batteriesincluded.com · Questions & Answers

Can we use AI to help with employee reviews without making our managers less human?

Use AI to prepare review materials, not to judge the person. A manager should remain responsible for interpreting evidence, hearing context, and having the conversation. Do not let an AI-generated rating become the default simply because it looks objective.

EOS emphasizes the human side of business, including clear roles and individual contributions. Apply that principle by separating administrative preparation from leadership judgment. AI can organize agreed goals, documented examples, and previous commitments. It should not infer attitude, loyalty, or potential from someone’s writing style or meeting participation.

Set these boundaries before building:

- Use relevant, documented work evidence rather than private messages or broad employee surveillance.
- Show where each factual statement came from, and distinguish facts from interpretations.
- Let employees correct inaccurate information before it influences a review.
- Require the manager to write the final assessment and discuss it directly with the employee.
- Keep compensation, promotion, and disciplinary decisions out of automated execution.

For a Batteries Included engagement, a sensible Lovable or Replit build would be a preparation workspace with evidence links and editable summaries, not an employee-ranking dashboard. Permissions and review controls would need to be deliberately implemented and tested.

Evaluate the workflow by whether it produces better-supported, more useful conversations. Faster paperwork is valuable, but not if managers stop listening. If a manager cannot explain an assessment without pointing to the AI, the tool has crossed the wrong boundary.

Category: AI that stays human

← All questions