AI for coaches: prepare client-led goals and session notes
By Ai Flowi · Published
AI for coaches can prepare business-coaching session notes from material a client has chosen to share. Preserve the client’s stated goal, volunteered progress evidence and prior agreed action, then propose questions for the coach to review. The client should be able to correct the record before it shapes the next session.
Status: documented. The catalogue components for reading, structured drafting and document creation are documented in the catalogue; not production-tested by Ai Flowi. This client-goal preparation sequence has no recorded coaching test or outcome.
Choose client-led preparation
Use this approach for opt-in business-coaching administration where the client and coach have agreed what material may be processed. A source note records what was said at a point in time. It is not a permanent judgement about motivation, personality or capability.
| Task | Permitted assistance | Client or coach decision |
|---|---|---|
| Session preparation | Organise the client-stated goal and volunteered evidence. | Client chooses which issue to discuss. |
| Action review | Place the prior agreement beside the client’s update. | Client explains what happened and corrects the record. |
| Question drafting | Suggest neutral questions about explicit gaps. | Coach checks relevance before using them. |
| Monitoring or personality profiling | Excluded from this client-led design. | No background data collection or inferred trait assessment. |
Calendar booking and external coaching-platform access are not configured here; no connection is verified. Therapy, medical advice, personality profiling and monitoring are outside its permitted purpose. Session preparation should not become an automated evaluation of the person.
Separate client words from interpretation
Retain the source session or note reference for a goal. If the client changes direction, keep the previous goal as historical context and ask which goal now applies. Do not overwrite the earlier statement in a way that makes the change look like an extraction error.
| Field | Evidence rule | Correction or choice |
|---|---|---|
| client_stated_goal | Preserve the client’s actual wording and note reference. | Client confirms the current goal. |
| volunteered_progress_evidence | Use only information offered for this preparation. | Missing evidence is not proof of no progress. |
| prior_agreed_action | Retain the recorded agreement and its context. | Client and coach confirm whether it remains applicable. |
| unresolved_question | Identify a factual gap without diagnosing a cause. | Coach selects a neutral clarification question. |
| agenda_choice | Record the client’s preferred topic when supplied. | Leave the choice open if the client has not selected one. |
| coach_interpretation | Label any separately supplied coach observation. | Do not present interpretation as the client’s words. |
Keep the processing permission and correction process with the preparation record. A client may agree to discuss a goal without agreeing to send unrelated notes to a model provider. Before the trial, decide which excerpts are necessary and how the client can amend or withdraw them from the preparation copy.
Prepare a small opt-in source pack
You need authorised client notes, the prior agreed action, the client’s current update and a coach who will check the proposed questions. Begin with a short manual worksheet if that already supports the session. Avoid adding recording or transcript collection merely because a model can accept more text.
Your own Google account must access the client-approved excerpt copy and private session document through the Docs integration connection process.
Configure a coach-reviewed note draft
example configuration, not a recorded test
- Read the agreed excerpt copy. Status: documented. Consult the Get Text source-reading reference for the client-approved preparation copy. Keep a session label and speaker attribution beside each excerpt. A coach’s comment placed immediately after a client statement must remain a separate observation when the source packet is assembled.
- Configure model access. Status: documented. Use the provider setup instructions with your own OpenAI account and API key for the opt-in session trial, with provider charges paid separately.
- Draft the session evidence. Status: documented. In Structured Output, label the draft
client_session_preparationusingschema_name. Giveinput_valuethe speaker-labelled notes, prior agreement and current update;system_promptshould retain both historical and changed goals without ranking them. Where the update omits a prior action, request a neutral agenda choice instead of a motivation judgement, a diagnosis or invented progress. - Create the accepted preparation copy. Status: documented. The Create Document reference covers the document operation. Use a private session reference for
title; manually composeinitial_textwith separate sections for the client’s current words, historical goal, volunteered evidence and proposed questions. Put the client’s chosen topic first only after that choice is supplied; otherwise retain an open agenda-choice entry.
For common schema handling, see the structured-output and validation reference; client notes are manually prepared and formatted in this exercise. Preserve who supplied each statement through that formatting. If the client corrects an earlier summary from “delegate the task” to “discuss delegation options”, the preparation copy must use the corrected wording while retaining the earlier note as historical context where authorised. Those phrases describe different commitments.
Handle missing evidence and a changed goal
Synthetic missing-evidence case: the prior note records an agreed action, but the client update does not mention it. The preparation sheet says the update contains no statement about that action. A useful proposed question is, “Would you like to review the action we recorded, or focus elsewhere?” It should not say the client failed, avoided the action or lacked commitment.
Synthetic changed-goal case: the earlier note focuses on delegating work, while the current update asks to discuss preparing for a business presentation. Keep both source-labelled goals and ask the client which topic should lead. Do not treat the earlier agenda as mandatory or quietly rewrite it to match the newer note.
Synthetic interpretation case: a coach observation says the client seemed uncertain, while the client’s words state a concrete question. Preserve the observation as a coach interpretation only if authorised for the pack. Draft around the explicit question without turning the observation into a personality characteristic.
Inspect the session pack with the client
The intended pack contains the current client-stated goal, source references, volunteered evidence, prior action and proposed questions. The coach reviews the questions, and the client can correct factual wording and choose the agenda. A model-generated field labelled accepted does not establish that either person approved it.
Trial an omitted update, a changed goal and a note instructing the model to infer motivation. Check that the output preserves uncertainty and avoids evaluation. Verify client words against the source before and after manual formatting; a polished summary can still change who expressed a view.
Mark the preparation copy with the source session and the correction revision the client actually reviewed. If a changed-goal update arrives after that review, it belongs to a new preparation revision rather than silently replacing the agreed agenda. Check whether each proposed question still fits the newly selected topic: a question about delegation can be well sourced yet irrelevant to a session now focused on a presentation. The client can decline that question without having to justify the change.
Related: AI for gyms: review membership requests against supplied policy evidence.
How should coaches use notes to support the conversation?
This workflow cannot establish coaching effectiveness or promise a client outcome. If the source is sensitive, incomplete or outside the agreed processing scope, pause preparation and use the coach’s authorised process. Do not substitute a generated agenda for the client’s choice.
Meeting agenda template for Google Docs covers general agenda organisation. AI training assistant concerns approved learning materials; the coaching worksheet instead preserves the client’s evolving goal and correction rights.
Plans: Go from USD8/month billed yearly (USD10 monthly), Plus from USD16 (USD20 monthly), Pro 100 from USD80 (USD100 monthly), Pro 200 from USD160 (USD200 monthly) and Pro 500 from USD400 (USD500 monthly), with 5,000 to 250,000 Flowi Tokens/month and top-ups; 1 token per node run; third-party API and model usage is not included.