Haven't taken the check yet? Do that first — 14 questions, about four minutes. It scores you on the three areas below and tells you which one to fix first. Already have your result? Jump to your weakest area: AI Usage · AI Governance · AI Learning.

01The model

Three things decide whether AI actually sticks in a team. They're not separate projects — inside a real team they're one problem, and a weakness in any one of them stalls the other two.

01AI Usage — where it helps, where it doesn't, and how people work with it day to day.
02AI Governance — a policy a team can actually adopt, not one that sits in a folder.
03AI Learning — training pitched at the people who have to use it, technical or not.

The check gives you a score for each and a combined grade. The work is almost always the same shape: find the weakest of the three and fix that one first. Doing all three at once, or doing the one you find most comfortable, is how programmes stall.

02AI Usage

Usage is whether AI has actually changed how the work gets done — not whether people have tried it. The test is simple: name a real piece of weekly work that runs differently now because of AI. If you can't, usage is still at the demo stage.

What good looks like
  • People know where AI earns its keep — and where it wastes time or misleads.
  • AI output is checked before it's used or sent. It's a habit, not a policy.
  • There's a known, supported set of tools — not everyone on whatever they found.
  • Real workflows run differently now. Not demos — the actual work.
Common failure
  • Enthusiast silos — a few power users, no spread, no shared practice.
  • Unchecked output — AI text going straight out, with the risk that carries.
  • Tool sprawl — everyone on a different tool, data going in all directions.
Your first move
If you're just starting

Pick one team and one workflow — a task people do every week. Work through where AI helps and where it doesn't, and make "check before you send" the rule. Adoption spreads from a visible win, not a mandate.

If it's partly there

You have enthusiasts — now make their practice legible. Capture what's working, agree a recommended tool set, and give the occasional users a reason and a route to go deeper.

If it's already strong

Protect it. Keep the recommended tools current, watch for over-reliance on unchecked output, and feed what you learn back into how new people are brought up to speed.

Go deeper

  • Write a one-page "where AI helps here" list, specific to your work — and a matching "where it has burned us" list. Keep both live.
  • Make review visible: a quick "AI-assisted, checked by [name]" note on anything that goes to a client or into a decision.
  • Name the supported tool set and who pays for it, so people stop quietly expensing their own.
🛠️

PromptForge puts structured prompting into everyday practice — 12 frameworks, auto-suggested, local-only. It's the difference between people guessing at prompts and having a repeatable way in. Free public beta.

03AI Governance

Governance is the part everyone agrees is important and few have actually done. The bar is not a legal document — it's a policy a team will read and follow, and one clear rule about data that is enforced rather than trusted.

What good looks like
  • One page people will actually read — not a twelve-page document nobody opens.
  • A hard line on what data never goes into a public tool — enforced with tooling, not trust.
  • A named owner with a mandate and a budget.
  • Near-misses logged and reviewed, so the policy improves from evidence.
Common failure
  • Policy theatre — a document exists; behaviour hasn't changed.
  • No owner — governance is "IT's problem", informally, on top of everything else.
  • Silent incidents — mistakes happen but aren't recorded, so nothing improves.
Your first move
If you're just starting

You're carrying real exposure with no controls. Write a one-page usage policy people will actually read, and set a hard line on what data never goes into a public tool — backed by a tool so it isn't just trust.

If it's partly there

A policy exists but isn't living. Give governance a named owner, make the "never share" rule enforceable rather than advisory, and start logging near-misses so the policy improves from evidence.

If it's already strong

Keep it current as tools and regulation move, rehearse the "how do you govern AI?" answer for clients, and make sure the owner has the mandate to say no when needed.

Go deeper

  • Draft the policy as a single screen: what's fine, what's never OK, what to do when unsure, who to ask. If it needs a second page, cut.
  • Define the data red line in concrete terms — name the systems and record types, not "confidential information".
  • Open a no-blame near-miss log. The goal is a short monthly review that changes one line of the policy.
🛡️

Guardian AI sits directly under this: it enforces the data rule at the browser before a prompt is sent, and doubles as a critical-thinking checkpoint — did a human actually think about this? Free public beta.

04AI Learning

Learning is what stops AI capability from living in three people's heads. Self-taught doesn't scale and it quietly leaves non-technical staff behind. The bar is training built around the actual work, and an active habit of sharing what works.

What good looks like
  • Built around real tasks and real prompts — not a generic "intro to AI".
  • Role-relevant. What a finance lead needs is not what an engineer needs.
  • An active practice of sharing what works — prompts, patterns, cautionary tales.
  • Non-technical staff genuinely included. AI is everyone's tool, not a tech thing.
Common failure
  • Generic training — an "intro to AI" that doesn't touch anyone's actual job.
  • One and done — a single session, then nothing, while the tools move monthly.
  • Left behind — AI framed as "a tech thing", so half the organisation opts out.
Your first move
If you're just starting

Run role-relevant training built around real tasks — not a generic "intro to AI" — and open a shared space for prompts and patterns so the learning outlives the session.

If it's partly there

Training has started but isn't sticking. Tie it to the work people actually do, make "sharing what works" a habit rather than an afterthought, and check that non-technical staff are genuinely included.

If it's already strong

Keep it fresh as the tools change, keep the sharing practice active, and consider formalising it into a short internal course so it survives staff turnover.

Go deeper

  • Replace the "intro to AI" with a 45-minute session per role built from three real tasks that team does.
  • Start a prompt library where entries are pasted from real use, with a line on what they're for. Kill the ones that stop working.
  • Pair one technical and one non-technical person to co-run the next session, so it doesn't read as a tech briefing.

05Do this in the next two weeks

  • Take (or retake) the readiness check and write down which of the three areas scored lowest.
  • For Usage: name one weekly workflow to run through with AI, and set "check before you send" as the rule for it.
  • For Governance: draft the one-page policy and the data red line. Name an owner — a real person, not a committee.
  • For Learning: book one role-relevant session built from real tasks, and open the shared prompt space.
  • Open a near-miss log and tell people it's no-blame. Put a 20-minute review on the calendar for four weeks out.
  • Pick the single weakest area and put a named owner and a date against its first move. Leave the other two as "next".

Still stuck?

If the weak area isn't moving — the policy keeps not landing, usage won't spread past the enthusiasts, training doesn't stick — that's the point to bring Jon in on that one problem. A short, hands-on look at how AI is actually used and governed here, then the specific move it needs: a workshop, a rollout, or a fractional hand on it.