In this episode of The Connector Podcast, host Koen Vanderhoydonk interviews Pieter, Head of Strategy at In The Pocket, a Belgian digital product studio that began as a mobile‑app agency 15 years ago and has grown to over 200 staff across Europe. Pieter explains the company’s evolution from building gimmicky mobile tools to delivering full‑stack digital products, cloud back‑ends, and strategic consulting. He highlights the current strategic focus on AI, especially agentic AI, and introduces the concept of “agent experience” – applying user‑experience principles to AI assistants in financial services. The conversation covers regional differences in AI adoption (the Netherlands vs. Belgium), the impact of the EU AI Act on strategy, and a concrete use‑case where In The Pocket is helping a financial firm automate customer‑support tickets with large language models while keeping humans in the loop to mitigate hallucinations. Pieter identifies the main barriers for banks: model hallucinations and high compute costs, but notes that smaller fintechs can gain a competitive edge because they can manage costs more easily. Data quality, security, and the need for cross‑disciplinary teams (design, technology, strategy) are emphasized as essential for successful AI projects. The episode ends with a brief pitch for organizations seeking AI‑agent expertise to contact In The Pocket.
Chapters
Welcome to the Connector Podcast, an ongoing conversation connecting fintechs, banks, and regulators worldwide. Join CEO and founder Koen Vanderhoydonk as you learn more about the latest available trends and solutions in the markets.
Welcome to another podcast from the Connector in collaboration with Fintech Belgium, and today I've got Peter with me. Peter from In the Pocket, a Belgium company, and you're head of strategy. So what is In the Pocket and what do you do?
Yeah, good morning, Koen. I'm a strategist at In the Pocket, so I basically help companies make decisions, make tough decisions. There's a lot of things coming at companies at the moment, and they sometimes need help to make those decisions. And that's what strategy is about, is making Smart decisions, knowing why you made those decisions, and moving forward.
I thought you were about to say tough decisions. Sometimes.
Sometimes, yes. No, no, absolutely. It depends. There are some— sometimes it's usually a combination of very easy decisions and tough decisions, short-term and both long-term. And things change quickly, quicker than they used to 5 years ago.
Yeah. Don't tell us, right?
Yeah.
Hey, tell us a bit more about In The Pocket.
So we are a digital product studio at the moment, but we actually started out 15 years ago as a mobile development agency. It was really the start of the mobile revolution. First iPhone got released and yeah, that was a huge, huge revolution there. Companies could finally make apps for phones, and so that's actually how we started. We started out with sometimes gimmicky, gimmicky things like we made an Astrid Bryant soundboard, for example. And obviously, in those 15 years, a lot changed. Now, you know, we work for really big organizations whose whose main product is actually a mobile application. But we moved far beyond mobile apps. It's actually a only a small part of what we still do. So actually we do anything, any digital product, cloud backend, but so also strategy. So even if there is no specific product involved, we help organizations make decisions.
Makes sense. What actually struck me in your explanation is that you mentioned the beginning of iPhones. So what does that mean? Your company's in business for a long time?
Yeah. So we're celebrating our 15th year.
Wow.
Congratulations. We started as 3 people. Now we're over 200 across Europe with studios in, in Bucharest, in Amsterdam, and in Lisbon. So yeah, obviously a lot changed in that time.
Yeah, that's a large success story actually. And how long have you been with the company?
I've been now at In the Pocket for 5 years. So within the strategy team, where we're looking at all different aspects of strategy, business strategy, but also data strategy, and lately also a lot of AI strategy, because obviously a lot of questions are about AI these days. I think it's one of the main evolutions that we've been seeing. We obviously had the mobile revolution where, yeah, a lot of changes came to companies. They had to rethink their products through that tiny device. And obviously now they see all the possibilities of of AI, also some threats, of course. And we need to figure out what you can do as a company to adapt to it.
Would you see this as looking at the world through a different lens, or is that too narrow to look that sort of, or to take that as a view?
No, I think it's definitely a different lens. It's a different company, a different world where we live in. And also the speed of change is increasing a lot, right? So you can't really sit this one out and say, okay, let's wait 2 years and we'll figure out what we're going to do then. That's too long, right? So if you look at when ChatGPT 3.5 got released, it's not that long ago. And since then, I mean, we've seen a huge boom in terms of technology and now also— This year, I think, will be a really important year in terms of AI agents that will, especially in the financial markets, start taking over a lot of tasks that people are currently doing or extra tasks that simply were not possible at the moment and will be possible through AI agents.
And how do you see like a product team like you are in the Pockets having a role in implementing agentic AI?
Yeah, I think that's an important question. So one interesting concept around this is that it's called agent experience. So if you've been in the design world, you know about user experience, right? So you, you think about the needs of a user, frustrations, and you try to basically take away all, all the friction to make the best possible, for example, smartphone app. And the same goes for agents, right? So people are already looking for financial information through chat assistants, and this will only continue. So that's a really important question, I think, for financial organizations to ask at the moment. Like, how will people interact with our products and services through an agent, through an assistant? And then you really need to think about, okay, how do we offer our services? Do we include them? Do we work together with them? Or do we try to fight them with our own agent, for example? So I think those are a lot of questions that financial institutions will need to ask now.
And you said earlier that you are at different locations. Do you see a difference in terms of speed of adoption on the different locations that you mentioned?
I think there are some local differences. Like, for example, in the Netherlands, you have a few players like Bunq that are really pushing the envelope. So obviously then banks in the Netherlands are trying to catch up quicker. In Belgium, I would say we're a bit slower. I think most banks are still looking at one another to kind of see, okay, who's going to be the first mover there? We obviously have KBC that is pushing a lot with their KATE product. But yeah, there are definitely some local differences. I think one thing, of course, is because it's such a global market is that we'll be able to adapt much quicker. Whatever is happening in the US or in any other market, especially in Asia, things are moving fast. Usually we can pick up also much faster, especially from the technological point of view.
Maybe a more sensitive question, because we're here in Europe and a lot of discussions nowadays about Deregulation. So how do you look at regulation when we talk about AI?
Yeah, so obviously it is extremely important. We also have the AI Act since last year. So we also help organizations to look at that and to really kind of figure out for themselves what is the best way to approach this. There are now more or less clear guidelines. And so you have a playing field in which, you know, like, okay, we need to move within those boundaries. And it also helps you to make better strategic decisions because, you know, yeah, moving in a certain direction can, can really set you apart from your competitors.
Are there any success stories that you can share from InThePocket and specifically on the more agentic AI level?
Yeah, we're currently working with a, with a financial company to, to help them speed up their customer support. I think customer support will be definitely one of the first areas where we see a lot of quite successful agents simply because it's very text-based. It's quite repetitive. People have the same kind of questions over and over. But it's really interesting there where an agentic flow can have those small differences. Like, for example, if people have multiple questions in a single ticket where a typical rule engine that we had before would kind of block on one of those questions.
Yeah.
Obviously here using large language models, we can really move forward. And one really important aspect there is where we try to stay very aware of how we combine human and artificial intelligence because obviously the cost of being wrong can be quite high here. If your agent sends out a bad email or gives out the wrong information, you don't want that.
Hallucination is not the right place.
No, no, exactly. And I mean, if the user kind of understands that it's a hallucination, that would be one thing. But obviously it would be even more dangerous if they think that this is indeed the actual explanation. So that's what we're still working on, keeping humans in the loop, making sure that that, you know, their customers are served in the best way. But, but so far it's been really promising. It saves people a lot of time and it improves the quality of the service that people are getting. So that's really a win-win situation.
What do you see as the biggest stumbling block for financial institutions to sort of take that step towards agentic AI?
Yeah, definitely. Like you need some good guardrails in place there. I think at the moment, 2 things. So first, there is the whole hallucination thing, the quality of the output. You don't know, I mean, these models are non-deterministic. You don't know exactly in which situations they're gonna fail, but you know that they might make mistakes. So that's one thing. The other part is the cost. At the moment, running these models, especially the state-of-the-art models, is still really expensive.
Mm-hmm.
So if you're a big organization with a few million customers, Those costs add up. So those 2 things will be resolved, I think, in the near future with better guardrails. Yeah, simply more reliable models. And obviously the price also will come down. But at the moment, those 2 things are still a bit of a struggle for a lot of companies.
No, and then I think the level of change of LLMs, it's a dazzling speed at the moment, right?
It's very fast. But at the moment, what is really interesting is that it puts smaller fintech players with an advantage, because if you have millions of customers running those kind of services, it is insanely costly. If you have a small user base, you're really starting out now, then the costs are manageable.
True.
So I think at this moment, it's a really exciting time to be a fintech player where you look for a specific niche, you look how AI can really help deliver a great service, and you can really outperform a bank 100 times your size.
Well, you mentioned Bunq earlier, and I had the pleasure and joy to have Ali, head of data and AI of Bunq, in my panel last week in Zurich. And I asked him the same question, and he said, for me, it's all about data structure. So he brought it back to the very old saying, Rubbish in, rubbish out.
Yeah, I mean, obviously that's where everything starts. It's also, I think, the part where companies can already start with experimenting. I mean, you need good data. And in general, like, I have a double feeling about this because I also think that those models will get better and better at sifting through data anyhow. So it may be a problem that solves itself eventually, but at the moment, of course. Yeah, you have a lot of data points. It also brings in a lot of data security questions because if you take in huge datasets, yeah, you need to be aware of what exactly went in.
Vulnerability goes up as well.
Yeah, yeah. So that's definitely something, yeah, some important questions that companies need to ask themselves.
Well, interesting times. Our podcast has been going on and we're almost at the end of it. So can I ask a strategist to be a salesperson for 10 seconds and tell us who should contact you and where to find you as InThePocket?
Yeah, I think anybody that has a lot of questions about AI and agents should definitely contact me. We have a really nice team that is cross-disciplinary, which I think is really important. AI is really a technology that requires people to bring together design, technology, and strategy. So you need a party that can really bring those together. And what's interesting for us is that obviously we can really go from an idea on a napkin to an actual product, and we're also able to build it and scale it. So I think that's quite unique.
Well, Pieter, thank you very much for sharing your story.
You're welcome.
Thank you also very much to the audience. And please stay tuned because more news from the world of fintech. Thank you so much.
Thanks for listening to another episode of The Connector Podcast. To connect and keep up to date with all the latest, head over to www.jointheconnector.com or hit subscribe via your podcast streaming platform.
