What the assessment checks
Most AI pilots that fail do not fail on the model. They fail because nobody owned the outcome, the data the workflow needed was out of reach, the tool was never approved for real work, or nobody recorded how long the work took beforehand, so nobody could say whether anything improved.
The twelve questions cover those failure points in six dimensions, two questions each:
- Sponsor and owner. A senior sponsor who will remove blockers and make the decision at the end, and a day-to-day owner with time set aside.
- Workflow clarity. One recurring piece of work with a clear output, and its steps and review points written down.
- Data and access. The inputs are reachable by the people doing the work, and you know which data may be used with which tools.
- Approved tools and policy. At least one tool is approved with accounts available, and people have written guidance they can follow.
- Baseline measurement. You know how long the work takes today and how its quality is checked.
- Team capacity and support. The pilot group has protected time, and someone collects what works and answers questions.
Each answer scores 0 (“Not yet”), 1 (“Partly”) or 2 (“Yes”). A dimension with full marks is in place; one with at least half marks is partly in place; below that it is a gap.
How to read your result
The overall band follows three rules, applied in order:
- If sponsor and owner is a gap, the band is Not yet: start with ownership, whatever else you scored.
- If no dimension is a gap and the total is at least 17 of 24, the band is Ready to pilot one workflow.
- Otherwise the band is Fix prerequisites first.
The next actions list every answer below “Yes”, with ownership first, then “Not yet” answers before “Partly” answers. Each action is specific to the question: a missing baseline produces “record time per task for two weeks before changing anything”, not a generic “improve measurement”.
Illustrative example
Illustrative example, not a client result. An operations director wants to pilot AI on drafting first replies to supplier queries. She answers “Partly” on workflow steps, data classification and protected time; “Not yet” on baseline time and quality checks; and “Yes” to the other seven questions.
The result: 17 of 24 points and five dimensions in place or partly in place, but baseline measurement is a gap, so the band is Fix prerequisites first. The top actions are to time a sample of real supplier replies for two weeks and to define one quality measure (for example, how many drafts the reviewer sends back). Next come the “Partly” items: complete the workflow map, including the review step, and write the data rules for supplier correspondence as allowed, allowed with review, and not allowed. None of this needs new spending. Two or three weeks later, re-running the assessment moves the band to “Ready to pilot one workflow”.
Assumptions and limitations
- It is a self-assessment. The rubric is published on this page and applied to your answers. Nobody verifies them. It is not an audit, a certification or a regulatory opinion.
- It does not predict results. A ready team can still run a pilot that shows no value. Being ready means the pilot can produce evidence either way.
- It assumes one workflow and one team. Answers about “the organisation” in general produce vague results.
- Weights are equal. Every question carries the same points, except that ownership overrides the band. In your context, data rules or approvals may matter more; read the dimension statuses as well as the band.
What to do next
If the result says ownership, settle that before anything else; no external party can supply it. If it says fix prerequisites, most of the actions are internal and cheap. Once you are ready to pilot, use the adoption measurement worksheet to set up the baseline, and the pilot scorecard to agree in advance what go, revise and stop will mean.
If you want an independent view, with someone examining candidate workflows, data access and owners inside your organisation, that is the paid AI readiness and implementation roadmap: a commissioned investigation with an evidence-backed shortlist and a pilot brief. If you are ready and want help running the pilot itself, see AI enablement. Unsure whether the gap is skills or the work itself? Try training versus enablement.