Designer, team lead,
AI builder.

I take AI past the pilot, and keep it running.

Most AI implementations stall on the distance between the tool and the daily work. I take a proof of concept to production in contained building blocks with a measurable outcome, start with flows instead of tools, and put the operational backbone in place from day one, including governance and EU AI Act readiness. Memortium has run that way for over a year, with thirty paying clients and two hours of my time a week.

Two things make that possible: my design experience and my AI build skills. Over twelve years of enterprise UX show me where work stalls and how people come to trust a system. The agentic build practice I run every day turns that into working production in days instead of quarters.

Cost discipline lives in the architecture, from the first line.

The token bill often grows faster than the value beneath it, because the heaviest model gets pointed at every task. In production I route each task to the model that fits it: cheap work cheap, the heavier models only where they pay off.

These choices are not made on instinct. I know the shape of the models I use daily and where a lighter one takes over without losing quality. Token economics and FinOps are baked into the architecture, so cost scales with value and not with volume.

My agents hold up outside the demo, too.

Agents that shine in a demo often fall over in production: a missed edge case, a handoff that breaks, a tool call with no safety net. The multi-agent setups I run hold up, with guardrails and human-in-the-loop where a mistake counts, and Mono Dash handling the orchestration.

The difference is knowing what an agent can credibly do, and where it turns into wishful thinking. I design the orchestration layer underneath: who hands off to whom, where state lives, where the approval gate sits. Evals and guardrails keep an agent fleet honest when it actually counts.

I design AI that people understand and trust.

Adoption is where most rollouts die. The tool is there, the licences are paid, and still nobody reaches for it. What is missing is a design for how the system fits the workflow, and whether people trust it enough to open it.

That is my foundation. Years of enterprise UX taught me exactly where adoption breaks, and I design for it: confidence indicators, a clear fallback, human-in-the-loop where it counts. I handle adoption and change with training inside the work itself, tied to the flow people just picked up.

A rollout only succeeds when everyone comes along, from Annie at reception to Johan in the boardroom.

On both sides of the bridge: building, and knowing what a choice costs.

There is a gap between the people who build AI and the people who carry its consequences: engineers who do not speak the business, leaders who do not know what a technical choice costs. I stand on both sides, from years of cross-functional work at Shell, Roche, Philips and ING and two years of 0-1 product ownership on my own ventures.

I translate a business goal into an architecture, and an architecture choice back into what it costs and returns. In a world where building takes days, the split between who advises and who builds disappears. I fill that role myself: designing and building in one motion, with stakeholder management and the team in control.

Governance belongs in the building block, not bolted on after.

The AI Act applies in full from 2 August 2026. Bolt governance on afterwards and you stall on questions no one can answer anymore: who saw which data, where a decision came from.

I build it in from day one. Per agent it is set what may run autonomously and what goes to a person, with logs and an internal owner who signs for exceptions.