How the assessment is scored

Fixed, and the same for everyone. They are not weighted: the index is the plain mean of the five, so no dimension quietly counts more than another.

Awareness

Knowing what AI can and cannot do.

Instruction

Giving AI clear, effective direction.

Verification

Critically evaluating AI output before acting on it.

Governance

Handling data, disclosure, and risk responsibly.

Integration

Building AI into real work that lasts.

Every dimension is scored one to four.

Each option in each question is mapped to one of the four levels. You pick an option, that option carries a level, and your score for a dimension is the mean of the three answers you gave for it. Three questions per dimension means a score lands on thirds rather than whole numbers, which is what lets it tell a 3.0 from a 3.3.

The mapping lives on the server and the browser only ever sends the option you chose, so the answer key is not readable in the page.

If any single dimension comes out at Unaware, the band next to your score is held at Exploring however high the mean is. The number is never altered, only the word.

An average hides its worst component. Someone Fluent on four dimensions and Unaware on Verification averages 3.4 and would otherwise read as broadly competent while being unable to tell whether AI output is true. That is the exact failure this measure exists to catch, so it is not averaged away.

• No percentile until a defined reference population exists. None exists yet, so no result shows a percentile or a “top X%” label. A percentile against whoever happened to take the assessment would describe early adopters, not a population. When a reference population is defined, this page will say who it is and how it was sampled.
• No participation label. With no percentile, there is no standing label of any kind.
• No movement claim on a thin round. Two measurement rounds are compared only when each has at least 5 responses on the same version of the assessment. Beyond that, a department shows movement only with at least 8 people measured in both rounds, and the organization figure needs 20. Below those, the arrow is not drawn at all rather than drawn faintly, and the page says which condition was not met.
• No arrow for a movement smaller than 0.2. An arrow is a claim that something changed, and it is not printed for a difference this small. That floor is a convention, not a derived bound: the standard deviation of this instrument is not established, so nobody can currently compute the point at which a difference becomes real. It is set where a claim stops being obviously unsupportable, and it should be revisited once reliability exists.
• No group results below 5 people. Leadership never sees an individual’s result. A department — or any other group — shows its results only once that many of its people have been measured, on the dashboard and in the evidence pack alike, and a larger group is withheld too when showing it would let a smaller one be worked out from the totals. This floor protects the people in small groups; it is not a claim about when a mean becomes statistically stable, which cannot be computed until the instrument’s reliability is established.
• No superlative on a tie. If several dimensions are equally weak, the result says so rather than naming one.
• Training follows the two weakest dimensions below Applied. For a department or the organization, the leadership view names the two dimensions furthest below Applied; one, when only one is below; nothing, when none is, and training still continues each quarter on the team’s two lowest dimensions, at the Applied level, working toward Fluent. Where a tie leaves the second pick unclear, it says so and leaders choose. Leaders also choose who is in each training group: the dashboard shows which teams are furthest behind and never picks people.
• This is not a credential. It is a rubric score and a description of a profile shape. It certifies nothing.

The archetype is interpretive. It is a description of the shape of a profile, not a measurement, and not a category anything has validated. It exists because two people who both average 3.0 can be very different practitioners, and a single number hides that.

It is assigned by a fixed set of rules, tested in order, from the five dimension scores alone. Nothing else feeds it — not the self-rating, not who you are, not how long you took. The rules that decide it are:

1. An average below 1.8 is The Onlooker, before any shape is read.
2. Verification at or below 2 while Instruction and Integration average 3 or more is The Fast Mover. Operating faster than you are checking is tested first because it is the shape that costs most.
3. Then, in order: heavy checking with little doing, strength across the board, personal wins that have not spread, capability outside the guardrails, and judgment ahead of practice.
4. If none of those fit, an average below 2.4 is The Experimenter.
5. Above that, the shape decides it. One dimension a full level or more clear of the next, at 3 or higher, is The Specialist. Otherwise, a gap of two levels or more between your highest and lowest dimension is The Patchwork, and a gap of less than two levels is The Generalist.

The thresholds are ours. They were set by judgment about which shapes are worth telling apart, and they have not been validated against anything external. They can and do change: the catch-all case once took 48% of results, which meant half the archetypes said the same thing, and it was split apart. The exact rules are in src/lib/archetype.ts, and the instrument document sets out what is and is not established about the scores underneath them.

Practitioner matching starts with your lowest-scoring dimension. Two things decide who appears:

• The dimension you scored lowest. Consultants are tagged by hand for the dimensions they work in, and only those tagged for your lowest dimension are eligible. Where several dimensions tie at the bottom, Verification is preferred, because it is the one where being wrong is most expensive.
• Rotation across everyone eligible. Which of them you see is fixed by your result id, so it does not reshuffle when you reload, but it differs between people. Demand spreads across the whole pool instead of collecting on whoever happens to sort first.

Anyone missing a photo, a description or a booking link is not shown at all, so a card is never a placeholder. Nobody is ranked by what they pay.

Disclosure. AIFluency operates The AI Fluency Coaching Hub and earns a fee on bookings made through it. We both measure the gap and sell help closing it, and you should read the recommendation knowing that.

This is a self-assessment of judgment across five dimensions. It is not a credential, and it has not been validated for use in selection, promotion, compensation or termination decisions. It has no established predictive validity for job performance, and the level mappings are ours rather than an external standard.

Use it to decide what to work on, and to see where a team is thin. Do not use it to decide who to hire, who to promote, who to pay more, or who to let go.

The level mappings are set by AIFluency, informed by how each dimension plays out in real work. They are published rather than hidden so you can judge them yourself.

None of the fifteen has been reviewed independently. Every level assignment is a considered first pass by the author and has not been keyed by an outside practitioner. Until that happens, no single result is used to make a claim about any individual.