Keelson - Performance and OKR Platform with the Works Agreement Built In

Home, OKR tree, objective detail, the manager review workspace and the calibration session. Every one of the 22 demo objectives opens its own detail page with key results, check-ins and a progress chart.

The AI draft reads only the written review inputs, links each sentence to its source, never fills a rating field, and blocks submission while inserted text is still unchanged.

Calibration shows rating distributions by gender, age band, working time and parental leave only for groups of at least 5 people. Smaller groups are suppressed.

4 company, 5 department, 8 team and 5 individual objectives with 45 key results, 35 employees in 6 departments, a 6-phase review cycle and a 17-person calibration session.

Co-Determination on Every Setting

Under section 87(1) no. 6 BetrVG the works council generally co-determines systems able to monitor performance, and section 94 covers assessment principles. Templates, rating scales and calibration rules cannot be used until they are agreed, and the product must collect no activity data such as logins, commits or messages.

AI That Must Not Decide

AI that evaluates workers is generally listed as high-risk under the EU AI Act, and GDPR Article 22 rules out ratings made by automated means alone. A pilot with a generic writing tool ended with managers pasting text in unchanged, and the works council asked for it to stop.

Fair Ratings Without Exposing People

The AGG prohibits discrimination, so a calibration panel needs to see whether ratings skew by gender, age, part-time work or parental leave, but only in aggregate and never in a way that identifies one person in a small team.

Who Sees What

Individual objectives, shared and private 1:1 notes, feedback, reviews and improvement plans each have different audiences. Two improvement plans had been discussed in a public channel by mistake, and every view now has to respect the viewer's role and reporting line.