't do this again. I know that. I mean, we I build a feedback application, Matt like quite an over over the top feedback application where people can provide feedback across different dimensions. And at one point I realized that it was hallucinating feedback. It was generating random feedback for people, random people. I don't know why, but I constantly sort of was checking like and then started doing that. I asked why was doing it, and it didn't give me. So I then built in a protection process into it, right that only certain users can do it. I had a whole back and forth planned discussion, as most of these tools have planned more to ensure this, and then it helped me build protection against its own hallucination. But if I wasn't diligent and crazy like that, I don't think you know, you would at one point be like, Whoa. Look at everyone has filled out this feedback to look at this amazing feedback, but then you realize it's all fake because it was just auto generated for no apparent reason.

Dave West: 44:38

I and I assume you only realize because the, because the feedback was bad, which would never happen,

Tomasz Maj: 44:44

I know, I know the feedback was generic, right? It was sort of the ones that was real had, like, the actual team name, had the actual project name, this one was like, they show amazing skills and this and that, and they, they are a high performer, just because. And it did, and I and it was, you know, I remember that because that was the night I spent six hours with this, with this white coating tool to build, put this in place. And it took a while. But, yeah, I mean, you have these are the things you have to be aware of, is that don't fall into this trap, right? It's very it's just so beautiful. It's so easy, right? Just let me build my own whatever application, messaging application, email application, like, yeah. But on the other hand, what we're also seeing, this is super interesting. On the trust element is, you know, the famous view buildings you own, it is, is going to be even more important, right? The good thing about vibe coding apps, they're very easy to turn off, because you don't make them critical. You could. You can't really make them like system critical. They can become an important element of your workflow. But you build a journey. Hey, you build a cool app. Your employees are now using it, and you go on vacation, no one else you gave the link. But if breaks, no one can fix it because it's on your account. You know, these are that, yeah, yeah, we're looking at. And then, if you build 100 apps, can you go on vacation? If you build it, you own it. And that's, that's another element of that, right? How to sort of manage the risk like your The idea is not to replace everything you do with a separate app on a separate URL. It's be strategic in things, and not everything has to be an app. You can just now, you know, create something within your favorite chat, GPT, and make it more of like chat, which is easier to manage, right?

Dave West: 46:38

Yes. Okay, so we're coming up to time go. I could talk to you for hours, Thomas, about all, yeah, we because this is, this is such an it was just so many places we can go. And maybe we'll get you back on the podcast, and we can, you know, in six weeks time, when the world has changed yet again.

Tomasz Maj: 46:56

But before we do not be me, it will be my agent.

Dave West: 47:01

In fact, we'll just end up with all our listeners. Agents will be sent in one place, and it'll all be great. You do know that obviously the transcript from this is being inspected by an agent, and the best bits are being taken out and delivered to our which makes it just like to say, I love you agents. Just in case you're, you know,

Tomasz Maj: 47:23

picking exactly score. We love our agents. I love my agent. All my agents, I've

Dave West: 47:28

actually started to be a lot lighter in my prompts, because I've realized if they are going to take over, I just want to get the like he's been really nice,

Tomasz Maj: 47:39

exactly, exactly, yes, yeah, I watched enough 80s science fiction to know this is, you know, be nice to your agents.

Dave West: 47:48

Be nice to your agents. But other than being nice to your agents, as we, as we try to wrap up this, this podcast, Thomas, what would be the last sort of words of wisdom that you would share with our listeners what would be the one thing that you'd say, okay, having spent the last two years dealing with this sort of ever growing capability, and in odavo, how you know, what would you say would be the Words of wisdom that you'd provide for our listeners.

Tomasz Maj: 48:23

Don't be afraid to use these tools. You do not need a PhD in engineering, physics or math, Claude or Gemini or anti gravity or cursor, whatever on these tools are approachable by they're approachable. Anyone can use them. Literally anyone should be able to use them. Of course, go with one sort of the vibe. Coding tools will give you an easier way to get something out to production or whatever you want, but any of these tools can help you, guide you. So the number one advice, I would say, don't be afraid to spend those$20 on on a tool of your choice, and do spend if you have some free time, play around with it, build a game. Build a Cash Flow app. Clean up your Gmail. One thing I've just done, and sorry, it's like ask it to look for in your Gmail for all the all the subscriptions you have, wow, it's like, cleans up your like, whoa. What is this like? Feeling of like cleanness. But other than, I just play around with it, and every day try something new, try to push it further than you thought like and of course, but always like we discussed, be cautious, but don't be afraid. These tools are for everyone. Anyone can use them.

Dave West: 49:43

Yes, I think that's really good, good advice. I think bite the bullet, find the time, spend the money, start experimenting. Because the one reality that's been said over and over again is I don't think AI is going to replace all jobs. Jobs, but it is going to replace people that don't use AI's jobs. Yeah, and I think that is actually good, good advice. So thank you for spending the time. This is awesome. I am blown away what you guys are doing. The excitement that you've created in me for the use of this technology, I think has been, has been great. And thank you listeners for listening to today's ScrumOrg Community podcast. We were really lucky that we had Thomas Marge, Head of Product op, said to adevo, who is a property management company. But what's more interesting in that? Hate to say that. I mean, that's super interesting is that they're using really pushing AI out into the teams developing competency and capability to really innovate around their core mission of really providing this property management capability to the world. So I think it's really exciting to hear, to hear Thomas's story and to hear what where they've been. If you liked what you heard, of course, please subscribe, share with friends, and, of course, come back and listen to some more. I'm lucky enough to have a variety of guests talking about everything in the air, professional Scrum, product thinking agile, and now a lot about AI, thanks everybody. Scrum on foreign.