Why We Don't Follow the Trends - Part 1

OLIVIER COLSON: 0:00

At some point, we start talking about AI and integrating that with everything, right? You could put it everywhere."

DAMIEN BOUVY: 0:07

Actually, if you put an AI that says dumb stuff says. That's crazy. There is also the problem when you try to do things too simple. Especially using AI, for example, is that you have no idea what it does. You kind of know it works, but the user has no way to intuit how it worked, why it did what it did. Like, if you start to automate stuff and try to make things intelligent, then people have no control over what happens in their ERP, which is kind of like a really important piece of software in a company. And so we avoid magic stuff most of the time when we can.

OLIVIER COLSON: 0:43

So basically the question is, you said that cryptocurrencies, or blockchain into Odoo. Could you tell us why?

DAMIEN BOUVY: 0:51

The fact that we say no is simply that there is The guy who paid with bitcoin for his pizzas in 2009. He must be really pissed off by now because I think he paid 10,000 bitcoins for a pizza, and that's like 15 million now or whatever. So you don't pay with Bitcoin, really?

OLIVIER COLSON: 1:07

Hi, everyone, and welcome to this new episode of Today, we decided to offer you the rebroadcasting of a live from our Twitch channel. This time, you'll be able to hear the first part of our conversation. Ready? Let's go! Hi, and welcome to this new episode of Odoo Unplugged. So today, we're going to talk about trends in development. We're going to talk about, uh, you know, those those cool things that everybody talks about. And should we, should we not do them? And what did we do with them at Odoo? I have two guests with me. So I guess you guys can introduce yourself, actually. So, Mehdi, you go first. All right.

MEHDI RACHICO: 1:43

Um. Hello everyone. I'm Mehdi, I'm a software engineer at the professional services department in Odoo. Uh, what we do usually is, uh, create customizations for clients that are not fully satisfied with the standard Odoo. So let's say, for example, we rework some of the accounting flow, and some of the sales flow. And, uh, yeah, we we usually encounter API integrations in our everyday work, and uh, yeah.

DAMIEN BOUVY: 2:09

And I'm Damien, uh, I'm in the product management Um, I'm in somewhat of a hybrid position because I have a tech background at Odoo, so I tend to manage more technical aspects of the product.

OLIVIER COLSON: 2:23

You used to be a developer when I started at the

DAMIEN BOUVY: 2:24

I was a developer, uh, and so my role now is to and to say no to a lot of stuff that he then has to implement. Uh, exactly in their team. Uh, so that's it. Okay.

OLIVIER COLSON: 2:37

That's a nice way to.

DAMIEN BOUVY: 2:39

Show.

OLIVIER COLSON: 2:39

Your interaction. Why not? Why not? Uh, so I suggest we start with, like the, the most obvious subject, actually, because, uh, you know, it's been like, what, one year that that media are all over the place talking about one thing, artificial intelligence. So there, there have been all those things with ChatGPT. Uh, I don't know if you guys tried it yourself, but it is truly amazing what that thing is able to do. Right.

DAMIEN BOUVY: 3:03

That's actually a good question. Do you do you use it in your everyday development?

OLIVIER COLSON: 3:07

Uh, that's that's the point.

DAMIEN BOUVY: 3:10

I know that I do use it, uh, even for stuff like I also use GitHub Copilot a lot, and it's I've been using it since the start, from the start and.

OLIVIER COLSON: 3:23

Recall what it is for people. Yeah.

DAMIEN BOUVY: 3:24

Github Copilot is basically an AI that So you start typing five characters of code, and it fills in like two hundred of them.

OLIVIER COLSON: 3:33

And does it, does it really work? I never tried it, but.

DAMIEN BOUVY: 3:36

If you tell it what to do, it's it can be really Uh, I've used it a lot in also in personal projects. Uh, and I was actually quite amazed because Odoo is, uh, you know, we have our own framework in JavaScript and in Python, and it knows them both. Like, I can ask it to do an old component. I can ask it to do like ORM models and it works. And it names things quite well, like the code is readable. It's not perfect, of course. Uh, I think like most AI at the moment, but it really it's really astonishing, to be quite honest. I got used to it, and now it's part of my and.

OLIVIER COLSON: 4:12

Every kind of thing. Were you programming with it? I mean, from a complexity point of view, uh, because.

DAMIEN BOUVY: 4:17

It helped me like recently, I did, uh, put my the automations, uh, of Odoo studio, which is kind of like low level stuff. And it did help me quite a lot. It's. Yeah, like for functional stuff, it works really well for more technical stuff, like if you were to do stuff into the framework itself, it's going to be hit or miss because it doesn't know everything, but still, uh, still, it's really quite good. I would advise anyone to use it if you want to work faster and more efficiently. Crazy.

MEHDI RACHICO: 4:50

I would add something on that exact point Like I also use a copilot from time to time. So if I'm getting started on a project and I want like the structure, the basic structure of the project, I. I just use copilot to make me, uh, to make my, uh, start faster. But I would actually, like, disable the completion from time to time when, like, the the, the flow gets complicated because copilot gets a bit annoying when it suggests something that you didn't even, uh. Uh, yeah.

OLIVIER COLSON: 5:15

And it makes sense, actually, because it knows So all the things really common, it knows them. And so it can give them to you.

DAMIEN BOUVY: 5:24

But I and it was trained on basically all of So it's not surprising that it knows Odoo. But still I would have thought that Odoo would be like buried beneath all the Django stuff that you can find on GitHub or all the, uh, the react stuff. And yet it can make the difference between the two frameworks. It's really, really quite good.

MEHDI RACHICO: 5:43

Also, something interesting. It learns from your style of coding as well. It tries to copy your style. Yeah.

OLIVIER COLSON: 5:48

Yeah. So what, uh, if you pretend to give, like, know if.

MEHDI RACHICO: 5:52

You tend, for example, to have a certain It copies like the same structure when docstrings for functions. Yeah. Crazy. Amazing.

OLIVIER COLSON: 6:00

Yeah, crazy. And so the next question is, uh, So maybe not Copilot, but with ChatGPT, or I like that I have a story about that actually, because, uh, last week, that's a true story. So that's very perfect timing, but I'm not bullshitting. Okay. That's a real story. Uh, um, last week I was at the restaurant with friends. So, developers, they were at university with me, and, um. Uh, at some point, we start talking about AI and tell me, "Oh, but as Odoo you probably you must be integrating that with everything, right? You could put it everywhere." And I was like, actually, well, to my knowledge, we put it on, on website, uh, we put it in the well in all the HTML fields. So in the text, you can, you can, you can ask uh gpt to, to, to just give you something. But basically, that's it. And to my knowledge we're not putting it everywhere elsewhere. Right.

DAMIEN BOUVY: 6:56

We have another integration that is very much Uh, and it's more for demo stuff. But if you configure a new website on Odoo, uh, it will ask us, hey, what's your sector of activity? What do you want from your website? Do you want to get more leads or just develop the brand? And in the past, we had some sample data that we would fill into the website to like show what it could look like to avoid the, like, blank page, uh, um, stress of for authors. Um, and now what we do is that we send this data that you put in, like, I have a shoe, uh, store, uh, activity. We send that to a platform that we own that will ask ChatGPT. Hey, can you, like, provide some demo content for someone who's making a shoe store website? And we pull that into the website that we output for you? Obviously, these are just placeholders, but that means that at least when you land on the website that we created for you, you have something that evokes your actual business and not some generic stuff, but it's like it's it's mostly toy, mostly a toy. Yeah, yeah it is.

OLIVIER COLSON: 8:01

Because actually I have another story with that. But uh uh um, so on my free time, I'm part of some, some fan club on some sci-fi series, whatever. Uh, and um, uh, we are redoing the website with, with Odoo. We have already used it before, but a way older version. So, uh, and so I start the thing to generate a new website, and I, he asked me what kind of website I want. So I'm just like, okay, let's, let's try just to see what it gives me. I'm just going to type the name of the series. So really nothing, it's not very informative. Uh, and it's, it's for me it's hard and it's funny because, uh, the result I got what's some, was some kind of, I don't know, an environment friendly website. There's this kind of, you know, very eco friendly thing, uh, but all the things that were said. So it looks like some religion stuff. It was really weird, but, uh, but, uh, everything was, uh, linked to the, the themes of the series. So still it had understood more or less what I was trying to do. Uh, just not that it was about a fan club, but it was funny.

DAMIEN BOUVY: 9:02

Actually. What I think, if I remember correctly, this content and then we ask ChatGPT to review the content generated, and we ask it something like, like is it okay, should should we use it. Do you think that what you output it is good enough. And then sometimes it will say no and then we fall back on the like stupid placeholders that we had before or.

OLIVIER COLSON: 9:22

Philosophical point of view. It's a bit.

DAMIEN BOUVY: 9:23

Crazy. Yeah, it's kind of funny to ask. An AI like, can you read what you just said and does it make sense? And then it will say, yeah, no, I'm sorry, I just bullshitted you just use the old data. It's gonna be better.

OLIVIER COLSON: 9:34

Yeah. No. Nice. So, uh, outside of these integrations, are they? So. So all these are already there in Odoo 17. Uh, are there things coming for Odoo 18? Do we plan to put it everywhere, like my friend suggested?

DAMIEN BOUVY: 9:47

No, I don't I don't think so. We had a discussion I don't remember when it was exactly a few months ago when we saw that AI was really not just a toy anymore, but was getting integrated into Windows, into GitHub, into like really professional stuff where you think, okay, it seems that we have reached a point where it can be really useful. We had a meeting with other POs and the management to decide, hey, what should we do? We had, of course, a ton of ideas. I remember that there was stuff like, hey, we could, uh, ask it to review some kind of analytics and suggest stuff we could ask it to like review resumes that people upload to apply for positions, which personally, I, I kind of hate the idea of automating this step, but I agree with you. That's a personal point of view.

OLIVIER COLSON: 10:30

But hey. Uh, on the other hand, uh, some some Uh, some structures might want these kinds of things. So that's also it's not because you leave the option to do something like that that you, you need to want to, to, to, to wish to make it, uh, exactly unavoidable for everyone. But still, I agree with you. Yeah.

DAMIEN BOUVY: 10:48

We I think we surfaced ideas like, you know, use So like, ChatGPT could review uploaded content by users and say, hey, this should be moderated by by human. Typically, I don't remember all the use cases we had in mind, but we had a lot. Uh, of course, the Odoo way is usually to start small, uh, and with the most useful stuff. And the strength of, uh, of these AI is at the moment is generating content. And so it was kind of abuse to put it in the website, because that's where you will use a lot of content and use and generate a lot of content. Um, but yeah, I. I'm not completely up to date on what is going on inside of R&D in other departments with regard to AI, so I'm not sure if we're going to, uh, already implement other stuff. Uh, there is also the question of how much it will cost us, because at the moment this integration is free, even in the community edition of Odoo. Uh, and at some point, maybe it will start costing us some actual real money. Um, but yeah. And what I remember that what surprised me at this meeting a few months back is that, uh, you know, in technology, things kind of move fast. New trends emerge all the time, and we often get asked immediately, uh, in the product department, hey, should we add this new technology to what we do? And the usual answer we give is, yeah, you know, let's wait a bit. Let's see if it's really useful. Let's see the use cases. Um, for example, uh, if you if you compare the speed at which we integrated AI into Odoo compared to the request that we've been having for years, which was integrating with blockchain or bitcoin, to which we always said no, because the use case is not really clear. Um, I mean, you can see the disdain for Bitcoin.

OLIVIER COLSON: 12:39

As I speak. I think let people in the chat judge Um, and so yeah.

DAMIEN BOUVY: 12:45

It was really surprising that I mean, basically, maybe two. Uh, and we we did it in a year, editing it to to Odoo. Um, actually, we did it in like a month, but it was a year after, uh, that it got really, uh, released and, uh, yeah, it was kind of surprising. I think it's probably, the speediest integration I've seen of a new technology in Odoo, which is kind of interesting in itself, and maybe but.

OLIVIER COLSON: 13:16

It's, it's, I think what makes it what made it Well, from a purely technical point of view, it's not. Well, we're calling ChatGPT. We're asking it to generate something and we're getting the result. So there is no big logic on our side going on. Uh, and uh, but I think it's it's working really nice and it's doing the job. Uh, and so that's, that's good. But I agree with you that, uh, you know, when there is something trendy like that, the reason why my friends were talking about that is because they would like to put that in their company, in the, the kind of software they work on. I'm not giving too much information away. Uh, and uh, and, and why do they want to put it there? Because they want to play with ChatGPT. Because they want to play with, uh, with the new thing. I mean, it's. Yeah. Come on. It looks cool. Yeah. Uh, so, so they want to do something with it because, uh, as a developer, they find it interesting. And I think that is also something that must be a factor in people wanting to put it somewhere. It's just that they want to play with it. Uh, and, and it is also our role as, like the editor to just say, okay, you want to play with it, but what's what for? Yeah. What for? Because frankly, yeah.

MEHDI RACHICO: 14:26

And yeah, we also use AI for the OCR module. We have an Odoo.

DAMIEN BOUVY: 14:31

Uh, we do use AI. It's not uh, it's not a LLM like ChatGPT. Uh, and we've been doing that for a few years now, like five years, I guess. Yeah. Uh, but it's vision. Vision AI and yeah, we maybe we have other use cases like this that I kind of flew under the radar, but, uh.

MEHDI RACHICO: 14:48

Yeah. And, yeah, we use a deep learning model, code. Yeah.

DAMIEN BOUVY: 14:55

So I don't think we train the models for. No, we don't.

MEHDI RACHICO: 14:57

Train the models. We just, uh, use it.

DAMIEN BOUVY: 15:00

Okay. Yeah. Because I remember that we have, that hey, you missed like this VAT field, which is in a weird place in this invoice, but I don't know if it basically just means that we save a template somewhere that the, the, the AI reuses, or if we retrain it, I don't think we retrain it. I mean, you probably use it.

MEHDI RACHICO: 15:20

Yeah, we integrate it into our code.

DAMIEN BOUVY: 15:23

Yeah. Now that you say that, I'm thinking, okay, don't remember it? I know that we have.

OLIVIER COLSON: 15:30

Bunch of places, but, uh, the I think the message speaking, uh, is, uh, putting AI for the sake of putting AI doesn't really make sense. It needs to it needs to be interesting for the, the project. Uh, otherwise I think we will just end up with, uh, people adding, you know, that that little thing that was on old versions of Microsoft Words in the late 90s, 2000, there are little Clippy, uh, asking you, hello, can I help you? And basically asking you to chat with it? Uh, so we could just reintroduce something like that. And frankly, we do. We want that.

DAMIEN BOUVY: 16:05

Um, it's funny that you speak of, like, uh, small the meeting we had, we discussed training an AI on our documentation and putting it on the website, you know, to answer questions on live chat, stuff like this. And I think, uh, no, actually, I know, I know that a partner did it. Um, I think.

OLIVIER COLSON: 16:27

For documentation, it might make a bit more Uh, the thing is, having that within the software directly when you're using it is a bit noisy. And it's not that it's entirely useless, but that that Clippy, uh, was using resources and which was just. What the hell? What the hell? What do I need that while I'm typing my text? I don't think so.

DAMIEN BOUVY: 16:48

And usually they tend to complicate stuff. I think they tend to suggest stuff that you don't actually need. They also, I mean, we if you know, I a little bit, you know, that they tend to hallucinate a lot and to create engineer. Yeah. Or create stuff that like if you ask uh, and I actually I did try it with custom GPT on on, on OpenAI. You can give it a data set to kind of it won't be trained on it, but it can access it to reply to you. And so I gave it all the Odoo documentation and I asked it questions and it got really good answers if you ask the right question. But if you ask a is it possible to do that, it will tend to always say, yes, it really wants to help you. You know, even. It will Say yes, yes, yes, sure. Do this, do this. And it speak of menus that don't exist or. And obviously it's not something that. We can really put into production. And I think there was actually a story recently in the media about this. I think it was a, an air company that deployed, uh, like an AI as a helpdesk agent on the, on the website. And the AI promised stuff to customers like, yeah, we can refund you, of course. And actually, no, it was not true. And so they had to go to, uh, to do a trial to, to determine that. Actually, if you put an AI that says dumb stuff on your website, you're responsible for what it says. I thought, that's crazy, but it's a good.

OLIVIER COLSON: 18:08

Point that that's also one of the of the limits. People forget maybe a bit about, uh, if you hire someone to answer the chat. Uh. You train this person, and you tell this person, okay, you can say that. You can say you cannot say that. And here are the correct things. If you deploy some AI and even if you train it with your own data, uh, you need to test it really, really well because you it's so complex. Actually, that's also the beauty of it. Uh, it's so complex that you lose a bit control on, on the actual output. It will give like 100% of the time, uh, because even if it tells right things, 90% of 90 or 95% of the times, if you have a lot of traffic on your website and the five persons, uh, it's just telling crazy bullshit or, you know, or racist jokes or whatever, but you'll get into trouble.

MEHDI RACHICO: 18:55

Exactly. And yeah, I mean, yes, that's the point. Like, as you said, like this. Like using AI just for AI is not good because as we see, like in Odoo, like we only use something if it's really needed for our, for our like use case. Let's say for OCR, it's really important that customers automate, uh, reading of invoices and put them into Odoo directly. So, and also this gets me back also to uh, to a previous company I worked for. So, uh, we were trying to predict, um, we were trying to predict, uh, uh, food production for companies to help them eliminate food waste. So basically, we have, like, a lot of time series data where we have like, for example, how much you sold in this week, how much you sold the next week, and you would have this data for 10 to 5 years, and you use like train this, these machine learning and even deep learning models to do this. And like we spend a lot of time training like the most complex deep learning models with transformers and everything. And the best results we got with with naive, uh, statistical models and for some of our clients. And we found ourselves wasting a lot of time doing something really over complex for what we do. And yeah, so this is, uh.

DAMIEN BOUVY: 20:00

That's kind of the problem. When you get a new

MEHDI RACHICO: 20:03

Exactly. Yeah, exactly. Yeah.

DAMIEN BOUVY: 20:05

And we actually, I think we had this discussion, when we started doing the predictive scoring of leads in CRM. Uh, there were some voices that were suggesting, hey, we could actually use AI to do stuff. And then we tried basically a bias theorem. Yeah. And it works super well. And you're like, okay, why make things complicated when you can just look at what happened in the past, take a few guesses exactly that are not bad but not overengineered, and then work with that. And it actually it works quite well actually.

OLIVIER COLSON: 20:41

Yeah. And I think it's, it's really important Uh, because a lot of the value of Odoo is that it is simple, or at least it looks simple for the user. That's, that's the main thing really. And, and and you know it as a product owner, there are a lot of things that you say no to all the times. Uh, why not? Just because you want to be rude with people, I hope. But but but. Just because, you know, uh, it would make the software more complex and, and we, we really try to, to hunt down all these opportunities, you know, uh, if we can choose not to put a button somewhere and have something, uh, a bit more intelligent than that, than that we do that all the time. And so clearly, when it comes to integrating something like that, we want to have something simple, something clear for the user, and not something you need to configure everywhere that will give tons of different outputs.

DAMIEN BOUVY: 21:28

Uh, there is also the problem when you try to do example, is that you have no idea what it does and I don't know if you like. In my code, I don't like black magic stuff like you kind of know it works, but the user has no way to intuit how it worked, why it did what it did. Like, if you start to automate stuff and try to make things intelligent, then people have no control over what happens in their ERP, which is kind of like a really important piece of software in a company. And so we avoid magic stuff most of the time when we can. So, an AI, in a lot of cases, falls into the magic stuff category.

MEHDI RACHICO: 22:09

Yeah, exactly.

OLIVIER COLSON: 22:10

And when there is magic, it needs to be logical So I have a few places in mind with accounting. Uh, where, where uh, some kind of configuration will trigger some, some behaviors that are not explicitly configured by the user. But, you know, even if, even if someone does not understand and, and ask the question to the support or whatever, which is okay, or looks at the documentation, uh, once explained, it seems logic, it seems it's normal that we do that. Uh, it's just we hide it just because we know that we will need that. Uh, so that makes sense. Um, we have a question in the chat, uh, about what? What you were saying about cryptocurrencies. Uh, earlier. You just opened Pandora's box. I'm sorry. Uh, so. I'm just going to display it, uh, up. So, uh, basically the question is, uh, you said that you've always said no to integrating cryptocurrencies, blockchain into Odoo. Could you tell us why, uh, and how do you choose what to add to a do and what not then?

DAMIEN BOUVY: 23:04

Uh, it's a loaded question. Of course, the for cryptocurrency. Uh, I can reply super easily. Uh, we are an opinionated software company, in a way. I mean, obviously, as you said, we try to, even if we don't like something for. We have modules to to track employees when they log in, log out where they are. We don't use it internally because that's not our point of view in most of our interactions with employees. But we did it anyway, even if we don't really like it. We know that people want it. In the case of cryptocurrency, a really small, small amount of people want it, and they believe in that. Uh, I'm going to call it business model, even though it pains me to do so, and they really hardly believe in it. But that's a really small fraction of the actual user population of Odoo. Um, and we had feedback even on that, I think, two weeks ago where the person told me, yeah, we, the use case of bitcoins, are now really coming into into the age of maturity. And I kind of laughed, to be honest, because I think the blockchain, uh, exists, what, since 2008. And besides the obvious usage on cryptocurrencies and ETFs and, uh, ETFs, uh, uh, you know, like the images, uh, I don't remember the NFT exactly. Um, which itself was dubious as well. I mean, we've all seen this bubble, uh, rise and explode.

MEHDI RACHICO: 24:33

Yeah, that was crazy. Obviously, there is.

DAMIEN BOUVY: 24:36

No actual, truly, uh, widely accepted use case And I will even reiterate some stuff that I see on Twitter come and go. Is that basically, blockchain is a linked list with a hash that stores the integrity of the chain. That's it. There's there's nothing really super complex there. So the fact that we say no is simply that there is no good use case. And if you say, ah, yeah, but in e-commerce, I want to accept bitcoin payments, that's not true. It's like saying, okay, I'm going to pay with, like uh, shares from Apple. It's not it's no longer a currency. It's like it's a trading stuff. It's a I'm not going to say pyramid scheme because I know it's loaded, and I know that, uh, people in the chat might get angry, but that's kind of my belief. Uh, and, uh, I mean, the guy who paid with bitcoin, uh, in for his pizzas in 2009, he must be really pissed off by now because I think he paid 10,000 bitcoins for a pizza, and that's like 15 million now or whatever. So you don't pay with Bitcoin. Really? And even like, I know that there was also a lot of discussion that yeah, Bitcoin is private. Uh, that's actually not true. It has been proven mathematically that the ledger is basically in the open. So if your address leaks at some point all your past history can be revealed. So that's not even a good way to use money. And to transfer it, you lose all the privacy of a banking system, of all the security of a banking system. So no, we we really don't see a use case for cryptocurrency at all and for blockchain itself without the shady currency part. We kind of have something that looks like a blockchain in the French, um. Accounting. Accounting. Because you have to like, uh, make sure that there is no modification in the past. So we did in fact create like a linked list with a hash that verifies that the integrity of the list. And that's it.

OLIVIER COLSON: 26:30

Because of a legal need, because.

DAMIEN BOUVY: 26:32

There was a legal need. Indeed. And it's not a bad one. Uh, sure. Sure. It's not really blockchain. It's blockchain.

OLIVIER COLSON: 26:37

Like it for the sake of doing it. It's really because yeah, they wanted something. Something to ensure that you couldn't you couldn't change everything everywhere. And so we had to do something.

DAMIEN BOUVY: 26:46

But to be honest, I have yet to see a real, Okay. If you get one, send it to me on Twitter or whatever.

OLIVIER COLSON: 26:55

Sounds like a clear answer. Um, okay. Uh, to go back on the on the way, we, we, uh, decide what to put or into Odoo or what we, what we should not put into Odoo. Uh, we have another question in the chat that I'm up displaying now. Uh, so how do you assess the impact of any trendy features on Odoo user Odoo user-friendliness? Uh, do you discuss with other users about that?

DAMIEN BOUVY: 27:23

It's a complicated question. We do, uh, quite, the sense that most UX designers would probably look down on my job and say, how are you deciding what to put and what to put? Not in the software. We don't do a B testing. We don't do extensive user research. Um, usually, we simply ask ourselves, what is the complexity of the development? In the case of ChatGPT, for example, as you said, it was super simple. You have basically one endpoint. You ask a question to the system, and it replies to you. So it's really easy to integrate. Um, in the case of blockchain, it's a nightmare because you have to handle a lot of stuff in screen asynchronously, um, stuff like this. And so we check out the complexity of the development. We check at the value we think it will bring to, to the users.

And then we have I'm going to say it: 28:14

I know that a lot of people will hate it. But we have the metric of whether is it good from a common sense point of view to do this. And we kind of of course, there is like variability between product owners, even within the team. We have debates, uh, where something someone will say, hey, we should do this. Then I would say to complex, someone say, yeah, but the value is enormous.

OLIVIER COLSON: 28:38

I think we also work a lot with, you know, we customer tickets, from our own experience as well, because Odoo is using Odoo. What a surprise.

DAMIEN BOUVY: 28:48

We do a lot of dogfooding.

OLIVIER COLSON: 28:49

I mean, for, for for accounting, for example, I feedback coming from our own accounting about things they want to do because yeah, I go do the accounting is quite complex. Uh, and so sometimes we realize that there is something to do or to change because of some this kind of feedback. So, uh, indeed, we don't have any real strict way of doing things like always, actually. Isn't it like the DNA of Odoo? Uh, but, uh, I think we stay very dynamic. Uh, and, and that's what makes it work, actually. And that will be the final word for this first part of the episode. I will now make a short break, and we'll be back in a few minutes. And that's a wrap for today's episode. Did this first part interest you? To listen to the second part, be there next week as we post the rest of the conversation in which we answer some of our viewers' questions. If you want to join us live next time, don't hesitate to follow us on Twitch to receive the notification of our next programs. The link is in the description, and if you're in the mood for more captivating content, I highly recommend checking out our other episodes. Until next time. Cheers!