AI readiness assessment
The AI readiness assessment, in five dimensions.
A two-week diagnostic that scores where an organization actually stands — and names the first project, which is almost never a model.
What is an AI readiness assessment?
An AI readiness assessment is a structured diagnostic of whether an organization can put AI into production and keep it there — scored across data foundations, workflow clarity, decision rights and governance, capability density, and adoption and trust posture. It deliberately does not assess models. Pilots rarely die of model quality; they die because the data describing the workflow is wrong, because nobody can say what the process actually is, or because the frontline has scar tissue from the last three rollouts. The output is a sequence, not a grade: the lowest dimension is the first project.
The method in full is in The AI Readiness Score: How I Assess a Product Org in Two Weeks.
What gets scored, and what the scores mean.
Each dimension runs 1–5. The anchors below describe a 2 and a 4, because those are the two states teams can actually recognize in themselves.
| Dimension | The question | Scores 2 | Scores 4 |
|---|---|---|---|
| Data foundations | Is the data describing your core workflows accessible, accurate, and fresh enough to act on? | Any operational question needs an analyst and a week. | Systems of record agree, every critical dataset has an owner, a new integration is a task not a project. |
| Workflow clarity | Can you describe the process you intend to automate, with numbers? | The process lives in three tenured heads and every walkthrough draws a different diagram. | Volumes, exception rates and handoffs are known: “step four is 30% of cycle time.” |
| Decision rights and governance | When the system is wrong with consequences, who finds out and who fixes it? | Governance by escalation: every incident convenes an ad-hoc meeting of everyone. | Quality bars set in advance, an owner per deployed system, a written path from wrong to fixed. |
| Capability density | Can the teams closest to the work read an eval and reason about failure modes? | All AI knowledge sits in one team everything else queues behind. | The median product team has shipped and operated something probabilistic. |
| Adoption and trust posture | What happened to the last three tools rolled out to the frontline? | A workforce with scar tissue from transformation programs past. | Frontline teams have visibly shaped a rollout and trust that a fallback exists. |
AI lands on whatever trust surface already exists. It does not create one.
How the two weeks run.
Deliberately unglamorous. Evidence first, interviews second, scores last.
- Days 1–3 · Read what already exists. Roadmaps, incident channels, ops dashboards, and the last three pilot post-mortems — whose absence is itself a data point.
- Days 4–7 · Walk the top two workflows. End to end, sitting with the people who run them, counting volumes, exceptions, and handoffs rather than accepting the diagram.
- Days 8–10 · Interview the leadership seam. Product, engineering, ops, finance, and frontline management — structured, comparable, and aimed at the seams where accountability changes hands.
- Days 11–12 · Score from evidence. Each dimension 1–5, anchored to what was observed. Self-assessment inflates by about a point; the artifacts and the walk are the correction.
- Days 13–14 · Sequence the work. The lowest dimension becomes the first project, with what has to be true to move it, and an explicit list of AI work that should wait until it does.
Want the shape of the answer before committing two weeks? The Lab hosts a free self-scored version of the same five dimensions.
Diagnosis, then the right practice.
If the first project is a leadership or operating-model problem, that runs through ProductExec — often as fractional CPO work. If it is an implementation problem, it runs through Enterprise AI Studio. If it turns out to be a monetization problem, start at AI SaaS pricing.
Questions this raises
What is measured in an AI readiness assessment?
Five dimensions, each scored 1–5 from evidence: data foundations, workflow clarity, decision rights and governance, capability density, and adoption and trust posture. Model quality is deliberately not one of them — it is rarely the binding constraint, and the lowest of the five dimensions is what decides whether anything reaches production.
How long does an AI readiness assessment take?
Two weeks. Week one reads the artifacts the organization already produces and walks the top two workflows end to end with the people who run them. Week two runs structured interviews across the leadership seam — product, engineering, ops, finance, frontline management — and scores each dimension from evidence rather than self-report.
Is a self-assessment good enough?
It is a good start and a poor finish. Self-assessment inflates every dimension by roughly a point, consistently and in good faith — people score the process they believe exists. The free scorer in the Lab is built for exactly that first pass; the artifacts and the workflow walk are the correction.
What do you get at the end of an AI readiness assessment?
A scored profile across the five dimensions, the evidence behind each score, and a sequence — because the real output is not the number. Your lowest dimension is your first project, and it is almost never a model. The deliverable names that first project, what would have to be true to move the score, and which AI work should wait until it does.
Find out which dimension is holding you back.
Score yourself in the Lab in five minutes, or bring the real thing and we will read the evidence together.