Humanising AI adoption — with an untraditional toolbox.

AI teams are full of people who can build. What they are usually missing is the combination this profile brings: someone who has architected production systems for two and a half decades and holds a degree in philosophy and philosophy of science — and uses both at once. The engineering half knows what it takes to ship; the philosophy half knows what questions the shipping quietly assumes: What warrants this output? Whose judgment stopped the regress? Which assumptions have expired? When AI enters an organisation, those become operational questions — and most teams have nobody whose job it is to ask them.

The mission is humanising AI adoption. Not slowing it down — shaping it so that people remain the point. That means multi-agent workflows dressed up for human interaction rather than chat boxes that transfer the burden to users. It means human checkpoints with real authority, quality processes that anchor knowledge in teams instead of dissolving it, costs in plain sight, and language people can use to name what is happening to their work. The talks and the working system on this site are that mission, demonstrated.

The craft behind it is harness engineering. Day to day he works on complex agentic harness development across different environments — the machinery of tools, policies, verification and delegation that turns a model into a dependable system, and the discipline of knowing what each layer stores epistemically. This site's experiment made a deliberate methodological choice from that background: stock Claude Code with no extra add-ons served as the harness, for the most honest impression of what Claude Code with Fable 5 does on its own — fewer than 500 human words of direction became the working studio you are browsing. The experiment write-up is on LinkedIn.

And the experiment's honest ledger is the profile's argument. What the model did not do is the finding: it never wanted anything, never questioned the roadmap, wrote thousands of lines and zero tests, twice reported success on changes that did nothing, and missed flaws any user catches at a glance. The experiment measured what the machine does alone — and what it revealed is the shape of the human it needs. That is the uncomfortable news for the dream of automating everything, and the entire case for this profile's work: humanising AI adoption is not sentiment. It is where the value is.

The road here

• 2026 —
  Senior Software Architect, Continia Software — shaping the shift from traditional software to agentic systems and development workflows.
• 2023 —
  Engineering Manager, Continia Software — building environments where AI agents get better on their own, not just better instructions to follow.
• 2020 – 23
  Senior Backend Engineer, Continia Software — cloud solutions with Azure and Kubernetes.
• 2013 – 20
  Technical Project Lead & Architect, Niras A/S — seven years of delivery and architecture in engineering consultancy.
• 2004 – 13
  Telenor, Norriq, Safetrack, Comlog — senior developer and software designer years: business integration, telecoms, and Internet of Things solution architecture.
• 2001 – 04
  Self-employed — independent software work straight out of philosophy.
• 1994 – 01
  Cand.mag. in Philosophy, University of Southern Denmark — the discipline that keeps asking what an answer is worth.

Why this page exists

Because the combination is rare, the mission matters, and the conversations it starts are worth having: exchanging ideas with people wrestling with the same questions, comparing notes on harnesses and human-first adoption, and the occasional talk. One standing exception: NGOs and volunteer causes working to humanise technology — for those, the door is open, pro bono.