How Huge ERP System CEOs Think About AI

Shamil Malachiyev: Good morning, everyone, and welcome to Beyond the Signature, where we cut through the AI hype and count what's really happening with automations and AI agents across the industry, straight from the people in the trenches doing the work. Where is the adoption really at? What tools are being used? What are the real challenges? Forget about those shiny LinkedIn posts, because this is AI reality. Today I'm joined by Domenico Cipolla, co-CEO of Xentral, a cloud ERP platform out of Germany that powers operations for over 1,800 companies across Europe. They're backed by Sequoia and Tiger Global, and right now they're in the middle of building AI agents directly into their ERP. Domenico, welcome to the show.

Domenico Cipolla: Hi, Shamil. Thank you very much for having me.

Shamil Malachiyev: Let's start with what you currently do at Xentral, the scale at which you operate, and the real AI challenges you're facing at the moment.

Domenico Cipolla: Sure. I joined Xentral almost to the day three years ago. Benedikt and Claudia, the two founders, are still in the company. They built Xentral initially for their own purposes, and they built it over years and years, bootstrapping it to a considerable size. But at some point they said, running a company with 200 people is very different from running a company with 20 people. We need help. I had been on the other side of the table, on the customer side, for 15, 20 years, so I have lots of experience operating ERPs, installing ERPs, implementing ERPs. I think it was a marriage made in heaven. They asked me, I joined, and for the last three years we've been rebuilding and repositioning Xentral from the very good beginnings it had into this new AI era. I think we're very well positioned for everything that is yet to come.

Shamil Malachiyev: With a lot of CEOs joining companies in the last two to three years, there's this question of what the company's AI strategy is going to look like. And it began at a stage where we had only just discovered this new technology, GPTs, and nobody really had an idea of where it was going to go, what solutions would be built, what would be used and what wouldn't. How did you approach that beginning, that anxious, scared moment of, okay, something's about to change, what are we going to do with it?

Domenico Cipolla: Maybe we're a particular bunch of people, but this whole holistic grand AI strategy, we didn't have it. AI was such a force, and the momentum with which it arrived didn't leave you a lot of time for strategising and thinking about the future. Everything changed from week to week. I'm currently on vacation, I've been gone for four days, and I'm already scared you're going to ask me about something that was posted yesterday and I'm out of the loop. So we approached it very differently. We weren't thinking about the grand strategy. We understood very early what AI could mean for us, and also what AI isn't. Because at the end of the day, within the context of a business operating and business automation platform, we don't believe AI is a value in itself. It allows us to make things faster and save a lot of effort. But an ERP system is quite deterministic.

So it wasn't about a grand strategy. Very early on we did two things. We started experimenting with it, both for our own purposes and for our customers. And the second thing, which is maybe what comes closest to a strategy, is that we were thinking about how best to position and modify our platform to really be able to capitalise on AI. There's this old saying, and I'm not sure what I'm allowed to say on your podcast, but if you've got nonsense coming into a system, you've got nonsense coming out of it, no matter how powerful the system is. So we were really trying to get ourselves and our system ready to fully capitalise on the power of AI.

Shamil Malachiyev: How did you make the evaluations? You have this technology, where do you put it? There are obvious solutions like customer support, but there's also legal, there's recruitment. How did you pick the departments and directions to implement AI?

Domenico Cipolla: I see a lot of, I come across a lot of people that have grand AI strategies and then you ask them, what do you actually do internally? And they do nothing. So the first thing we did was think about our internal use cases, because that's the least painful area to experiment with AI without negatively affecting our customers. The obvious areas were software development, but also our own knowledge management and our internal processes. Very quickly and very early on we were developing solutions internally to automate repeatable processes, make them less error-prone, and use AI there.

Within that context, we discovered that if you have an outdated tech stack, you can have the best AI and you're still not going to be able to fully leverage its power. So we made lots of internal adjustments to our own tech stack. Once we felt comfortable with that, I remember a meeting where we were trying to determine the first use cases we wanted to do for our customers. It's a very easy matrix in the end. How manual, repeatable, repetitive, no-brains-involved is a process, and how many of our customers are using it, how often. Whatever ended up in the top right quadrant, meaning it's really not that intelligent and it's causing a lot of friction or a lot of work for our customers, we prioritised. Think, for instance, of payment reconciliation. You're getting invoices by email, fax, letter, PDF. At the same time you have your incoming payments in the bank account. Matching those doesn't take a lot of brains, but it takes a lot of effort and a lot of time.

Shamil Malachiyev: Yeah.

Domenico Cipolla: So we prioritised those processes and took them one by one. The important thing within the context of an ERP is that we're not talking about AGI. We're talking about AI. At the end of the day we need to be able to do a process a thousand times with the power of AI, and the result that process yields a thousand times needs to be identical every time. The moment one of the thousand yields a different result, we're going to be in trouble and our customers are going to be in trouble, because we're talking about audit trails, system-of-record capabilities and so on. So for us AI is a powerful tool, but it's a continuation of automation and machine learning. AI is the next step in that. I'm not downplaying its role, I think there are lots more things we can do now, but it's an evolution from machine learning: getting processes more automated, more repeatable, safer, more secure and faster.

Shamil Malachiyev: Can you give us more detail? When you saw the opportunity with invoice reconciliation, what tools did you look at to automate the process, and how did you make it more deterministic within your teams, to make sure you don't get that one in a thousand where an invoice goes through when it shouldn't have?

Domenico Cipolla: I wouldn't say we made it more deterministic. We made it faster. Let's start from the artefact. The original process is an accountant checking the bank account and checking the invoices, thousands of invoices and thousands of entries in the bank account, and trying to match them. Typically you have an invoice number, but people mistype it, sometimes they forget it, some people leave it out on purpose. So there's a lot of manual work. There's a heuristic that tries to match, but it's very flawed. Out of a hundred invoices, you still need to manually touch 50, 60 or 70, depending on the quality.

One of the first things you can do with AI is the OCR, the recognition of text in files, translating it into structured information you can match. That's a big step, and those tools have become much, much better with AI. In the past, we've been using a third party tool. Now we've done our own one because with AI, we almost vibe coded it over a weekend. And it's a much more powerful tool than the one that we paid for in the past. That's number one.

Number two, and this is why I qualified AI as an evolution of machine learning: one really important thing is the learning. Machine learning does the same process over and over. With AI we now take into consideration how customers have reacted in the past. We thought this invoice and this entry didn't match, because the data didn't, but the last seven times we suggested that, the customer manually said no, they do match, because of a PNC. So the system learns, and that's what increasingly improves the quality of the output.

The third thing: as you mentioned in your intro, we have more than 1,800 customers. So we have a lot of learnings across customers as well, without jeopardising data protection, and we can generate learnings that help all our customers. So with AI, the structuring of text, the matching to the bank account, the speed at which you can do it, the learning within a customer and the learning across customers are opportunities we didn't have in the past. Or if we had them, they were extremely complicated.

Shamil Malachiyev: When we look online, we see a lot of information about AI adoption among Fortune 500 companies, which have to be cutting edge. When you look at the real market, especially the European market, how do you see AI tooling adoption rates within SMEs compared to what's being advertised?

Domenico Cipolla: I don't know how it is with enterprises. I suspect there are lots of statistics saying 95, 97, 99 percent of projects are failing and don't have an ROI. What I can say about our customers: Xentral is a business automation platform specialised for retail companies. Our ICP starts at two, three, four million euros, where our customers typically hit a wall, a growth wall, not because they lack the opportunity to scale but because they lack the platform, the single source of truth, for taking the next step in terms of workload, resources and complexity. That's where they come to us. Our ICP probably ends at 100 or 150 million. We have loads of customers doing 200, 300 million euros in revenue, but our sweet spot is probably between 20 and 150 million.

They don't have big IT departments with 10 or 15 people constantly fiddling with new things and experimenting. They are technically, or technologically, not particularly sophisticated, I would argue, and hence they really rely on service providers such as ourselves to pass on the benefits of AI. They're not experimenting themselves. I'll qualify that in a second, but they're relying on us. And this reliance combined with limited resources, no time or money for an 18-month AI implementation project, makes them very no-bullshit. They say, we see everybody talking about AI. What can you guys deliver to us?

Shamil Malachiyev: Yeah.

Domenico Cipolla: Let's not have it next Monday, but the week after. They're very fast to adopt, but also very demanding on speed. So in the features we've already deployed, we mentioned the OCR part. We have another one that's super obvious. We had a really powerful analytics platform, but in the past you needed to be quite good at SQL, at coding. Sometimes customers had to commission a partner of ours to write reports for them. Nowadays it's just a language, and if you have a translator from natural language to the coding language, our customers end up doing reports by saying, I'd like to see this, split into these categories, over the last two weeks, please generate it. The system generates a report. They say, that's exactly how I want it, but I'd also like just the articles above a certain threshold.

There are lots of those smaller, very practical tools that really alleviate the day-to-day operations of our customers. Sometimes I ask myself whether they actually know it's AI. I don't think they even realise, or they realise afterwards. Because at the end of the day, they don't care whether it's AI, whether it's 10,000 people sitting somewhere magically doing the work in the background, or some other technology. So I would argue that real adoption of AI in our space is predicated on our customers not having a lot of resources. There's no time for bullshit or wasted money. Combined with the pressure everybody is under. We had COVID, then the Ukraine war, now the Iran war, all the supply chain issues, price bumps everywhere. I've been in the business for quite some time, and I don't remember such an extended period where retail businesses have been under so much pressure.

Shamil Malachiyev: Yeah.

Domenico Cipolla: The combination of limited resources, really tough times and not being technically sophisticated means they rely on us. And I think we're doing a good job at focusing on things that really matter to our customers, and not getting lost in some major AI transformation project that yields nine months of implementation time and no results.

Shamil Malachiyev: In your own experience, when you develop something that's obviously helpful, like allowing people to use natural language to bring analytics into their dashboard, do people naturally start using it? Or do you usually have to message them: hey, this is the new function, please try it, or get in touch and we'll show you how? Because we keep pushing all these new technologies with natural language and MCPs connecting to all these systems, but are we really seeing everybody jump to use them? People had their old ways of building those reports. How easy is it to persuade them?

Domenico Cipolla: Very good question, and there's a lot in it. Forget AI for a second. If we launch a new module, let's say we have 2,000 customers with 20 or 30 users each on average. Take 25 and that's 50,000 users using the system on a regular basis, daily, weekly or monthly. You can have the most superior new module compared to the old one, and getting people to switch from one to the next is not done in a second, because people are just used to doing things a certain way. That change management aspect is typically quite complicated. We're lucky because our system is super intuitive, and because our ICP is used to being fast and adaptive. In the size bracket our customers are in, their processes, routines and structures are not set in stone, because they're still evolving themselves. One day they're offline only, the next day they open an online shop. One day they say they'll never be on Amazon, the next day they have a TikTok shop. So they're used to constant change, and with the pressure everybody has been under, probably even more so over the last couple of years.

So it's not easy to get 50,000 people onto a new process or a new module, independent of AI, but we've been quite lucky with it because we've really innovated our product over the last couple of years. With AI it's no different, because as I mentioned, very often they don't even know whether it's AI or not. They also don't care. We're not marketing things as the new AI function. We just call it new functionality. It shouldn't matter to anybody how the engine under the hood is operating.

The one thing that's certainly better now is that you can support the change management process with AI. Let me put it this way. A lot of people see the need for a new ERP system. Our biggest competitor right now is Excel sheets and Google Sheets and some old, outdated legacy system. Most people stick to those system dinosaurs because they're afraid of changing to a new ERP. It has this aura of a one-and-a-half-year implementation project that costs hundreds of thousands of euros. To be honest, they're right; I've done this a few times in my career. Then there's this new era of modern systems such as Xentral, where we onboard customers in two to four weeks if they're super small and uncomplicated, and at most two, three or four months depending on the speed of the customer. Rather fast and rather cheap.

Now with AI, a lot of the individual building blocks of switching from one system to the next, or implementing an ERP for the first time, can be helped along: data mapping, data migration, configuration. What I'm getting at is that you can now conceptualise tools that weren't there before, that alleviate the change management process and get to adoption faster. The one thing that helps, and I get a lot of critique for it, is providing an end of life for the old feature. You're forcing customers to their luck. With our new analytics module, the first wave is the early adopters who want to do it. Then the important customers that are too big, so you help them. But the absolute laggards, the ones who wouldn't migrate no matter what, at some point you need to give them a push by announcing end of life and end of support for the old module. And typically, once they've migrated, customers say, we should have done this a long time ago. Last sentence on this: the less work our customers need to do to keep the system up to date and migrate to new things, the better. With newer modules there isn't even anything being done on the customer side. All the changes happen under the hood, so customers don't need to get involved to move from an old module to new technology.

Shamil Malachiyev: Sometimes, looking at the current tooling that even large companies use, I feel like in 10 or 15 years we'll still have a lot of companies on Excel spreadsheets instead of all the new technological advancements we integrate.

Domenico Cipolla: Cloud software has been around for what feels like a hundred years, and people are still on outdated legacy systems. All the banking software in Germany runs on code some guy developed who's long dead, and if that ever crashes nobody knows how to safeguard the whole thing. So there's a lot of inertia in the market, for sure.

Shamil Malachiyev: When thinking about AI and the way users will interact, there are quite a lot of ideas floating around. Some say the UI stays within your SaaS platform, it just gets a lot more agentic features. Others say the future of work is everyone working through Claude, Cowork, accessing all your features through their interface. What future are you planning for? What are you seeing?

Domenico Cipolla: That's a good question as well. I'm not sure we have a final position on it. We're focusing on the things we can affect. We understood the power of AI early, so we made all our systems, modules, our whole tech stack, accessible to it. When we're developing agents, they have all the endpoints. We have a data layer, a data foundation, that can operate across the entirety of the tech stack, and on top of that we're now developing our agents. In addition to the tools I mentioned before, where customers don't even know whether it's AI, we're now deploying our first agents. We currently have 50 customers testing them, and we're launching them in April and May. It's a dunning agent, an invoice reconciliation agent, a reordering agent, a customer support agent, a returns agent, and so on. Product feedback is fantastic. People are starting to queue to be in the next wave.

Those agents sit on top of the data layer I mentioned. It wouldn't be a big thing for us to open that data layer to external agents as well. If you had an agent that works across your CRM, your marketing analytics and your ERP, and the other two systems were as open as we are, it could access everything across the board. Our position is that AI is just a means to achieve something. The real power lies in the business logic and the understanding of the business logic, the business processes, the requirements of our customers. Developing an agent without that context of the business setting our customers operate in would yield very, very inferior agents. So our approach is, let's build those agents. And we'd like those agents to also access systems outside our ERP, and be that agentic layer that optimises and operates across systems.

Rather than have a third party do that, if we're fast, if we're good, and if we really understand our business processes, there could be an agentic layer that we develop, working on behalf of our customers across different systems. One thing people need to understand: if I think about the tech stack of our customers, there are two gravitational points. There's the online shop, where a lot of the marketing analytics, CRM, emailing and payments go. And there's the ERP in the background, where a lot of the CRM, warehouse management, inventory and accounting sit. It's very different from a point solution, which is quite isolated, with a couple of inputs and a couple of outputs.

Shamil Malachiyev: Inventory.

Domenico Cipolla: The ERP is pretty much at the core of the business operation. So I think we're destined, together with the shop, to be that data hub, that system of record, or system of reference as we call it, where all the data flows in, and the agentic layer sits on top of that.

Shamil Malachiyev: I'd also be interested in your experience when you talk to people about AI and AI agents. What are the common things everybody gets wrong? How do we explain to people what AI agents are not?

Domenico Cipolla: That's an easier one. Without giving away too much, I'm currently looking at a software company together with a private equity fund that's interested in buying it. There's a commercial due diligence provider, a technical due diligence provider, the private equity investor, and I'm sitting on those calls as an industry or software advisor. The level of the discussion around the AI ability of that company is, if I were to say it's scratching the surface, I'd be very generous. Everybody is just thinking about, yeah, with AI, they're going to be able to do this and this and this. And nobody is understanding that if you pour AI into a really outdated system with outdated data structures, with every data being super isolated, it's not like you do that and then magically you have the super powerful software. There's a real difference between the quality of the tech stack and the impact it has on AI. That's number one.

Number two. You asked me before about our strategy. I see a lot of companies with grand strategy papers, and if you ask them how they use AI internally, they say, we're mandating everybody to use ChatGPT at least once a day. And they think that's going to be the trigger event for their AI strategy. So, to answer more concisely: everybody needs to be extremely aware that AI is just a means. It's not a value in itself. That's number one. Number two, the quality and accessibility of the tech stack play an important role. If you have nonsense coming in, the best AI is only going to generate nonsense coming out. Those are the two on the sceptical side.

The third: I heard a lot of scepticism at the beginning in my industry, along the lines of AI is never going to be able to do this. I think that's also wrong. People are hugely underestimating the impact AI is going to have in our industry. It's been lurking around for the last two or three quarters, but now I really feel it's manifesting itself in powerful use cases that generate real value. So I'd be very cautious not to underestimate AI. As a matter of fact, we're banking on it. We're betting a lot of our product development roadmap on it. And the fourth, after two negatives and two positives: whatever was impossible yesterday seems already outdated today, because things change so fast. I joked before: I've been on vacation over Easter and haven't followed all the posts, and on LinkedIn everybody seems to reinvent the wheel every day. I wouldn't be shocked if yesterday a new model or a new functionality launched and suddenly something new is possible. Keeping a close eye on developments, filtering out what's nonsense and what's actually actionable and valuable to you, then implementing that into your own product roadmap, is going to be key.

Because, to answer slightly differently, there's a lot of panic around this. There's a lot of euphoria when you speak to people one or two degrees removed from it: potential investors, shareholders, consultants, advisors. Everybody listens to the same podcasts and reads the same LinkedIn articles, and I think that's a very distorted and small perspective on reality. The reality I see on the ground is very, very different.

Shamil Malachiyev: Well, this is exactly the reason for this podcast. That's exactly what we try to do. I have two questions on that. First, for anybody listening who's in their department, in their company, trying to understand, with senior management or shareholders saying, okay, we need to integrate AI agents into departments. You mentioned knowing that your data structure is ready, that the tech stack is there. How can these people evaluate whether the data structure within their department is ready for AI to be involved?

Domenico Cipolla: The most practical advice I'd give is to think of an internal use case. It shouldn't be too big, but it should have some real practical implications. Then just try to get it done. If you get it done, it means the data layer is there, you have the endpoints, you have the access rights. Let me give you an example. I'm not a technical person at all.

Shamil Malachiyev: No.

Domenico Cipolla: But I'm curious, so I like to experiment myself. When OpenClaw came out, I tried to get it running. No chance. It asks you whether you want to do this on your machine or on a virtual server, and it's an endless series of questions where every time I needed to use ChatGPT to find out what they mean and what the advantages are. Then Claude Cowork came out, and I said, okay, this is OpenClaw for dummies, I need to be able to do this. I got much further. Then, I live in Germany, I have an old university account, which is my private email account. One of the things I wanted to do was connect my email and trigger some automations based on it. The practical reality was, because it was my university account, I asked the admin of the university: Claude Cowork would like to access this, could you please give authorisation? And the answer was, that's a very good idea, thank you very much for your request, but no.

The analogy is: try a use case, try to do something internally within your tech stack, and see where you hit a wall. When you hit that wall, remove it. Once you've done this two or three times, you can abstract from it. It's not that you're going to go experiment by experiment every time you hit a wall; you try to understand what those walls are and get a macro perspective on them. In our case, a couple of years ago we still had a pretty monolithic, redundant, spaghetti-code tech stack. Separating that out into, I don't want to say microservices, but very clear business objects, very clear business logic, an application framework, the right API endpoints, that isn't done over a couple of weeks. That's a complicated and protracted IT transformation. But we've done it, and it's a prerequisite for the new era in which AI is going to play a much bigger role.

It's funny, because of this email account example: the compatibility of the components of your tech stack with AI is going to play a major role. It's going to determine quite a lot who survives this and who doesn't. My answer now is, okay, I need to switch my email account, because I don't think that IT admin at the university is going to change his mind anytime in the next 25 years. So I have three options. I stick with it and make my peace with it. I change my email account to somewhere more AI-friendly. Or I forward everything to another account, but then the reply comes from a different account, which isn't ideal. It's interesting. When we jumped on this call I saw you're using Fireflies for transcription. I'd been using something else, but when I wanted to connect that transcription software with my Claude Cowork, it didn't work, so I changed to Fireflies as well, because there's a native integration. It's very interesting how that ecosystem of modern components is emerging, and the ones staying behind are sooner or later going to fail.

On your earlier question about our own agents versus third-party agents, and whether we have a final decision: if everybody is going one way, we're not going to stand in the way of our customers' wishes. We believe we're going to develop faster, more efficient, cheaper agents that understand the business logic much better. But at the end of the day, we're servants of our customers, and we'll do whatever we need to do for them to thrive and be more successful with us as their ERP provider.

Shamil Malachiyev: By the way, a small thing on Claude Cowork. If you need access to your university emails and there's no existing capability to connect the systems, I just open it in Chrome and log in with my credentials. It opens the emails, it can scroll through them, it can give you all the information. But it takes a long time.

Domenico Cipolla: I agree, but I said I'm not a technical person, and I'm a little bit of a design person. That's not the user experience you want, having to open Chrome and use an extension. If we invented AI, and now we need to take all these extra detours to make it work, that's not the idea.

Shamil Malachiyev: Yeah, that's the worst one. The other question I wanted to ask is about hype. We look at the advancements and every week there's some new thing coming out, models are getting better, hardware is getting more efficient. Then when you look at real adoption, where people are actually using it, the technology is advancing geometrically but adoption goes very slowly, almost at the start of the adoption curve. When we look at company valuations and all the LinkedIn posts, they almost say the technology is there and adoption is almost there too. Based on that, do you believe we live in an AI bubble?

Domenico Cipolla: Bubble, you always need to define it. I said I'm on vacation, but I did see a super interesting comparison, I think yesterday or the day before, between the internet bubble and the so-called AI bubble now. They were talking about optical fibre, the overproduction, and how it took years to catch up. Whereas nowadays, with all these data centres being built out, their usage is very, very high. So I don't know about an AI bubble. It clearly feels insane that companies are adding a trillion of ARR per day and being valued at more than a trillion, and then two weeks later everybody's saying AI is the worst company in the history of mankind because it's all hype. There's a lot of volatility. That initial hype is coming down again, and I'm sure it's going to be there.

To answer your question: I don't get involved. We're running a really healthy business that looks at the fundamentals of what AI can bring to us and our customers. We keep our heads down. We filter out the noise. We stay quiet in meetings where external people create more panic than we'd like them to. I think this really unifies us at Xentral, because I feel a very similar sentiment across the company. What I like most about it: I mentioned our internal use cases, the use cases for our customers, and I joked about the company mandating ChatGPT logins once a day. What I like most is that we have a lot of curiosity in our teams. In the people department, once or twice a year we have a talent review process where everybody is assessed, gets feedback and a development plan. Somebody in the people department just used Claude Cowork to automate that process and collect all the inputs, rather than doing a lot of that work manually. She developed something that helps her about a day a week in that process. Somebody in finance did something similar with the monthly close. People in customer success prepare a dossier about the customer with all the information before every customer meeting.

What I mean is there's a lot of curiosity and ability to experiment across the teams, which is not mandated top down. We have really talented people in their departments who think, what am I doing, how can I automate this, can I use this? I'm sure a lot of those efforts fail; that's the nature of experimenting. But that curiosity, people starting this on their own, sometimes in their own time, is really remarkable. A lot of good is going to come from it for those people, for us as Xentral, and in turn always for our customers.

Shamil Malachiyev: How do you facilitate that? Obviously it's good when people naturally come up to you saying, please give us Claude team accounts, we want to do some things. But say half a year goes by and you don't see anyone doing anything innovative. How do you try to be like, hey guys, let's do an internal hackathon? Or mandate a login to ChatGPT twice a day?

Domenico Cipolla: That's a really good question. I have two hearts in my chest on this one, as we say in German. Facilitating could very quickly turn into impeding. There are these green shoots everywhere, and they're coming. The consequence is that they're not going to be synchronised with each other at all. Somebody is going to use Claude Cowork, somebody is going to use OpenClaw, somebody is going to use Claude Code. Everyone is going to do something and a lot of good is going to come out of it, but it's super heterogeneous. So the question is, should we step in and say, from now on you need to use Claude Cowork, only that one is paid for, this is our application framework, before you release anything it needs to be QA'd by Peter, and by the way we shouldn't have infrastructure costs that run away? You have a higher chance that things speak to each other, that it's better maintainable, that security is better. But on the other side, that facilitation, that regulation, that rule-based approach... The question is when the right time is. Right now I'm erring on just letting them have fun. When we're all a bit more experienced, that's the right word, about these things, then maybe it's a good time to bring the efforts together. But right now, I'd err on the side of experimentation and let them have fun.

Shamil Malachiyev: Perfect. Thank you. I want to thank you for joining the podcast and sharing all of this. It's been incredibly informative, your view into how to adopt and how to facilitate. I personally got a lot of notes that I'm going to take back to my teams. Hey guys, ChatGPT three times a day. I want to thank you for being such a fantastic guest and sharing the reality of what AI adoption and thinking look like. Thanks, Domenico.

Domenico Cipolla: I very much enjoyed this. Thank you very much for having me. As you put it during the conversation, this topic is right on point. It's about taking a step back and looking through the noise of everything being posted everywhere. I think AI is extremely powerful, but panic has never helped anybody understand what those elements are, how you can make use of them, what's useful for you, and then chart your own way forward. That's really the recipe for being successful in this AI era, and that's what we're trying to do. So thanks again, Shamil, for having me.

Shamil Malachiyev: Thanks, man. Appreciate it.