AI Session Notes for Coaches: Consent, Privacy and a Practical Workflow
Use AI session notes with explicit consent, minimal data, human review and clear retention. Includes a practical workflow and vendor checklist.
- Agree recording and AI processing before capture begins.
- Name the purpose, provider, outputs, access and retention period.
- Treat generated notes as drafts linked to their source.
- Separate private coach notes from client-facing recaps.
- Delete raw records when the stated purpose no longer requires them.
Coaches can use AI session notes responsibly when recording and analysis are agreed in advance, data is limited to a stated purpose, generated text is human-reviewed and each record has clear access and deletion rules. AI notes should remain draft interpretations. They are not automatically accurate, confidential by default or suitable for every audience.
What are AI session notes?
AI coaching notes are generated summaries, themes or action candidates derived from audio, video, a session transcript or another authorised record. They can help a coach recall decisions and prepare follow-up. They can also invent wording, overemphasise one topic or strip a statement from its context.
The record type matters. A transcript attempts to represent spoken content. A summary selects. A private note reflects the coach's judgement. A shared recap is written for the client. Combining all 4 under the label notes makes consent and access harder to explain.
AI can draft the record; the coach remains responsible for what the record says and who sees it.
What should a client know before recording begins?
The client should know who records the session, which provider processes it, what outputs are created, why each output is needed, who can access them and when they are deleted. They should also know what happens if they decline. A meeting bot joining the call is not a substitute for a clear conversation and documented decision.
The ICF Code of Ethics places agreements, confidentiality and legal compliance within professional standards. The European Commission's GDPR principles overview covers lawful and transparent processing, specific purposes, data minimisation, storage limitation and security. Legal requirements vary by jurisdiction and context, so this article is not a substitute for qualified advice.
| Consent item | Question to answer | Record |
|---|---|---|
| Capture | Is audio, video or both recorded? | Participant decision and date |
| Processing | Which transcription and AI providers receive data? | Named providers and roles |
| Purpose | Is the output for recap, preparation, quality review or another use? | Specific authorised purpose |
| Access | Can the coach, client, sponsor or provider see each output? | Access matrix |
| Retention | When are media, transcripts and notes deleted? | Applied retention schedule |
| Choice | What changes if the client declines or withdraws? | Alternative workflow |
Consent should describe the actual data flow, not merely say that AI may be used.
How much session data should an AI tool receive?
Use data minimisation in coaching: send the smallest authorised input that serves the stated purpose and keep it only as long as needed. A full transcript may be justified for a traceable recap. It may be unnecessary for a reminder about one accepted action.
The European Commission's data minimisation guidance says personal data should be adequate, relevant and limited to what is necessary. The ICO's purpose-limitation guidance also recommends specifying the purpose from the outset and reviewing it to prevent function creep.
- Do not connect unrelated client records to improve a generic prompt.
- Remove sensitive detail that the output does not need.
- Disable provider training or secondary use where the agreement requires it.
- Set different retention periods for raw media, transcripts and approved recaps.
- Test deletion and export before trusting policy language.
The safest session record is the one you no longer need and have actually deleted.
How should coaches review generated notes?
Review generated notes against the source before sharing or using them for a material decision. Check speaker attribution, names, dates, claims, accepted actions, inferred emotion and missing context. An action belongs in the record only when the client accepted it.
| Review check | Failure example | Correction |
|---|---|---|
| Speaker | Coach suggestion appears as client decision | Return it to suggestion or remove |
| Certainty | Possible concern becomes stated fact | Label inference or quote the source |
| Action | Idea becomes assigned task | Confirm acceptance, owner and review point |
| Privacy | Private detail appears in shared recap | Remove it from the client-facing record |
| Relevance | Old topic dominates the next brief | Close or deprioritise it explicitly |
The European Data Protection Board's 2024 opinion on AI models discusses anonymity, legitimate interest and unlawfully processed personal data in AI development and deployment. The practical lesson for a coach is narrower: do not assume that a provider calling data anonymous answers every question about the original recording, account, output or use.
Human review must test meaning and audience, not just spelling.
What is a practical AI session-note workflow?
- Agree the purpose, providers, outputs, access and retention before the first recorded session.
- Confirm consent when the session begins and whenever participants or processing change.
- Capture through the authorised account and verify the recording is attached to the correct client.
- Generate a transcript and draft notes only for the stated purpose.
- Review source accuracy, accepted actions, sensitive details and audience.
- Separate private notes from the approved client recap.
- Share only the approved output and provide a correction route.
- Apply deletion dates to raw media, transcript and generated drafts.
- Carry only relevant open threads into the next session preparation.
MentPass does not record calls itself. It imports recordings, transcripts, summaries and action items from the coach's connected Fathom account, files them against the relevant client and lets the coach review recaps before sending. The processing arrangement still needs to be reflected accurately in the practice's agreement and privacy information.
A responsible workflow has a decision point before capture, before sharing and before retention.
What should you ask an AI coaching-note vendor?
- Which legal entity processes the data, in which role and in which locations?
- Which subprocessors receive audio, transcript, prompt or output data?
- Is customer data used to train or improve shared models, and can that use be disabled?
- How are workspaces and clients separated?
- Who can access raw media, transcripts, prompts and generated notes?
- Can each record type be exported and deleted independently?
- What evidence confirms deletion and backup expiry?
- How are incidents, model changes and subprocessor changes communicated?
Do not accept a security badge as the whole answer. Ask for the processing terms and test the product controls that matter to your workflow. The right vendor should make it easier to explain the data flow to a client in plain language.
If the vendor cannot explain the data path, the coach cannot explain it to the client.
Frequently Asked Questions
Can coaches use AI to take session notes?
Yes, when the processing is lawful, agreed, limited to a stated purpose and human-reviewed. The coach remains responsible for accuracy, confidentiality, access and retention.
Do clients need to consent to AI coaching notes?
Recording and AI analysis should be explained explicitly before processing begins. The required lawful basis and form of consent depend on jurisdiction and context, so obtain qualified advice where needed.
Are AI-generated coaching notes accurate?
They can be useful drafts but may misattribute speakers, omit context or turn inference into fact. Review material claims and actions against the authorised source.
Should a coach keep the full session transcript?
Keep it only when the stated purpose requires it and for the defined retention period. A reviewed recap may support continuity with less confidential data.
Can a sponsor see AI coaching notes?
Only when the coaching agreement and applicable requirements permit that access. Client-facing, sponsor-facing and private coach records should be separated.
What should an AI recording consent form include?
It should identify capture type, purpose, providers, outputs, access, retention, deletion and the effect of declining or withdrawing. Record the participant's decision before capture.