I help organisations run AI they can stand behind.
Whether you already run AI in production and aren't sure it holds up, or you're about to start and want to avoid the expensive mistakes. I mostly work on what is already running: finding out whether it works the way you think, and what needs to be in place before anyone else relies on it.
SøkTilskudd and SøkStøtte are the proof: tools I have built, put into production and remain accountable for myself. Below are the ways I usually work, but you don't have to pick one. Tell me what you're dealing with and I'll say what I think the right next step is. I take on a limited number of engagements at a time, so each one is done properly.
AI trust review.
I review the AI you already have in production and answer one question: does it work the way you think, and can you show that to a customer, an auditor or a regulator? Two to three days. You end up with findings, a prioritised action list and a set of tests you keep running yourselves.
A good fit if you've rolled out coding agents and found that guidelines on paper stop nothing, or if customers, caseworkers or regulators are about to meet the AI directly. A poor fit if you want a general AI strategy: this is a technical review of something already running.
When the AI Act starts to apply to you, what takes time to build is the testing and the documentation, not the paperwork. The review covers part of that: findings with evidence, a prioritised action list, and a set of control questions you keep running.
How it works
- 01
Where is AI used — and what is at stake?
Where AI is actually used, which decisions it helps make, and what happens when it gets things wrong. We begin with the consequence, not the technology.
- 02
Written or enforced?
Most organisations have rules. Few have rules enforced in code. I go through what actually stops an error, and what only exists in a document.
- 03
Can it say "I don't know"?
Can every claim be traced to a source? I build a set of control questions with known answers for your field and run them against the system, including questions it should refuse to answer.
- 04
Findings ranked honestly
Every finding is marked verified or assumed, and I say what I did not have time to look at. A review that implies it covered everything is worth little.
- 05
I walk you through it
Findings, sequence and what is urgent. One hour, with the people who will actually do the work.
Three steps further, if you want them.
The review stands on its own. But if it finds something, this is the way forward. I have built all three for my own products first. I use governance and operations every day; the advisor is in pilot. The scope for you is set once we know what you actually need.
- 01
Governance in place
I install and adapt my control setup to your stack and risk profile. Guards that stop an error where they can, and shout loudly where they can't. And tests that prove they work: I deliberately weaken each guard and require the test to go red. A green test that has never been attacked is an undocumented promise.
Engagement · typically 1–2 weeks
- 02
An advisor on your own rules
An AI advisor built on your own body of rules. The ground truth is curated and dated, not a vector search across everything you have, and every fact has a source I can point to. In the pilot I measure the hard part: not whether it makes things up, but whether it stays silent about something that should have been said. That is the error that costs most, and the one fewest people measure.
Pilot first · I measure before I promise
- 03
Operations
I keep the knowledge base current, the tests running and the guards in step with change. A knowledge base that isn't maintained goes wrong without anyone noticing — and then nobody gets told.
Ongoing agreement · scope and price set after the pilot
Workshops.
Three topics I deliver as workshops, tailored to your organisation. Format to suit you, on-site or online.
AI for product managers
From prompt to process.
For PMs and product owners who want to integrate AI into product work without losing the craft. Strategy, prioritisation and value realisation with AI as a tool.
Tailored to your team, in person or online
Request a proposalAgent-driven development in practice
Hands-on with agent-driven AI tools.
For teams that want to build faster with agent-driven tools. Practical hands-on with the setup we use to build Kveck. Workflow, governance and quality assurance.
Tailored to your team, in person or online
Request a proposalAI Governance in practice
Evals, red teaming and responsible AI.
For leaders and domain leads who have to ensure quality, safety and ethics in AI systems. Built on Responsible AI Leadership and practical governance work at Husbanken and Kveck.
Tailored to your team, in person or online
Request a proposalBuild with us.
A structured programme that delivers a working MVP, not a pitch. We take one well-defined problem and build all the way through to a validated solution.
How it works
- 01
What needs solving?
We start with the problem, not the technology. Who is the user, what is the pain, and what's needed for this to actually create value. Often as a short workshop on-site.
- 02
Sketch and trade-offs
Architecture, model choice and governance frameworks set against each other. What to automate, where humans still need to be involved, and what it'll cost to run.
- 03
Building
Agent-driven development with AI tools, built hands-on. I build the solution itself, but in close dialogue with you so you understand how it fits together.
- 04
Testing against reality
Evals with golden queries, red teaming and user insight. I measure whether the solution is actually better than the baseline, not just whether it's new.
- 05
What now?
Scale, adjust or stop. A clear answer on what should happen next, with numbers on cost, risk and what it takes to operate.
Advisory and short-term engagements.
A sparring partner, short studies and hands-on support when you need someone who's done this before, not a full project.
Typical engagements:
- AI readiness assessment: Mapping of where the organisation stands, which processes are ready for AI, and what needs to be in place first.
- Governance study: Evals, red teaming, prompt injection defence, GDPR and responsible AI. Practical recommendations, not boilerplate policy.
- AI PM coaching: Ongoing sparring for product managers integrating AI into product work, or building AI-driven products for the first time.
- Agent implementation: Hands-on help setting up agent-driven AI tools in your development flow.
Built on practice, not theory.
I build SøkTilskudd and SøkStøtte from the ground up with agent-driven AI tools, and run the whole stack myself: architecture, evals, governance, operations. What I learn there goes straight into workshops, build programmes and advisory.
Underneath: 12 years of product management from Husbanken, LO and Compass Group, among others. Master of Management from BI with specialisations in Responsible AI Leadership, Digital Security and Strategic Business Development, plus certification as an AI Product Manager. Read more about my background →
Ready for a no-obligation chat?
Tell me briefly what you're dealing with and I'll say what I think the right next step is. I reply myself, and I'll be honest if I'm not the right person.