Approach
Clarity for data, automation and AI.
We work with data, automation and applied AI under one principle: proportionality, control and clarity. No hype, no scaremongering, and without turning every project into a bureaucratic exercise.
Three principles
There is no method without clarity.
Before we talk about tools, we agree on how to decide. These are the three principles we apply on every project.
Proportionality.
The controls we add to a process are proportional to its impact. An internal classification is not documented like a decision that affects people. We avoid over-engineering and we avoid under-engineering.
Control.
Every automated step has a human owner, a review point and a clear way to stop it. AI adds capacity; it does not replace it.
Clarity.
Whoever uses the system understands what it does, with what data and under what rules. Whoever audits it — internally or externally — finds what they need without having to ask.
How we apply it
Concrete practices, not slogans.
What these principles mean in the day-to-day of a project.
Responsible AI for SMEs.
We design AI use cases around the real size of the problem: which decision it supports, with what data, what happens when it gets it wrong and who reviews it. No abstract frameworks — concrete criteria, written into the project itself.
Data governance without bureaucracy.
We agree the bare minimum: who owns each dataset, what quality is expected, how it is updated and when it is deleted. We are not after a data committee; we want every dataset to have a clear owner and a clear lifespan.
Traceability and minimal documentation.
Every automated flow records what came in, what went out and what it decided. We document just enough for the team to maintain, audit and change it without us. If it cannot be understood without us, it is not finished.
Human oversight.
On each flow we define where a human reviews: before deciding, before communicating, before acting. Review lives in the design, not as an add-on once something breaks.
Clear limits on generative AI.
We decide which tasks use generative AI, what data it may work with and which outputs require human validation. We also decide what is not done with it. These limits are written down and shared with the team.
Tools and vendors chosen with clear criteria.
We choose tools for how well they fit the case, not for hype. We weigh where data is processed, what dependencies we create and what happens if a vendor changes terms. Portability matters.
Automation
Automating is not delegating blind.
A good automated process is an understood process, not a hidden one. So before automating we define the input, the output, the owners, the limits, what counts as an error and how it stops. Then we measure: what it saves, what fails and when it stops being useful. Useful automation is the kind the team controls — not the kind that replaces the team while nobody reviews it.
European framework
AI Act: when it comes into play.
The AI Act is the European Union's regulation on artificial intelligence. It classifies AI systems by level of risk and sets different obligations depending on that level. Most SME projects fall into lighter categories — productivity, internal assistants, analysis — while others, depending on the use case, may fall into more demanding ones.
On each project we identify which category a use case might belong to and which good practices make sense to apply from the design stage: transparency, oversight, documentation, risk management. We do this as an engineering and management criterion, not as a legal opinion.
This page explains how we work; it does not constitute legal advice. For legal decisions on AI Act compliance or other regulation, we recommend specialist counsel.
If this approach fits, let's talk.
We start by understanding where the friction is and what is worth prioritising.