Coaches get the most from AI by recording each session with consent (or dictating straight after it), letting a note-taker produce the transcript, then running one fixed prompt that turns it into their own notes template and a short follow-up email they edit before sending. Add a 30-minute Friday review across all clients, and a session's write-up can shrink to about ten minutes.
The notes the AI produces are a draft of your record, not the record. Summarising tools happily turn a client's "I might try asking for a day off" into "Client committed to requesting leave", and a follow-up email built on that line puts words in their mouth. So every step below keeps you as the editor, and there is a dictation route for clients who would rather not be recorded.
Before the first session: consent, a template and one home for notes
Three decisions made once save hours of tidying later.
Consent wording clients can comfortably refuse
Put AI capture in your coaching agreement and ask again at the first session. If you hold an International Coaching Federation credential, its Code of Ethics already expects confidentiality terms to be discussed and documented in writing before coaching begins, and AI recording belongs in that same conversation. Wording to adapt:
With your agreement, I use an AI note-taking tool to transcribe our
sessions so I can listen instead of writing. The transcript is used
only to prepare my session notes and your follow-up email. I delete
transcripts within 30 days and keep only my notes. You can ask me to
switch it off for any session, or any part of one, and it won't
change our work together.
The last sentence matters most. A client talking through a health scare or a falling-out with a business partner may want the recorder off for ten minutes. Make that feel normal, and mark the gap in your notes.
A notes template the AI fills rather than invents
AI writes better notes when you hand it the shape. Keep the template short enough that you will actually reread it before the next session.
SESSION NOTES | [client initials] | session [n] | [date]
Focus the client brought:
What moved (their words, quoted):
What I noticed (my observations, each line starts "Coach:"):
Commitments (only what the client said they would do, in their words):
Unresolved or parked:
Next session:
Flags for me (energy, tone, anything to follow up with care):
Separating the client's words from your observations is the most useful line in that template. It stops your interpretations quietly becoming "facts" about the client three months later, when neither of you remembers who said what.
A capture tool, and a single place notes live
- Zoom: meeting summaries are still built into paid Zoom Workplace plans at no extra cost. Zoom retired the "AI Companion" name in June 2026, so look for plain labels such as meeting summary in your settings.
- Google Meet: "Take notes for me" writes notes into Google Docs for hosts on eligible Workspace plans and for Google AI Pro or Ultra subscribers.
- Otter.ai: Pro costs $8.33 a user a month billed annually, or $16.99 month to month. Useful if you coach across several video platforms.
- Fathom: the free plan records and transcribes without a cap, while the fuller summaries and AI action items sit on paid tiers. Premium is $16 a month billed annually or $20 monthly.
For in-person sessions, skip the recorder. Dictate a two-minute voice memo as soon as the client leaves: what they brought, what shifted, what they said they would do. It is a smaller input than a transcript and, for many coaches, a better one, because you have already done the filtering. If you want a side-by-side of capture tools, see how the main AI note-takers compare for small teams, and before any client call, read whether note-takers are safe on client calls.
Dictation needs one habit to work with the notes prompt: say "quote" before the client's own words. Otherwise everything in the memo is in your voice, and the model can't tell what the client said from what you thought. An illustrative memo after an in-person session with the translation agency founder: "Session six. Came in wanting to talk about the freelancer again. Quote, I know I need someone, I just can't face the interviews, end quote. I think it's less about interviews and more about handing over the big client. Quote, maybe I'll ask the agency I used last year, end quote. Next session in a fortnight." Without the markers, a first run put "less about interviews and more about handing over the big client" under "What moved" as though the founder had said it. With them, that line lands under "Coach:" where it belongs, and "maybe I'll ask the agency" stays a maybe.
Then choose one home for finished notes: your coaching platform's client record, or a private folder per client in your business Google Drive or OneDrive. Not the note-taker's own library, where transcripts pile up indefinitely.
Monday: a five-line prep sheet for each client
Before the week starts, paste each client's last notes (the finished notes, not the transcript) into your AI assistant and ask for a prep sheet. For a full week of sessions this takes about ten minutes.
Here are my notes from [client initials]'s last two sessions.
Write a prep sheet of no more than five lines:
1. What they committed to, in their own words
2. What they parked last time
3. One opening question that refers to their commitment without judging it
4. Anything I flagged to follow up with care
5. Where we are in the programme (session [n] of [total])
Don't add interpretations that aren't in my notes.
For the web design studio founder in the example later on, an illustrative prep sheet comes back like this:
1. Committed to "block Thursday mornings for sales, no client calls"
and "maybe draft a price rise email, but not send it yet".
2. Parked: whether to let the junior designer run client check-ins.
3. Opening question: "How did the Thursday mornings go in practice?"
4. Care: mentioned sleeping badly since the big client left.
5. Session 5 of 8.
Notice that line 1 kept "maybe" and "not send it yet". If your prep sheet says "committed to sending a price rise email", the notes upstream have already hardened the client's words, and the fix belongs in the notes prompt, not here.
Save your template and tone rules in a Project (ChatGPT and Claude both have them) or a Gem (becoming a skill from November 2026) in Gemini, so you are not re-explaining the format every Monday. Using initials rather than full names in prompts is a cheap habit that removes a whole category of risk.
Session days: the ten-minute write-up while it's fresh
Straight after each session, while you still remember the room, run the transcript or voice memo through the notes prompt:
You are helping a coach write session notes. Use the template below exactly.
Rules:
- Quote the client's own words for "What moved" and "Commitments".
If they said "might", "try" or "maybe", keep that word.
- Only list a commitment if the client said it. If I suggested something
and they didn't agree, put it under "Unresolved or parked".
- Put any interpretation under "What I noticed", starting "Coach:".
- Leave out health details, names of third parties and anything the client
asked to keep off the record. Write [OFF RECORD] where that happened.
- No diagnosis and no psychological labels.
Template: [paste]
Transcript or voice memo: [paste]
Here is an illustrative slice of what that prompt returns for a PR consultancy founder, and what the ten-minute check changes:
What moved: "I keep saying yes to retainer clients who don't pay on time."
Coach: Energy dropped noticeably when discussing the largest client.
Commitments:
- Will raise late payment with the largest client this week.
- "I'll look at whether the retainer terms need a payment clause."
Unresolved or parked: [OFF RECORD] 10 minutes
Next session: 14 October
Two fixes. The first commitment isn't in the transcript: the founder said "I should probably raise it", and the coach had suggested a date. It moves to "Unresolved or parked". The "Coach:" line is correct but vague, so the coach adds what actually happened: a long pause and a change of subject when the client's name came up. The off-record marker stays exactly as it is, with nothing summarised around it.
Spend your ten minutes on three checks. Are the commitments really commitments? Did anything from an off-record stretch leak in? Is anything missing that you would want to remember? Transcripts miss what happened in silence. A client who went quiet for a long while when their co-founder came up will not appear in any AI summary unless you add that line yourself.
The same-day follow-up email
Clients read short follow-ups. Aim for under 150 words: thanks, what they committed to, one reflective question and the next session date. The AI draft saves the blank-page minute; your edit keeps it personal.
Draft a follow-up email to [first name] from me, their coach, using the
Commitments and Next session sections of these notes. Under 150 words.
Warm and plain, no exclamation marks, no coaching jargon. Include their
commitments in their own words, one open question to reflect on before
we meet, and the next session date. Don't summarise the whole session
and don't tell them what they learned.
Notes: [paste]
An illustrative before and after for the same PR founder. The AI draft:
Hi [name], thank you for such an insightful session today! It was great
to see you realise that saying yes to everyone is holding the agency
back. As agreed, you'll raise late payment with your biggest client
this week. Remember, growth happens outside your comfort zone!
See you on the 14th.
After a two-minute edit:
Hi [name], thanks for today. You said you'd look at whether your retainer
terms need a payment clause, and you're thinking about when to raise
late payment with your largest client. One question to sit with before
we meet: what would you want a new client to agree to on day one?
Next session: 14 October, 10am.
The edit removed a realisation the client never voiced, softened a commitment back to what was said, cut the slogan and added the time. It also dropped the exclamation marks the prompt asked the AI to avoid, which is a reminder that instructions reduce drift but don't eliminate it.
One thing never to let through: an email that tells the client what their breakthrough was. "You realised your perfectionism is holding the studio back" might be true, but it is theirs to name. Delete it and ask a question instead. The same principle runs through keeping the personal touch when you coach with AI.
If a commitment has a date attached, put a mid-week nudge in your calendar or CRM yourself. A one-line Wednesday message ("How did the conversation with your account lead go?") is where much of the value between sessions lives, and it only works if it arrives on the right day.
Friday: a 30-minute review across every client
This is the step most coaches skip, and the one AI makes practical. Paste the week's finished notes (initials only) into one conversation and ask:
Across these session notes from this week:
1. List each client's commitments and whether their next session is booked.
2. Flag anyone whose commitments have repeated for three or more sessions
without progress.
3. Flag anyone who mentioned stress, exhaustion or a major life event, so I
can check in with care. Don't interpret; point to the line.
4. Note who is within two sessions of the end of their programme.
5. Don't suggest what I should do in sessions.
An illustrative extract of what comes back for three of the twelve clients:
- [WD], web design studio: commitments "block Thursday mornings for sales" and "draft a price rise email, not send". Next session booked. No flags.
- [TA], translation agency: "hire a freelance project manager" has appeared in sessions 3, 4, 5 and 6. Next session not yet booked.
- [PR], PR consultancy: flagged for stress: "I want to stress-test the new retainer terms before sending them".
The second entry is the useful one, and the missing booking is worth a message today. The third is a keyword false positive: the founder was talking about testing contract terms, not feeling stressed. Delete it, and don't be tempted to widen the prompt to catch "anything that sounds like stress"; that produces more false alarms, not better care.
Read the output against your own sense of each client. The repeated-commitment flag earns its keep: a client who has "planned to delegate client reporting" for four sessions running needs a different conversation, not another reminder. If you want the week's commitments to land in a task list without copying them across, the method in turning meeting notes into tasks automatically works for coaching notes too.
The fourth item pays off at programme ends. When a client is two sessions from finishing, ask the assistant for a one-page arc of their notes from session one: the focus they arrived with, the commitments they kept, the ones that kept returning. You will still write the closing reflection yourself, but you start from the whole engagement rather than whatever you remember from the last fortnight, and the renewal conversation gets easier because both of you can see the distance travelled.
For the web design studio founder at session 7 of 8, an illustrative arc reads: arrived saying "I'm the bottleneck on every project"; kept Thursday sales mornings from session 4 onwards; the price rise email drafted in session 5 and sent in session 6; letting the junior designer run client check-ins came up in sessions 3, 5, 6 and 7, and "was successfully delegated in session 7". That last phrase is the one to check. The session 7 notes say the junior designer ran check-ins for two weeks as a trial and the founder "isn't sure yet". Change it to what the notes say, because the closing session is where a client will notice you describing their progress more generously than they would.
What stays out of the AI entirely
- Safeguarding or risk-of-harm disclosures. Write these up by hand and follow your referral process.
- Health and mental-health details, beyond a neutral note that something was discussed off record. Under the GDPR and comparable laws these are special-category data, which carry stricter duties than ordinary client information.
- Third parties by name: the difficult co-founder, the employee being managed out. Use roles instead.
- Anything you've agreed not to share with a sponsor, in employer-funded coaching. Don't let a shared AI summary become the sponsor report by accident. It happens more easily than it sounds: if the sponsor's HR contact was on the calendar invite for a three-way kick-off and the sessions then run on the same recurring invite, a note-taker that sends summaries to all invitees will send them session two's summary too. Book the one-to-one sessions on a new invite with only the client on it, and check the tool's summary-sharing setting before the first recorded session.
- Anything at all on a consumer plan with training still on. On ChatGPT Plus or Claude Pro, switch off the model-training setting in privacy settings; business plans don't train on your content by default. For client details that must go in, anonymise them before you paste.
On retention, set a monthly reminder to delete transcripts older than 30 days from the note-taker and keep only finished notes. The shorter the transcript trail, the less there is to lose if an account is ever compromised.
Four weeks of the routine in a twelve-client practice
To make the time concrete, here is a hypothetical business coach with 12 active clients, including the founders of a web design studio, a PR consultancy and a translation agency. Some meet weekly and some fortnightly, so call it ten 60-minute video sessions a week. These figures are illustrative; time your own first week.
| Task | Before, per week | With the routine, per week |
|---|---|---|
| Session write-ups | 10 x 25 min = 4 h 10 min | 10 x 10 min = 1 h 40 min |
| Follow-up emails | 10 x 12 min = 2 h | 10 x 4 min = 40 min |
| Session prep | 10 x 5 min, ad hoc = 50 min | 10 min on Monday |
| Cross-client review | none | 30 min on Friday |
| Total | about 7 hours | about 3 hours |
Running cost: Fathom's free plan or Otter Pro at $8.33 a month (annual billing) for capture, plus one chat assistant at around $20 a month. How the month might unfold:
- Week 1, run both ways. Write two sessions' notes by hand as usual, then compare them with the AI version. Expect to rewrite the prompt at least once.
- Week 2, tighten the commitments rule. In this sketch, the AI turned "I'd like to" into firm commitments in three of ten notes until the "keep their word" rule went in.
- Week 3, first Friday review. It flags the translation agency founder, whose "hire a freelance project manager" commitment has now repeated for four sessions. That becomes the next session's opening question.
- Week 4, trim. The prep sheet drops from five lines to three because the coach never used the last two.
The four saved hours matter less than what fills them. In this illustration, the Wednesday nudges become the new habit, and they are what clients comment on.
Signs the notes are drifting from what the client said
- Firmer verbs than the client used. "Will" and "committed to" where the transcript says "might" or "I'll think about it".
- Vocabulary that isn't yours. "Breakthrough", "journey", "limiting belief" appearing in notes when you never say those words in sessions.
- Interchangeable notes. If three clients' "What moved" sections could be swapped without anyone noticing, the AI is filling a template rather than summarising a session.
- Missing "Coach:" lines. When observations stop being labelled, interpretation is creeping into the factual sections.
Once a fortnight, pick one session at random and read the transcript against the finished notes. It takes ten minutes, and it is the only reliable way to know the shortcut is still accurate.
Keep a tally so the check tells you something over time. An illustrative fortnight's audit of one session: three commitments in the notes, all found in the transcript, one with a firmer verb than the client used ("will" for "want to"); nothing from the off-record stretch in the notes; one "Coach:" observation that was really a quote from the client, relabelled. One hardened verb in three commitments is the drift the prompt is meant to stop, so the next step is to reread the commitments rule in the prompt, not to add a new one. If two audits in a row come back clean, you can safely move to monthly.
Questions coaches ask before letting AI near session notes
Should I tell clients their follow-up email was drafted with AI?
Tell them at the start that you use AI tools for note-taking and drafting, as part of your consent conversation, and you won't need to label each email. What matters more is that you edit every email so it reflects the session you actually had. If a client asks, be straightforward about what the tool does and what you do yourself.
Can I use a free AI note-taker for client sessions?
You can, but first read what the free tier does with recordings, where they are stored, and whether they are used to improve the product. Free tiers change often. For client work I'd rather pay for a plan with clear retention controls, or skip recording and dictate a voice memo after the session into a tool whose settings you have checked.
How does this change in sponsored or team coaching?
When an employer pays, three parties have expectations about what gets shared. Agree in writing what the sponsor receives, usually attendance and broad themes rather than session content, and keep AI summaries separate from any sponsor report. Some note-takers email a summary to every meeting invitee by default, so check the sharing settings before the first session.
Can AI spot patterns across a client's sessions that I'm missing?
It can surface repetition, such as a commitment that reappears for weeks or a topic that comes up in most sessions. It cannot tell you what those patterns mean for this person. Treat its observations as prompts for your own reflection or supervision, not insights to hand to the client, and never let it label a client psychologically.
Further reads
- Are AI Tools Built for Coaches Worth Paying For Over ChatGPT? — Whether coach-specific platforms beat a general assistant for this routine.
- How Coaches Use AI to Build Workbooks, Exercises and Programmes — Turn the patterns in your notes into exercises and programmes.
- How Coaches Use AI to Qualify Discovery Call Bookings — The same approach applied before a client ever books.
- How Much Does an AI Note-Taker Cost per Seat? — What note-takers cost once you add seats or switch plans.
- How to Stop AI Tools Training on Your Business Data — Switch off model training before any client material goes in.
- How to Build a Shared Prompt Library for Your Team — Store your notes and follow-up prompts so they stay consistent.
- Should Coaches Build an AI Version of Their Method for Clients? — When an AI version of a coaching method helps clients and when it undermines you, what to build it on now custom GPTs are retiring, and the guardrails it needs.
- Can Therapists Use ChatGPT for Session Notes? — Personal plan, de-identified shorthand, business plan or a therapy note tool: where each lands, and why removing a name rarely removes the client.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Zoom product announcements and community notes on the June 2026 AI feature renaming; Google Meet Help, Take notes for me; Fathom and Otter.ai pricing pages; International Coaching Federation Code of Ethics and ethics guidance.