Skip to content
be-human.ai
← Back to the journal

AI by industry

AI Agents for Executive Coaches: Session Notes and Client Homework

AI for executive coaches: how agents can draft session notes and track homework in client-owned memory, with coach approval before anything is sent.

Be Human typographic cover with the title: AI Agents for Executive Coaches: Session Notes and Client Homework
Conceptual illustration created with AI · Be Human

AI for executive coaches is most useful when an agent handles the memory work around each engagement: structured session notes, homework that clients can act on, and follow-ups that the coach approves before anything leaves the practice. The decision that matters most is ownership. Goals, commitments and reflections should sit in a record that the coach and client control, with clear permissions, rather than in a general chat history or a vendor system the practice cannot easily export.

Key takeaways

  • Session notes and homework are the working memory of a coaching engagement, so they deserve the same care as the coaching itself.
  • An agent can draft recaps, track commitments and prepare reminders, but the coach should approve anything the client sees.
  • Memory should be scoped per client, stored where the coach can export it, and deletable on request.
  • Start with one role and one checkable success criterion, such as whether every commitment from a session has a visible owner and date.
  • Ownership matters more as assistants improve, because more capable tools can reach more of the record.

Why session notes are the working memory of coaching

An executive coach's value comes partly from what happens in the room and partly from continuity between conversations. Clients return because the coach remembers what they said they would do, which pattern keeps resurfacing, and which conversation with a manager stalled. That continuity usually lives in scattered places: a paper notebook, a notes app, calendar descriptions, a few lines in an email thread. Each location is a small risk. Notes are hard to find when they are needed, and hard to hand over if the practice changes hands or a client asks for their file.

The reason to treat this as a design problem rather than a tidying task is that the notes are not only administrative. They hold the client's own words and commitments, recorded in a form the client can review. An agent that drafts, tracks and reminds can make that record consistent, but it can also make an unreviewed record quietly wrong. That is why the approval step belongs in the design from the start.

For a parallel view of the same workflow from the operator's side, the guide on AI Agent for Business Coaches: Session Notes and Follow-Up covers how session notes and follow-up connect in practice.

What the memory should hold, and what it should leave out

A useful coaching memory is narrower than a full transcript. For each client, it should hold:

  • Goals in the client's own words, with the date they were first stated
  • Commitments made in a session, each with an owner, a plain description and a review date
  • Homework assigned, its purpose, and what the client reported back
  • Themes the coach is tracking, labeled as working hypotheses rather than facts
  • Logistics such as session format, time zone and the channel the client prefers for reminders

Equally important is what stays out. Raw audio and full transcripts are rarely needed for follow-up, and each copy multiplies the places where sensitive material can sit. Health details, family circumstances and information about third parties should be recorded only when the client has agreed and the coach needs it for the work. When in doubt, store a short summary written in the coach's voice and keep the original under the coach's own control.

A step-by-step plan for session notes and homework

Start small, and keep a person in the loop at every point where the record changes.

  • Define the record boundary: which fields the agent may write, which it may only read, and which it may not touch.
  • Agree with each client, in writing, what is captured, how long it is kept, and how they can see or delete it.
  • After each session, have the agent draft a summary from the coach's bullet notes, clearly marked as a draft.
  • The coach edits and approves the summary before it is saved to the client's file.
  • Convert approved commitments into homework items, each with a plain purpose and a date.
  • Send reminders only through the channel the client chose, using wording the coach has approved.
  • At the start of the next session, review what was committed and what was reported, so the coach can ask directly rather than reconstruct it from memory.
  • Once a month, review the file: remove stale items, correct errors, and confirm the client still agrees to what is stored.

A simple success criterion keeps the plan checkable. Every commitment from a session should have an owner and a date within one working day of the call, and the coach should have approved every outbound message. If both conditions hold, the system is doing its job. If they do not, the problem is usually in the approval step, not in the model.

An illustrative scenario from a leadership engagement

Illustrative scenario: a coach works with a director who is preparing to lead a team through a reorganization. In one session the client commits to a direct conversation with a peer before the next leadership meeting, and says the conversation has been avoided for months. The agent records that commitment in the client's words, drafts a short homework prompt about what to prepare, and proposes a check-in reminder before the meeting. The coach changes one phrase in the prompt, because the client had asked for a question rather than an instruction, and approves the reminder. When the next session opens, the summary lists the commitment, its date and the client's own report. The coach can ask what happened without relying on memory, and the client can see that the record matches what they said.

Who owns the memory, and why that matters more over time

Ownership is a practical question with three parts. The client should be able to see what is held about them. The coach should be able to export the record in a standard, readable format. And either party should be able to remove material without waiting on a vendor's support process.

These questions grow more urgent as AI systems improve. Capable assistants will infer more from less: a pattern across a few notes, a shift in tone across months, a theme the coach has only half formed. That capability is useful, and it is also the reason the store of notes needs a clear owner. A more capable system with access to an unclear record can produce confident conclusions from material no one agreed to use that way. For a practice, the question is not only which model is strongest today but who decides which memory it may read tomorrow.

The NIST AI Risk Management Framework is a sound starting point for these tradeoffs because it frames risk around governance and context rather than any single tool. For a coaching practice, governance can stay small: a written note of what is stored, who may use it, and how often it is reviewed.

Permissions, prompt injection and the approval boundary

Coaching records invite a specific risk. A client may paste a document, forward an email or type a message that contains instructions, and an agent that treats everything it reads as a command can be steered by that text. OWASP's guidance on prompt injection is clear that content from outside the system should be treated as data to be assessed, not as instructions to be followed. Access control matters for the same reason. An agent working on one client's file should not be able to read another client's file, even when the request sounds reasonable.

A short decision guide helps the coach and any operator agree on the boundary:

  • If an action changes a client's record, the coach approves it first.
  • If an action sends anything to a client, the coach approves the wording and the channel.
  • If a request asks for notes belonging to another client, the agent refuses by default and logs the attempt.
  • If a client asks for deletion, the material is removed and the client receives written confirmation.
  • If a new tool or integration is proposed, the coach reviews what it can read and write before it is connected.

How Be Human approaches this in a managed setup

Be Human designs these systems as a managed service. A first engagement defines one useful role for the practice, usually session summaries and homework tracking, along with the memory fields the agent may write, the approval steps the coach keeps, and a success criterion the coach can check each week. Operations then cover monitoring of what the agent writes, periodic evaluation against sessions the coach has already reviewed, permission reviews as the practice changes, and controlled improvements the coach requests. The coach keeps the record in an exportable form and can ask for changes or removal at any time.

If you are weighing where an agent could help first, the free Be Human AI diagnostic can point to the one task worth handing over.

Frequently asked questions

Can an AI agent write session notes for an executive coach? Yes, as long as the agent drafts from the coach's own bullet notes and the coach approves each summary before it is saved. The draft saves time on structure and continuity, while the judgment about what matters stays with the coach. Summaries should never be filed or sent without that review.

Who should own the notes a client shares with a coach? The client and the coach should both have clear rights over the record, with the coach able to export it and the client able to see and delete what is held about them. Ownership should be written into the client agreement rather than left to a vendor's default settings. A practice that cannot explain where its notes live should not automate them yet.

Should clients see AI-generated summaries of their sessions? Usually yes, provided the client agreed to this at the start and the summary is written in plain language the coach has reviewed. Clients who can see the record can correct it, which improves accuracy and builds trust. Material the coach keeps for professional reasons can be held separately, with a clear note of that choice.

How much of a session should be recorded or transcribed? Less than many tools suggest. Follow-up usually needs commitments, goals and homework, not a verbatim record. Recording or transcription should be an explicit choice made together with the client, and the default should be the shortest record that supports the work.

Your next step

See where an AI agent fits in your company first.

Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →