AI consultancy · Brussels · ULB spin-off since 2005
Mentis helps organisations in regulated sectors across Belgium and Luxembourg put AI to work: readiness assessments, sovereign on-premise architectures, and generative AI grounded in your own documents — built by a founder-led team of AI researchers who have shipped production systems since 2005.
Start here
No account, no sales pitch — pick the depth that suits you.
18 questions, 4 minutes, instant maturity radar across six dimensions — strategy, data, architecture, AI Act governance, people, security. Take the scorecard →
Live agents tuned to finance, legal, healthcare, public sector, logistics and professional services — use cases, regulation, sovereignty, in your language. Start a conversation →
The 4-week, fixed-price diagnostic: board-ready maturity report, scored opportunity portfolio, AI Act exposure register, 90-day plan. See the offering →
What we do
Strategy, compliance, data science, engineering and operations — one senior team, end to end. The AI Readiness Assessment above is where most engagements start.
Risk classification of every AI system you run or plan, an exposure register, pre-filled documentation and a governance charter — by an Independent AI Expert for the European Commission. Learn more →
LLM assistants, document automation and retrieval-augmented generation over your own knowledge — with citations, access control and evaluation, on-premise or EU-hosted. Learn more →
Forecasting, risk scoring, NLP and computer vision trained on your data, benchmarked against your metric, explainable enough for your compliance team. Learn more →
Reference architectures for on-premise LLM inference and EU-hosted stacks; pipelines, monitoring and retraining so models keep working after launch. Learn more →
Route, scheduling, load and resource optimization modelled on your real constraints — savings in kilometres, hours and service levels, measured on your data. Learn more →
We design and build custom AI applications end to end — document pipelines, APIs, integrations and interfaces — secure, tested, documented and handed over with the code. Learn more →
Industries
Each industry page comes with a live AI advisor tuned to its use cases and rules.
Why Mentis
Our founder advises the Commission on AI. When we classify your systems under the AI Act, it is not a reading of the text — it is practice.
A spin-off of the Université libre de Bruxelles. Twenty years later the scientific habits remain: measure first, then decide — and say "not yet" when the data says so.
On-premise LLM inference and EU-hosted architectures are our default, not an exception — for clients who cannot send data to US clouds.
KnowledgeRAG and OptimizeFlow run in production. Operating them keeps our consulting honest about what works.
Nine senior people, founder-led. The person who scopes your work is the person who delivers it — at a fixed price, not a bench of juniors billed by the day.
Most engagements start with the AI Readiness Assessment or a scoped pilot — a clear success metric, your real data, results in weeks.
We deploy on your cloud or on-premise, work within GDPR and the AI Act from day one, and hand over the code.
Every model comes with monitoring, documentation and a retraining path — the unglamorous parts that make it last.
You work with the people who build it. Support, iteration and knowledge transfer to your team for as long as you need.
Track record
A sample of the domains our team has worked in over twenty years. See the case studies →
Scheduling technicians and vehicles under time windows and skills constraints — the work that became our OptimizeFlow product.
Assistants that answer from your own documents with citations, powered by KnowledgeRAG. Fewer hallucinations, traceable sources.
Storage, search and analysis of large genetic datasets for research and healthcare teams.
Models that attribute outcomes to campaigns and channels so marketing budgets go where they work.
Early-warning scoring of company insolvency risk from financial statements and market signals.
Demand-aware price recommendations that balance margin and competitiveness.
Failure prediction from sensor and maintenance-history data, so interventions are planned instead of emergencies.
We build with
Open source / open technology
Enterprise providers
Technologies we work with, chosen per project — not a list of partnerships or certifications.
See the full stack →Hands-on tutorials and opinions from our team on NLP, deep learning and where AI is heading.
The most common reason an AI idea dies is a sentence: "our data isn’t ready." It is usually half true — and the half that is true is smaller and cheaper to fix than people fear. The second dimension of AI readiness: what "ready" means, per use case, and why your documents may be your biggest asset.
Boards want AI, vendors sell AI, pilots multiply — and nothing reaches production. The first of six dimensions of AI readiness explains why, and what to fix first: a scored use-case portfolio, an owner, and a decision on what AI is actually for.
Understanding How CNNs Can Tackle the Challenges of NLP and Document Classification.
How Brain Architecture, Data, and Compute Converge to Shape the Future of Artificial General Intelligence.
A 30-minute readiness debrief gives you an honest read on feasibility, data, compliance exposure and a first step — usually within the week.
Mohamed Ben Haddou
Founder & CEO · Independent AI Expert for the European Commission
You talk to Mohamed, not a sales team — and the person who scopes your work is the one who delivers it.
