Performance Radar uses artificial intelligence to score sales- and customer-call transcripts. This page explains how AI is used, why its outputs are advisory, and the human oversight we require.
How we use AI
For each call you submit, the transcript is analyzed by a large language model (Anthropic's Claude) against the scorecard you configure. The AI produces a per-skill score, short supporting evidence, and recommendations. A human manager sets the scoring criteria and reviews the results — the AI assists that judgment, it does not replace it.
AI outputs are advisory, not verdicts
Outputs are probabilistic. They may be incomplete or wrong, and they are not legal, HR, or professional advice. Treat every score, excerpt, and recommendation as one input to a human decision — never as ground truth.
Not a sole basis for decisions about people
You must not use a Performance Radar score as the sole basis for any employment, disciplinary, compensation, promotion, or other significant decision about an individual. A qualified person must review and make the decision. Where scores inform employee evaluation, you are responsible for compliance with applicable automated-employment-decision and worker-evaluation laws (for example, NYC Local Law 144 and the EU AI Act).
Your data is not used to train AI
Transcripts you submit are processed only to generate your results and are not used to train AI models. We never store the full transcript — only derived analysis, keyed to a one-way hash, for up to 30 days. See our Data Handling Overview for details.
Which AI provider
Scoring is performed by Anthropic (Claude) via its API; Anthropic is a SOC 2-audited sub-processor. Infrastructure (hosting, storage, identity) is Cloudflare. Both publish compliance documentation, available on request. See our Trust Center.
Questions
Email info@projectmoneyball.com and we'll be glad to help.