M.A.I. Consulting
About the practice

A boutique built for one job.

M.A.I. Consulting makes organisations AI-fluent for the work they do. Five services, one framework, fixed prices agreed in writing. We do not build AI systems, and we do not sell anybody’s software.

What the practice does

We work on the people who use AI.

Most AI failure in an organisation is not technical. Tools get bought, a generic session gets delivered, nothing changes in the work, and a policy written to eliminate risk pushes real use into private accounts. We address that failure and nothing else. Every service starts from it and ends with something the organisation can show.

Diagnose before training anybody

A scored baseline across ten dimensions tells you where the gap is. Training a department that is already fluent, while another is uncertain, is the most common way this budget is wasted.

Write the permissions first

A policy that only forbids makes necessary work non-compliant, so the work happens anyway and the organisation carries the risk without seeing it. Ours says what is allowed, with which data, by whom.

A playbook for each team, built on their own tasks

A working playbook built from that department's own tasks and documents, holding what to delegate, how to ask, what to check and what to log. It is what the team works from afterwards, and it is neither a course nor a curriculum.

Leave behind a file you own

The score, the policy, the playbooks and the attendance record are yours. They are also the documentation a donor, an auditor or a board asks to see.

The method

One vocabulary across every deliverable.

Delegation, Description, Discernment and Diligence: the AI Fluency framework published by Anthropic with Rick Dakan and Joseph Feller. We use a public framework rather than a proprietary one so that you can audit the method instead of taking it on trust.

Delegation

What to hand to AI, what to keep human, and who decides.

Description

How to ask, with your organisation's context already in the request.

Discernment

How to check an output before anyone relies on it.

Diligence

What to log, what to disclose, and who owns the result.

The same four words are the axes of your score, the sections of every playbook, the clause groups of your policy and the modules of every session. No translation is needed between the diagnosis, the training and the rules.

What the work stands on

Public frameworks, recognised standards, and our own instrument.

Certified in this material

The practice is certified across the AI fluency, responsible-AI and applied-tooling curricula of Anthropic and of the United Nations College with Microsoft, including the qualifications for teaching AI fluency as well as using it. Member of the Anthropic Claude Partner Network.

Mapped to the standards buyers are asked about

ISO/IEC 42001:2023 · NIST AI Risk Management Framework 1.0 · EU AI Act (Regulation (EU) 2024/1689) · GDPR · OECD AI Principles · UNESCO Recommendation on the Ethics of AI.

A diagnostic we built ourselves

Ten dimensions, sixty scored questions, 300 points normalised to 0–100. Published free on this site rather than kept behind a proposal, because a maturity model you cannot inspect cannot be checked by the people it scores.

Open the extended assessment

How engagements run

Fixed price, fixed scope, agreed before we start.

Priced up front

You see an indicative range in the estimator before we speak, and a fixed number in writing after a 30-minute call. Scope changes of more than 20% are re-quoted. Payment is 40% on signature, 60% on delivery.

Three to twelve weeks

Most engagements run three to eight weeks. Support retainers run three to twelve months and are billed monthly.

English, French or Arabic

Sessions and materials in any of the three. Arabic material is written in Arabic from the first draft rather than translated.

Geneva-based, delivered anywhere

Remote or on-site. Courses for individuals and organisation representatives, fewer than ten participants, run in person or online; in-person courses include accommodation and hospitality arrangements. Travel outside Geneva is billed at cost.

Limits

What this practice is not.

It is not a large firm. There is no bench and no 24-hour turnaround. Capacity is finite and engagements are scheduled accordingly. If you need thirty trainers in four countries next month, we are the wrong call.

It does not build your AI systems. No model fine-tuning, no data pipelines, no custom software. We do configure agents and tools on the platforms your teams already use, and that is where it stops. All five services are about the people who use AI.

It does not do compliance work. No regulatory compliance services, no certification, no legal advice. Our documents use the vocabulary of recognised frameworks so that they are legible to the people who ask about them, and whether that satisfies a particular legal obligation is a question for your own counsel. Where compliance is what you need, we will say so on the call rather than take the work.

It sells no software. No licences resold, no partner commissions, no quota. That is what makes “stay on the free tier” something we can say to you.

It will point you elsewhere when that is right. Free introductory courses on this framework exist, published by the people who wrote it. Subsidised programmes exist for smaller nonprofits. Where one of those is the right first step, the discovery call will say so.

It is early. Founded in 2023, and the client list is short. What you can inspect before committing is the instrument, the framework, the sample deliverable and the estimator, which is more than most consultancies put in front of you before a proposal.

Thirty minutes is enough to establish whether this fits.

We will tell you which of the five services you need, or that you do not need us yet.