AI that rates warehouse staff is high-risk, and workers must be told first

If you use AI to hand out tasks or rate staff, the EU AI Act treats it as high-risk.

Annex III point 4(b) is the key line. It covers AI that makes decisions on work terms, promotion or dismissal. It also covers AI that allocates tasks based on a person's behaviour or traits. And it covers AI that monitors and evaluates how workers perform and behave. Point 4(a) adds hiring tools that filter CVs or rank candidates.

For an online retailer, this can catch a lot. Think of pick-rate scoring in the warehouse. Think of shift tools that rank people. Think of dashboards that flag "low performers" in customer service.

The Article 6(3) exit is narrow here. A system that profiles individuals is always high-risk. Scoring named staff on their performance usually counts as profiling.

As the employer, you are the deployer. Article 26 gives you the usual duties: follow the instructions for use, assign trained people to oversee the system, keep the logs, and monitor it. One duty is just for employers. Under Article 26(7), you must inform workers' representatives and the affected workers before you start using the system at work. Use your normal national rules for informing staff.

What is not in scope? Simple rule-based routing, like sending orders by warehouse zone, is not AI under the Act. Forecasting at team level, without scoring individuals, is also outside point 4(b).

A practical first step: list every tool that scores, ranks or routes staff. Include vendor features you switched on in warehouse or HR software. Mark the ones that act on individuals. Then build the worker notice into the rollout plan, not after it. Annex III duties apply from 2 December 2027 under the current timeline.