AI in performance reviews: what European works councils now insist on seeing

The performance review tool the works council reviews twice

Performance management software entered its modern category around 2014 when Lattice and 15Five built the first wave of continuous feedback tools that replaced the annual review template most HR teams had inherited from the 1990s. Leapsome, founded in Berlin in 2016, built the European answer at a similar moment. Culture Amp, the Australian-incorporated player, expanded into Europe through the engagement survey side of the same category.

By 2022 the category was mature enough that mid-market European employers had largely standardised on one of these tools. The pitch was reasonable. The work the HR team did once a year in spreadsheets and document templates could be done continuously through a structured platform that captured the right data and surfaced the right reminders. The tools did not historically make decisions. They captured what managers and employees wrote, structured the calibration meetings, and produced the documentation a compensation cycle required.

What changed in 2024 and 2025 was the addition of AI features that crossed an analytical line the previous generation of tools had not crossed. Manager assistants that drafted feedback. Calibration suggestions that highlighted potential bias. Performance review summarisers that condensed twelve months of one-on-one notes into a recommendation. The vendors shipped these features fast because the market demand was clear and the technical lift was modest. The unintended consequence was that the tools moved from documentation systems to decision-support systems, which is the line the EU AI Act draws around high-risk employment use cases.

How the vendors responded structurally

Leapsome, with its European customer base disproportionately exposed to German works council requirements, made the most visible structural response. The 2025 Leapsome AI release was carefully scoped to summarisation and drafting assistance rather than to predictive recommendations. The features that could plausibly be classified as automated decision-making were either not built or were explicitly opt-in with documentation about what the model produced and what a manager was expected to verify. The works council documentation that Leapsome ships with the AI features is the kind of artefact a German Betriebsrat can review in one consultation rather than three.

Lattice, with its larger American customer base, shipped more ambitious AI features earlier. The product release notes treated the AI capabilities as competitive parity with the broader HR tech category. The European customer base reacted with friction. A 2025 survey of Lattice's European customers reported that thirty-one percent had disabled or restricted at least one of the AI features after works council or DPO review. The product is good. The European deployment pattern shifted in a way the American customer base did not see.

Culture Amp made the decision to invest more heavily in the engagement and survey side of its product than in the performance review AI features. The strategic reading was that the engagement data is less likely to fall inside AI Act high-risk definitions than the performance review data, and the European customer base would value the engagement deepening more than the performance AI features anyway. The 2025 release pattern reflected that bet.