AI where it makes sense

AI ISN'T A
PRODUCT STRATEGY.

Sometimes it transforms the product. Sometimes a normal algorithm is better, cheaper and more predictable. The job is knowing the difference — and building either one properly.

Built and hosted in the EU. GDPR and the AI Act are part of the design, not an afterthought.

Pencil study of product details: chamfered edge, recessed button, speaker grille, cross-section

THE QUESTION ISN'T
"CAN WE ADD AI?"

It's whether the feature should exist at all — and whether a model is the right way to build it. A clever AI feature can solve a problem nobody has. A boring rule can quietly save someone an hour a day.

The honest split

When AI wins — and when it doesn't

Both answers get you a better product. Only one of them gets sold to you elsewhere.

AI earns its place

Messy human input — free text, speech, photos, documents with no fixed format
Judgment at volume — classifying, prioritising, summarising thousands of items
Rules nobody can write down — patterns that shift, exceptions that multiply
Work that blocks people — the step where everyone waits for one person

A normal algorithm wins

The answer must always be identical — invoices, pricing, compliance, payouts
The rule fits on one line — and hasn't changed in five years
Being wrong is expensive — and there's no human checking the output
It's there to impress — a demo feature nobody asked for
What I build

AI inside real products

Not chatbots bolted onto a homepage. Intelligence that removes steps people didn't want to do anyway.

Understanding what comes in

Messages, documents, tickets, photos and speech turned into structured data — classified, prioritised and routed without anyone reading them first.

Deciding and predicting

Recommendations, forecasts, anomaly detection and smart notifications — so the product acts before the user has to ask.

Doing the work

Drafting answers, filling systems, moving tasks between tools. Automation with a human in the loop wherever the cost of being wrong is real.

The user shouldn't see a model. They should see a product that suddenly asks fewer questions.

How I approach it

Four steps, in this order

The interface comes last. That's deliberate.

1

The system

What the product actually does, where the work piles up, and which step is worth removing first.

2

The data

What flows in, what it's allowed to touch, where it's stored — and what happens on a bad day.

3

The rules

Access, confidentiality, audit trails, human oversight. In the EU this isn't optional, and it shapes the design.

4

The interface

Minimum controls, maximum intelligence behind them. Understandable in five seconds.

EUROPEAN BY DEFAULT.

Hosted in the EU, built with GDPR and the EU AI Act in mind from the first architecture sketch. Sensitive data stays where it belongs, oversight is designed in, and what the system decided is always traceable.

Your idea, your data, your product — that order matters.

Questions

Before you ask

The short version. Everything else — just ask.

We have an idea for an AI feature. Where do we start?

With a prototype of the one part that matters most. You'll know in weeks whether the intelligence actually helps — before committing to the full product.

What if AI turns out to be the wrong tool?

Then I'll say so, and we build it a simpler way. Talking you out of an unnecessary model is part of the job, not a lost sale.

How accurate will it be?

Depends on the task and the data — and I won't promise a number before measuring it. What I will design from day one is what happens when it's wrong.

Where does our data go?

EU hosting by default, with the sensitive paths designed so nothing unnecessary leaves your control. We agree on the boundaries before anything is built.

AI WHERE IT MAKES SENSE.

Tell me the problem. I'll tell you honestly whether a model is the answer.

TELL ME YOUR IDEA

No polished brief required. Don't let the budget stop the conversation.