Robbie Tilleard: ‘Build for Today and You’re Already Obsolete’

Robbie Tilleard caught me preparing for our interview the way everyone prepares now, reading his company’s website, and gently called it out. “What are you doing? Just use Claude. Nobody reads any more.” He was joking, but the joke is also his entire business. If a machine can read the website, summarise it, and tell me what to ask, then a machine can also read your help docs, know your customer, and actually solve their problem. Tilleard is the GM for EMEA at Lorikeet, and the banner above his booth at Money 20/20 Europe made a bold promise: “the universal AI concierge that gets it done.”

The concierge framing is the whole pitch, and he explained it with a hotel. Bad customer support is the front desk handing you a thick folder of restaurants and saying “good luck.” A concierge says, “Great to see you again, I know you love Italian, I’ve booked the place around the corner and there’s a car waiting.” One dumps information on you; the other gets it done. The interesting question is how you let an AI behave like the second one in an industry where a wrong answer is a compliance problem.

The harness is the whole trick

The technology that unlocks a concierge experience is the large language model, but Tilleard was adamant that the model is not the product. The product is what Lorikeet builds around it, what he calls a “harness.” It is a useful physical word: you take all of that raw intelligence and training and wrap it in constraints and guardrails, so a regulated business can be confident the agent achieves the compliance outcomes it has to hit.

That harness is why his customers are the hard cases rather than the easy ones. Summ, a crypto-tax firm, uses Lorikeet to automate a reported 90% of its queries (a figure worth attributing to them), and this is genuinely thorny territory, how you file taxes on a portfolio of crypto investments. Magic Eden, the NFT marketplace, uses it to drive resolution across its support. The common thread is companies that are complex, regulated and compliance-bound, the exact businesses that would never hand customer support to a generic chatbot. They trust it, Tilleard argued, precisely because of the harness, not despite the LLM.

“We’ll just build it ourselves”

The obvious objection in 2026 is the one I raised: with the right setup, can’t any company wire up Claude and a few tools and do this themselves? Tilleard’s answer was confident and specific. The question, he said, is whether you are spending every waking hour on the frontier of this technology because his team is, and almost no one else’s is. Chat support, he conceded, a couple of good engineers might pull off. Voice is another planet. Building a genuinely good voice agent is brutal, experimental frontier work that he would not recommend unless you can spend millions a year and put five or six squads, twenty or thirty people, on it full-time.

So it comes back to focus. Why would a payments company, or a media outlet like ours, peel attention away from its actual product to build an AI concierge on the side? He has the same conversation with firms like TapTapSend and Airwallex, whose job is getting money to people on time, not building a support platform. And he has the receipts: “Our sales pipeline is full of people who said they were going to build this themselves and have come back.” A demo in this space is easy. Getting it right in a regulated, compliance-driven industry is much harder and slower than anyone expects.

The button is about to disappear

Where Tilleard got genuinely interesting was the near future. He thinks fintech is at the edge of a UX shift as big as anything happening in the back-office infrastructure. Open any of your finance apps today, and you are pressing buttons, lots of them, to do anything. Within five years, he predicts, that interface dissolves.

“You’ll simply go onto the app and say, I want to do X, and it will do X for you, without pressing five buttons to make a transfer. Maybe even better, you’re doing it from Claude or ChatGPT or Gemini. You say, what would it take to send this transfer? And it just does it.”

He is appropriately skeptical of the louder hype, noting that he has not yet seen anyone paint a truly practical vision of agentic commerce as a finished product, with the hard part being not the infrastructure but the user experience and how it all knits together. What has changed since last year, he said, is that people have stopped merely talking about this and started shipping it, running real experiments in production and learning what works.

Never build for today

The discipline that keeps Lorikeet ahead is a single rule, and it doubles as the best line of the interview: if you build a product for where the technology is today, you are already obsolete. He made it concrete. Back in December, a new generation of models arrived that were markedly better at “task adherence,” the ability to reliably go step one, step two, step three. Lorikeet had bet on that improvement in advance, switching its core product from rigid structured workflows (a box, an arrow, another box) to “natural-language workflows,” where you simply write out the steps you want in plain language, no technical skill required, no getting trapped in a flowchart.

When they first rolled that out, the technology could not quite carry it. Then in December, it suddenly could, and because they had been ready, they were ahead of the market instead of scrambling to catch up. That is the whole philosophy in one anecdote: skate to where the models are going, ship in small experimental steps, and be standing there when the capability arrives. Tilleard’s hat, I noticed, read “no dramas.” For a company operating on the bleeding edge of AI in regulated finance, that is either wishful thinking or the entire point. After fifteen minutes with him, I am fairly sure it is the point.

This interview was recorded live at Money 20/20 Europe 2026 at the Amsterdam RAI. For more on the agentic-AI shift, watch the full Cryptofocus series, including our conversation with Money 20/20’s own Bryony Naylor on the year’s big themes.