AI at work: pointers for prevention specialists and ergonomists

An AI changes a work situation

Step 1Observe real work

1. 1.1

   Have you described the work before and after AI, using concrete situations?

   If no

   Start from real situations: tasks that have moved, checks, corrections, exceptions, decisions that are influenced. Activity often shifts towards checking outputs and handling the most complex cases.

   Sources: Homepage · prevention specialist profile · Law and governance · role of prevention

2. 1.2

   Do employees approve an agent’s outputs or actions one after another?

   If yes

   Analyse supervision as a vigilance task: approvals per hour, time and authority to refuse, session length, breaks, rotation, responsibilities set out in writing. Record it in the DUERP (the company’s occupational risk assessment document) as a situation exposing workers to psychosocial risks.

   Source: Reading · Human in the loop

3. 1.3

   Are tasks from other occupations entering the job thanks to AI?

   If yes

   Document what enters and leaves the occupation, the time actually saved, the training needed and who is responsible for checking. Being able to do a task with AI does not automatically mean having the skill.

   Source: Reading · Occupational boundaries

Step 2Read through the six Gollac dimensions

1. 2.1

   On which dimension do the warning signs appear?

   Each card gives the question to ask for the DUERP and a measure at the source, taken from the prevention fact sheets.

   1 · Work intensity and working time

   Warning signs: raised targets, evening work, cognitive fatigue, training outside working hours.

   Question for the risk assessment (DUERP): “Has the time saved through AI been reallocated? To what, and who decided?”

   At the source: measure the real workload, checking included, and extend the right to disconnect to AI.

   2 · Emotional demands

   Warning signs: more complaints, cynicism, more frequent tense exchanges, requests for a transfer.

   Question for the risk assessment (DUERP): “What proportion of the interactions still handled by people is now tense or distressing?”

   At the source: share the work between people and AI according to the remaining emotional load, keep simple access to a human contact.

   3 · Autonomy

   Warning signs: a feeling of carrying out tasks without making decisions, recommendations followed without scrutiny, being stuck when the system fails.

   Question for the risk assessment (DUERP): “Who can reject the tool’s recommendation, and what happens when they do?”

   At the source: the right to disregard the tool’s recommendation without penalty, employees involved in configuring the tool.

   4 · Social relationships at work

   Warning signs: challenges to assessments, fewer informal exchanges, undeclared AI use, complaints of unfairness.

   Question for the risk assessment (DUERP): “Can a decision affecting a worker rely solely on an AI-generated indicator?”

   At the source: no fully algorithmic individual assessment, criteria that can be explained and contested.

   5 · Value conflicts

   Warning signs: talk of “botched work”, workarounds of the tool, ethical disagreements left unspoken, departures.

   Question for the risk assessment (DUERP): “Can an employee report an AI result they consider unethical, and is the report followed up?”

   At the source: define the quality of work with AI collectively, and write down how responsibilities are shared.

   6 · Job and work situation insecurity

   Warning signs: rumours about the occupation, repeated questions at the works council, sleep problems, early departures.

   Question for the risk assessment (DUERP): “Do workers know what the project will change for their role and employment over the next two years?”

   At the source: inform early and consult the works council, include AI in strategic workforce planning (GEPP), deploy in reversible stages.

   Sources: Prevention fact sheets on psychosocial risks and AI · Psychosocial risks · six dimensions

2. 2.2

   Do the effects differ by task, occupation or person?

   If yes

   Identify the exposed groups and document the differing effects: a positive average can hide negative effects, and an average gain must not become a uniform target.

   Sources: Reading · AI and real work · Law and governance · role of prevention

Step 3Propose and follow up

1. ·
   Prioritise the measures

   Act at the source first, on the organisation and on the choice and configuration of the tool; then detect deterioration early; then support the people affected. Training employees to manage their stress does not make up for an organisation that exposes them to it.

   Labour Code L. 4121-2

   Source: Psychosocial risks · prevention at three levels

2. 3.1

   Do you yourself use AI to prepare a DUERP, a company risk sheet or an action plan?

   If yes

   Assess it task by task (omissions, corrections, quality of the action plan), with no identifying data in an unauthorised tool, and do not let an automatic summary replace field analysis.

   Sources: AI in OHS services · uses · AI in OHS services · responsibilities

3. 3.2

   Are collective monitoring indicators in place?

   If no

   Propose some: hours, sick leave, errors, incidents, workarounds, changes in skills, recorded at T1, T2, T3 and at each change of tool. They serve to adjust the organisation, never to assess people.

   Sources: Assess an AI project · Psychosocial risks · early detection

Record in the DUERP and reassess whenever the tool changes