The demo always works. The organisation is the hard part.

In 2018 I built a voice-driven AI banking assistant with a colleague at Swedbank, where I ran product for payment infrastructure and Nordic cash management. We demoed it to the whole company — top management in the room. It never shipped. And honestly, for a conservative bank in 2018, not shipping it was the right call, by the rules banks had to play by.

What stayed with me wasn't the decision. It was the realisation that no amount of better technology would have changed it. Eight years later the models are extraordinary, and I keep watching the same film: pilots that impress, then stall. The constraint is the operating model — how work gets defined, approved, funded and delivered. That's the thing I fix now.

I've spent sixteen years inside Nordic banking, as an employee — not an observer.

At Swedbank (2010–2020) I ran IT delivery for core banking and lending — 120+ people across five countries, including the euro implementation in Estonia — then digitalisation of sales and customer service in Stockholm, then product for payment infrastructure and Nordic cash management.

At Luminor (2020–2023), after the largest merger in Baltic banking history, I was Head of Corporate Product Development. We migrated payments off Nordea and DNB so the bank could stand on its own, and I built the product management organisation from scratch — taking the organisation from project-centric to product-centric. In my experience, that shift delivers roughly 40% faster time-to-market.

Most recently I was VP Payments & Cards at Bankdata (2023–2026), Denmark's largest banking IT provider — eight banks, roughly a third of the Danish population. My second product-centric operating model transition, this time with becoming AI-first woven through it: legacy modernisation, developer productivity, and the product lifecycle itself.

Along the way: direct work with regulators and central banks in four countries — Denmark, Lithuania, Latvia and Estonia — on PSD2, instant payments, AML and card-scheme compliance. MIT Sloan's AI programme. An Executive Master of Management from BI Norwegian Business School. Board accreditation from the Baltic Institute of Corporate Governance.

I also mentor the founding team at EndoBud, an AI health-tech for women with endometriosis, as they work through architecture and AI under GDPR Article 9 special-category data rules.

The vocabulary of this field — product operating model, empowered teams, feature factories — was popularised by Marty Cagan and SVPG, and I use it gladly. But the model Tekhouse installs is my own, built and tested inside regulated Nordic institutions: an evidence gate before any build, work classified before it's queued, a clear contract between product and technology leadership, and an agentic delivery layer with a cost ledger on every dispatch.

Tekhouse is deliberately small: a core team and a network of senior consultants — ex-Swedbank, ex-Luminor, ex-Bankdata and beyond — who subcontract to each other when a job needs one specific pair of hands. When a strong engagement lands, we staff it with people who have done exactly that work inside exactly those institutions.