Methodology
How the Stay Useful Score is calculated, and what it cannot tell you
This page describes exactly how your result is produced. If any part of it does not match what you saw, treat that as a defect and tell us.
What this is
The Stay Useful Score is a structured self-assessment of practical AI preparedness. It reflects the answers you select at one point in time.
It is deterministic. The same answers always produce the same score, stage, gap, and recommendation. There is no model, no randomness, and no personalised generation involved in your result.
What this is not
- A psychometric instrument.
- An objective skills examination.
- An employment-screening tool.
- A prediction of redundancy.
- A certificate of AI competence.
- A measure of intelligence, potential, or personal worth.
What the 17 questions do
- Questions 1 and 2
- Establish your context. They are never scored and can never change any number in your result.
- Questions 3 to 5
- Measure AI Exposure, which is reported separately as context and never enters your Overall Score.
- Questions 6 to 8
- Measure AI Fluency.
- Questions 9 to 11
- Measure AI Leverage.
- Questions 12 to 14
- Measure AI Proof.
- Questions 15 to 17
- Help us personalise the supporting explanation and improve the product. They are never scored and can never change your diagnosis.
How the numbers are produced
Each scored answer is worth 0, 33, 67, or 100 points. The options are ordered from least to most developed practice.
Each dimension is a weighted average of its three questions. The Overall Score is a weighted average of Fluency, Leverage, and Proof only.
Calculations use full precision internally, and every displayed figure is rounded to the nearest whole number. Your stage, your labels, and your biggest gap are all read from the numbers you can see, so a displayed number and its label can never contradict each other.
The weights
| Dimension | Question weights |
|---|---|
| AI Exposure | Question 3 at 40 percent, Question 4 at 35 percent, Question 5 at 25 percent |
| AI Fluency | Question 6 at 40 percent, Question 7 at 30 percent, Question 8 at 30 percent |
| AI Leverage | Question 9 at 30 percent, Question 10 at 40 percent, Question 11 at 30 percent |
| AI Proof | Question 12 at 40 percent, Question 13 at 30 percent, Question 14 at 30 percent |
The Overall Score is 35 percent Fluency, 40 percent Leverage, and 25 percent Proof. Exposure is not part of it.
The bands
| Displayed score | Stage |
|---|---|
| 0 to 29 | Starting Out |
| 30 to 49 | AI User |
| 50 to 69 | AI Operator |
| 70 to 84 | AI Builder |
| 85 to 100 | AI Multiplier |
| Displayed score | Exposure label |
|---|---|
| 0 to 24 | Low Exposure |
| 25 to 49 | Moderate Exposure |
| 50 to 74 | High Exposure |
| 75 to 100 | Very High Exposure |
| Displayed score | Dimension label |
|---|---|
| 0 to 24 | Starting |
| 25 to 49 | Developing |
| 50 to 74 | Capable |
| 75 to 100 | Strong |
Why Exposure is separate
Exposure describes the kinds of tasks in your context. It does not predict whether your job will disappear.
A high Exposure score does not mean your role is at risk, and a low one does not mean AI is irrelevant to you. Because it describes your context rather than your capability, including it in a capability score would be misleading, so it is excluded.
How your biggest gap is chosen
Your biggest gap is simply the lowest of your three displayed capability scores. Exposure is excluded.
If two or three dimensions are level, we name all of them rather than picking one, and the recommendations address each. Ties are never broken by rounding, by internal decimals, or by the order the dimensions happen to appear in.
How your recommendation is chosen
Your recommendation comes from a fixed set of rules applied in a fixed order, using only your displayed scores and your gap list. The first rule that matches is the one you get.
Your context, motivation, blocker, and desired outcome may change the examples we show you. They can never change your score, your stage, your gap, or which recommendation you receive.
Known limits
- It is self-reported. People who are new to a subject sometimes rate themselves generously, and people who know a subject well sometimes rate themselves harshly. The result is only as good as the honesty of the answers.
- It measures practice, not outcomes. A high score means you describe habits associated with reliable AI use. It does not verify that you have used AI well, and no result should be presented to anyone as verified capability.
- The bands are considered judgements rather than empirically validated cut-offs. They have not been through formal psychometric validation, and we do not claim they have.
- A low Proof score frequently reflects limited access, permission, or opportunity rather than limited ability.
- It is a snapshot. If your work changes, your answers should change with it.
What happens to your answers
While you are answering, your progress is stored only in your own browser, and it expires after 24 hours. Nothing is sent to us until you complete the assessment.
If you ask for the emailed report, we use your email address to send it. That is a separate choice from marketing, which is optional, unchecked by default, and never required for your result, your report, or support.
Your answers are never sent to an AI model. Your result is calculated by fixed arithmetic.
The optional free-text question is never required, and you should not put confidential, personal, client, patient, student, financial, health, or employer-restricted information into it.
Version
Every result records the assessment, scoring, and diagnosis versions used to produce it. When those change, existing results are not silently recalculated, because a score produced under one version is not comparable to a score produced under another.
Assessment version 1.1.0. Scoring version 1.0.0. Diagnosis version 1.1.0.