We give each department one page: what to use AI for, what never to, and exactly how. And we write the policy that permits it.
It is that they have too much of it, none of it about their job, and no permission to act on any of it.
People leave knowing what AI is. They still do not know what to do with it on Monday, in their own role, with their own files.
Around 7.5% of employees report having received substantial AI training — while heavy users report more burnout, not less.
A blanket prohibition makes necessary work non-compliant. So the work happens anyway, unsupervised, on personal accounts.
Roughly two-thirds of professionals have used AI at work believing it was not permitted. About a third of organisations have any AI policy at all.
Every new assistant adds a login, a habit and a decision. Past a certain point, adding tools makes teams slower.
Productivity rises from one AI tool to two, slows at three, and declines beyond that (BCG / UC Riverside, 2026).
Buy one, or buy a package. Fixed price, fixed scope, agreed before we start. Prices shown are nonprofit rates.
“Where do we actually stand, and what do we fix first?”
A scored report across ten dimensions — fluency, governance, data protection, workforce impact, internal politics, psychological safety, future-readiness — with a ranked 90-day priority list, mapped to ISO/IEC 42001, the NIST AI Risk Management Framework and EU AI Act Article 4.
“What are our people allowed to do, and how?”
A board-ready policy that starts with permissions, not prohibitions: permitted uses by role, the red lines, a data classification table, disclosure rules for funders and beneficiaries, an escalation path and an approved tool list. Written so that required work stays compliant.
“What do I do with AI in my job, on Monday?”
One organisation-wide session on the shared ground rules, then a 90-minute lab per department built on that department's real tasks and real documents. Each lab ends with that team's one-page cheatsheet — see a sample.
“How do we stop this decaying in six months?”
Monthly office hours for whoever is stuck, cheatsheet updates as the tools change, coaching for your internal champions, and a re-score at six months so you can show the board what moved.
Delegation, Description, Discernment and Diligence — the AI Fluency framework published by Anthropic with Rick Dakan and Joseph Feller. It is public, so you can check our work.
What to hand to AI, what to keep human, and who decides.
How to ask, with your organisation's context already in the request.
How to check an output before anyone relies on it.
What to log, what to disclose, and who owns the result.
These are the four axes of your assessment score, the four sections of every team's cheatsheet, the four clause groups of your policy, and the four modules of every session. One vocabulary — so the diagnosis, the training and the rules all say the same thing, and nobody has to translate between them.
Fixed price, fixed scope, agreed before we start. Nonprofit rates shown; rates for private-sector and international organisations differ.
Prices exclude VAT and travel. Payment is 40% on signature and 60% on delivery. Scope changes of more than 20% are re-quoted rather than absorbed quietly.
You should use them. Free introductory courses on this framework exist, published by the people who wrote it. Subsidised cohort programmes exist for smaller nonprofits. If one of those is the right first step for you, we will say so on the call.
Then come to us for the four things a course or a cohort structurally cannot do: write your policy, use your team's actual tasks and documents, clear your own data rules, and produce the evidence a regulator or a donor asks for. A curriculum taught to two hundred organisations at once cannot tell your finance officer which tool is approved for which data, in your organisation.
Or that you do not need us yet. We will tell you that too.