How Boon coaching AI features use employee data

• Structured program data: goals, themes, action items, and survey scores.
• A coach's own notes, only when the coach asks for help tidying them.

No session audio or transcript reaches an AI model. None exist, because Boon does not record or transcribe coaching sessions.

Anthropic is the primary AI provider, under zero data retention terms. OpenAI is a fallback, with training opted out. Customer data does not train third-party models. Both are listed with Boon's other sub-processors on the Boon for enterprise page.

Between sessions, employees get a summary of their goals and action items, resources suggested for their focus areas, and Practice Space. Every participant still has a human coach.

In Practice Space, a leader picks a real conversation, such as giving feedback, handling a conflict, or a performance review, rehearses it with AI, and gets scored feedback. It is for practice between sessions and does not replace the 1:1 coach. It lives in the employee portal and is also a free public tool at /learn/practice, with no sign-up. Practice Space conversations are visible only to the employee, never to the employer.

Employees can use their coaching program inside Claude or ChatGPT through Boon's MCP connector, after signing in with a one-time code sent to their work email. Boon returns only that employee's own coaching data, and logs which tool was used, never the conversation.

The conversation itself runs on the employee's or employer's Claude or ChatGPT account and falls under that account's AI terms, not Boon's. The one exception is conversation rehearsal, which Boon runs on its own AI provider under the terms above.