Exploratory tool · Before, during and after deployment
Assess and monitor an AI use through real work.
Document a specific use at T0, review observed facts at T1, then monitor all nine axis scores over time.
Local processing. Answers remain in this browser and are not transmitted by the questionnaire. On a shared device, remember to delete the draft.
01 / A continuous approach
Assess before. Test during. Review after.
The same use case, population and observations must remain identifiable throughout the project.
Make assumptions open to discussion.
Describe what could change in the work and what is still missing for a decision.
Open the assessment ↓Test without creating lock-in.
Observe ordinary cases, exceptions, review workload and the ability to reverse course.
Review the observed effects.
Revisit the nine dimensions after the pilot and over time, while recording changes in context.
Open follow-up ↓The questionnaire measures neither probability nor clinical severity. It helps identify situations to investigate with workers, their representatives and prevention professionals.
02 / Guided assessment
Examine a specific use case.
One introduction
Choose one use case.
Answer for the most exposed group and the reasonably most demanding situation. If information is missing, say so: do not estimate it.
- the summary score for each axis;
- insufficient or non-applicable information;
- the two source answers behind each score;
- the actions and decision entered by your group.
Your draft stays on this device. You can resume, export or delete it at any time. On a shared device, export and then delete it before leaving.
Import its JSON export to resume the same project without re-entering its context. New answers remain blank.
Step 1 of 4 · Frame
Describe the use case.
This information gives context to the report. Do not enter any identifiable data.
Dimension 1 of 9
To what extent does each statement describe the planned project?0 = not at all true · 4 = completely true. Use NE when information is missing rather than estimating.
Step 3 of 4 · Review
Review before producing the report.
Review answers, missing information and contextual notes.
Step 4 of 4 · Act
Discussion priorities before the pilot.
This profile describes the answers entered. It does not measure a risk level and is not authorisation to deploy.
Summary scores for the nine axes
Each point is the mean of the axis’s two numeric answers, out of 4.
View detailed radar values
| Axis | Score / 4 | Answers used |
|---|
Score and detail by axis
The mean summarises both answers. Any answer of 3 or 4 remains highlighted as an individual point requiring attention.
Information to complete
Non-applicable situations
Decision documented by the group
No option is preselected or inferred from the answers.
Exportable action plan
Suggestions can be edited. Add an owner, a due date and the expected evidence.
Plan to review. Some answers have changed since it was prepared; review each action before export.
| Measure to take | Owner | Due date | Status | Evidence or review criterion | Action |
|---|
Collective follow-up history
A descriptive timeline for managers, HR, workers and prevention professionals — never a ranking of teams.
T0 describes anticipated risks; T1 describes observed facts. They can be read side by side, but not as causal before/after evidence. Quantitative changes are comparable only between T1, T2 and later assessments using the same grid.
View all 18 answers
03 / From results to decision
The questionnaire opens the discussion. It does not close it.
Six decisions must be understandable, discussable and traceable before the pilot, then reviewed after use.
Actual need
What work-related problem are we trying to solve, for whom and with what expected result?
Scope
What will the tool be able to do, what will remain human and which uses will remain excluded?
Data and version
What data will flow and which exact version will actually be tested?
Oversight
Who will review, with which skills, authority and available time?
Participation and appeal
How will workers, their representatives and occupational health be involved?
Reversibility
Which effects will trigger correction, suspension or termination of the project?
Test on a small scale, observe the real work, retain the ability to reverse.
- Describe the baseline situation.
- Include ordinary cases and exceptions.
- Limit the scope and duration.
- Decide before any expansion.
04 / Long-term follow-up
Make follow-up a tool for managing work.
A single snapshot is not enough. After the pilot, the same grid helps managers, HR, workers and prevention professionals jointly review workload, room for manoeuvre and actions taken.
- T0Before the pilotDescribe the use, baseline situation and anticipated risks.
- T14 to 8 weeksExamine the first observed facts and correct pilot conditions.
- T2At 3 monthsReview supervision workload, incidents and actual adoption.
- T3At 6 monthsReview organisation, skills, autonomy and open actions.
- T4+At 12 months, then every 6 to 12 monthsMaintain vigilance while the use, model or work continues to change.
Keep what makes change interpretable.
- a stable project reference, the system and its version;
- the observed period and the team or occupations concerned;
- changes in scope, objectives, staffing or organisation;
- agreed actions, their owner, due date and status.
Monitor work, never rank people.
- do not use the profile to assess an individual worker or rank teams;
- never convert NE or NA to zero or aggregate the nine axes into one total score;
- involve users, their representatives and prevention professionals in interpretation;
- reopen the assessment after a major change to the system or task, or after an incident.
Outside this schedule, complete a new assessment after a major change in model, data, objective, population or organisation — and whenever an incident or field signal warrants it. For shared HR use, this local tool does not replace a secure solution with access rights, defined retention and an audit log.
Prepare a follow-up assessment ↑05 / Method and limitations
Nine dimensions, kept separate.
The short online grid has 18 questions. The long instrument has 27 items — three per dimension — in two parallel T0/T1 forms and remains a separate development programme.
Tool status
This grid is exploratory and non-psychometric. It is neither a diagnosis, proof of compliance nor a complete regulatory risk assessment.
Each axis score is the mean of two numeric answers. It structures discussion and does not combine with the other axes into a total score.
Answers are self-reported. They must be compared with observation of real work and the perspectives of the people concerned.