The whole engagement is four weeks and it is printed below. Nothing to install, no integration work. We need a headcount list with role bands and an internal sponsor who can send two emails. If it does not fit, you will know in two minutes rather than two meetings.
Which role bands exist, which carry oversight duties, and what AI is actually in use.
Twelve to eighteen minutes per person, on any device, in work time. Reminders at day four and eight.
Scored across five constructs, banded by role. Segments below five respondents suppressed.
A ninety minute session with your sponsor and whoever owns compliance, then the file, dated and versioned.
Nothing to install and no integration work. We need a headcount list with role bands and an internal sponsor who can send two emails.
| Deliverable | What it is for |
|---|---|
| Evidence file | The record. Who was assessed, when, against what, and what the results were. Structured to be produced without assembling it under pressure. |
| Measures rationale | A written statement of the measures you adopted and why they are proportionate for an organization of your size, sector and AI usage. Under an obligation of effort, this is the document that carries the most weight. |
| Role band report | Results per band across all five constructs, with the calibration gap surfaced separately. Where capability is weakest, and for whom. |
| Benchmark position | Position against organizations of comparable size and sector, with the sample size behind every comparison stated plainly. Our benchmark set is still small, so early comparisons are reported as indicative and labeled that way. |
| Sentiment summary | Trust, anxiety, and perceived impact, themed from free text. Reported separately from literacy and never blended into it. |
| Assignment plan | Which people receive which training, derived from their own results rather than a blanket rollout. |
Our foundation program is eight modules plus a short welcome, about ninety minutes end to end, included with every engagement above the baseline assessment. It is short, plain-English, and built for people who have to work alongside AI rather than build it.
| Module | Covers | Time |
|---|---|---|
| Welcome and Overview | Why this matters and how the program works. Not numbered, and not assigned by result | 5 min |
| 01 The 5-Layer AI Cake | How modern AI is built, from the chips up | 12 min |
| 02 Large Language Models | What a model actually does, and why it can sound confident and still be wrong | 11 min |
| 03 The 5 AI Races | Talent, compute, models, adoption, sovereignty | 11 min |
| 04 AI Governance | Policy, integrity, data, and who owns the decisions | 13 min |
| 05 AI in the Workplace | How AI is reshaping sales, marketing, and internal operations | 12 min |
| 06 Agents and Workflows | Multi-step work beyond chatbots, and what it unlocks | 10 min |
| 07 The Future of Work | Durable shifts in jobs and skills, sorted from hype | 9 min |
| 08 Humanoid Robots and Physical AI | Where this is heading once it leaves the screen | 8 min |
We publish this mapping because it is the honest version. The foundation program covers understanding thoroughly. Judgment and oversight need practice, not explanation, so those are delivered as short role-specific modules assigned only to the people whose results show they need them.
| Construct | Covered by |
|---|---|
| C1 Comprehension | Modules 01, 02, 06, 08. This is the program's core strength |
| C2 Applied judgment | Module 02 introduces it. Role-specific practice modules carry it |
| C3 Boundaries and disclosure | Module 04, extended with your own policy where you have one |
| C4 Oversight capability | Role-specific only. Assigned to people who review or can override AI-informed decisions |
| C5 Calibration | Not a training target. Calibration is measured, and it is what tells us who needs which of the above |
Article 4 asks for measures proportionate to role, context, and existing knowledge, and since the July 2026 amendment it is the measures themselves that are assessed rather than any level your people reach. An optional course library evidences that you made something available. An assignment derived from a measured gap evidences that you identified what each group needed and acted on it. Those two things look similar on a slide and completely different in a file.
So the foundation program is assigned by result, not pushed to everyone by default. Someone whose comprehension already scores well skips most of it and receives the judgment material instead. Someone in an oversight role receives content the general population does not.
Priced per engagement rather than per seat, because a literacy baseline is a piece of work with a beginning and an end, not a platform subscription. Bands are by assessed headcount, and the tiers differ by what you get rather than only by how many people you have.
Legal advice. We are not a law firm and we do not opine on whether you are compliant. Where an engagement raises questions that need counsel, we say so and will work alongside yours.
| Item | Rate |
|---|---|
| Headcount above your band | $18 per assessed person |
| Additional languages beyond English (Baseline only, included in Program and above) | $1,800 per language per assessment |
| Extra re-measurement outside your schedule | $4,500 |
| Governance and policy workshop | $7,500 |
Only if they are confident the answers cannot be traced back, which is why our anonymity controls are structural rather than promised. No segment under five respondents is ever reported, and individual responses are never disclosed to you. That is set out in full on the methodology page.
No, and it should not try. This measures a specific capability and a specific set of attitudes toward AI. It sits alongside whatever listening platform you already run.
Possibly more than you think, since scope follows where AI output is used. Run the assessment and it will tell you in three minutes. If it says you are out of scope, believe it.
Then you have a dated baseline and a defensible plan, which is a materially better position than having nothing measured at all. Article 4 does not require your people to reach any particular level, so a low first score with documented, proportionate action reads far better than an absence of evidence.
No. It softened the standard and deferred the high-risk obligations, but the literacy duty survived a proposal to weaken it much further and remains binding on every provider and deployer. The Article 4 page sets out what changed and what did not.
Eight questions. It will tell you whether Article 4 applies to you and what your evidence position looks like today, including if the answer is that you do not need us yet.