Keep AI backstage and the moments clients pay for human. Let it prepare session briefs, transcribe and summarise, draft recaps, workbooks and admin. Never let it make the observation, the challenge or the check-in that proves you remember them. A useful test: if the client would feel short-changed learning AI wrote something, write it yourself.
Coaches who lose the personal touch rarely lose it through AI preparation. They lose it through the messages clients actually read: recap emails that could have gone to anyone, weekly check-ins that arrive on a timer, feedback on a client's plan that's thorough and bland. The fix is specific. Every client-facing message carries at least one line only you could have written, and anything sent on a schedule gets looked at before it goes.
The coaching journey, split into backstage and front of house
Map each touchpoint and decide who does it. This version assumes a one-to-one business coaching practice; adjust it for group programmes.
| Touchpoint | AI's part | Your part | Risk if fully automated |
|---|---|---|---|
| Enquiry reply | Draft from your notes on the enquiry | Add one line about their situation; send | Sounds like every other coach |
| Discovery call | Research the business beforehand; qualify bookings | The whole call | No relationship forms |
| Onboarding pack | Draft the welcome guide, forms, scheduling | A personal welcome note | Low |
| Session prep | Brief from past notes and homework | Decide the focus | Low, if you read the brief |
| The session | Transcribe, with consent | Everything that happens in it | The product disappears |
| Recap email | First draft from the transcript | The insight, the challenge, the tone | High: this is what clients reread |
| Between-session check-in | Remind you who needs one | Write it, or record a voice note | High: it proves you're paying attention |
| Feedback on a client's plan or pitch | Spot gaps, check numbers | What you'd actually tell them | High: generic advice reads as disengagement |
| Invoices, scheduling, reminders | Most of it | Exceptions | Low |
The pattern: AI does well where the job is gathering or drafting, and badly where the job is judgement or attention. Discovery-call qualification is covered in how coaches use AI to qualify discovery call bookings.
Take the enquiry reply, the first thing a prospect reads from you. An illustrative enquiry from a café owner: "Two sites now, I'm working 70-hour weeks, and a friend who worked with you said you'd tell me what I don't want to hear." The AI draft thanked them, described the coaching programme in three paragraphs and offered a discovery call. All accurate, and interchangeable with any coach's reply. The coach kept the booking link and the practical details, cut the programme description to one line, and added: "Seventy hours across two sites often means the second site still runs through you. That's worth an honest look on the call." Thirty seconds of editing, and the reply now shows someone read the enquiry.
Group programmes shift the balance slightly. With eight or ten owners in a monthly session, AI is genuinely useful for spotting themes across the group ("four of you mentioned staffing for next season") and for drafting the shared recap. The personal touch in a group comes from naming people: a line that credits one member's idea, or a question addressed to the person who went quiet. Draft the shared part with AI, then add two or three of those lines yourself. Members notice when the recap mentions them by name for something they actually said, and they notice when it never does.
What "personal" means to a coaching client
Clients describe good coaching in terms of being seen: the coach remembered the detail from three sessions ago, noticed the pattern they couldn't, pushed back at the right moment, and sounded like the same person every time. A language model, by design, produces the most likely next words. That makes it excellent at the average recap and poor at the one observation that matters to this client this week.
Here is the difference in practice. A coach works with the owner of a small tour operator who, for the third session running, has put off raising prices. The AI draft of the recap, and the coach's edited version (illustrative):
AI draft:
"Great session today! We discussed your pricing strategy and explored
some options for increasing revenue. Your action points are to review
competitor pricing and consider a price increase for next season.
Keep up the good work!"
Coach's edit:
"You said something today I want to hold you to: 'I know the price
is wrong, I just hate the conversation.' That's the third time the
price rise has slipped, and I don't think it's about competitor
research any more. So one action, not three: write the email to your
two biggest group clients announcing the new rate, and send it to me
before Friday. We'll look at their replies together."The draft is accurate and useless. The edit uses the client's own words, names the pattern, and makes one specific ask. The AI still saved time: it pulled the quote out of an hour's transcript, which is exactly what backstage means.
Using AI to remember more about each client, not less
The best use of AI in a coaching practice is memory. Keep a single running file per client: goals, commitments made, what got in the way, recurring phrases, family or business events they mentioned. Before each session, ask for a brief:
From this client's file and the last two session transcripts, give me:
1. Commitments they made last time, and whether homework came back
2. Anything they said they were worried about
3. Phrases or patterns that have come up more than once
4. One question I could open with
Quote the client where possible. If something is unclear, say so.
Don't suggest advice.A sample brief for a campsite owner (illustrative):
1. Commitments: draft a job ad for a site manager (not received);
call the bank about the overdraft (done, per her email 12 Sept).
2. Worries: "If I hire someone, I'll spend all winter training them."
3. Pattern: third mention of doing the shower-block cleaning herself
"because it's quicker".
4. Opening question: "What would you stop doing first if the site
manager started on Monday?"
5. She mentioned her daughter is starting university in October.Two checks before relying on it. The job ad wasn't "not received": the client sent it as a message in your scheduling app, not by email, so the file was incomplete; always confirm homework yourself. And item 5 isn't in the prompt's list; it's a personal detail the model included because it seemed relevant. That's useful here, but it shows why you read the brief rather than skim it. The full notes routine is in AI session notes and follow-ups for coaches.
A word on privacy. Coaching conversations include things clients tell nobody else. Record only with consent, keep transcripts in an account where model training is switched off, and don't paste client names into tools you haven't checked.
Sessions also wander into other people's business. A client may say, "Strictly between us, my head chef is off with depression and I don't know if he's coming back." That's a third party's health information, and it doesn't belong in a transcript held by an AI tool or in the client's running file. Agree a simple signal at onboarding ("can we go off the record?") that means you pause the recording, and if something slips through, delete that part of the transcript before any summary runs. In the client file, note only what you need for coaching: "staffing uncertainty in the kitchen; sensitive, don't raise first".
A solo coach on ChatGPT Plus or Claude Pro (both $20 a month) can switch off training in privacy settings; business plans such as ChatGPT Business or Claude Team don't train on content by default but need at least two seats, so they cost a sole trader $40 to $50 a month.
Recaps and check-ins that still sound like you
Give the model your voice as rules, not adjectives. "Warm and direct" means nothing to it; a list of things you do and never do works. A filled-in voice card for one coach (illustrative):
I always: use the client's first name once; quote something they said; end with one question or one action, never both; write in short paragraphs.
I never: say "great session", "crushing it" or "journey"; use exclamation marks; give more than two actions; mention other clients.
Length: recap under 150 words; check-in under 60.
Sign-off: just my first name.
Paste the card into every recap prompt, or save it in a Project so it's always applied. Building a fuller guide is covered in how to build a brand voice guide that AI can follow.
Check that the card is doing its job with a blind test once a quarter. Take five recent recaps for different clients, hide the names, and read them side by side. If you can't tell which client each one was written for, the personal lines have gone missing. In one illustrative run, two of the five could have been swapped without anyone noticing, and both had gone out in a busy fortnight when the coach approved drafts on a phone between sessions. The fix was timing rather than prompting: recaps now go out the next morning, edited at a desk, instead of within the hour.
The realistic mistake here is automation creep. A coach sets up a scheduled Wednesday check-in, AI-drafted from the last recap, sent automatically. For six weeks it works. Then one goes to a guest-house owner whose business partner walked out on Monday: "Hope the week's going well! How are you getting on with the marketing plan?" The client replies with one line, cancels the next session, and later says she felt like she was on a mailing list. The rule that prevents it: AI can remind you who's due a check-in and draft it, but nothing personal sends without you looking at it on the day.
A check-in that passes that look is short and tied to one thing the client said. The AI reminder flags the tour operator from earlier as due, with the note "committed to drafting the new-rate email to their two biggest group clients and sending it to you before Friday". The coach writes, in under a minute: "[first name], it's Thursday. Send me the rate email as it stands, rough is fine, and we'll finish it together." Under twenty words, no exclamation marks, one ask, and it could only have come from someone who was in the session.
Workbooks and exercises: AI drafts, you tailor
AI is very good at producing a first draft of a workbook, a reflection exercise or a twelve-week programme outline. The personal touch comes from the tailoring. A generic "identify your ideal customer" exercise becomes, for a campsite owner, "list the five groups who booked most last season, and the one you'd happily lose". It takes the coach five minutes and turns a worksheet into something that feels written for the client. The drafting workflow is in how coaches use AI to build workbooks, exercises and programmes.
Client plans: AI checks the sums, you give the verdict
When a client sends a business plan, a pitch deck or a pricing sheet, AI makes a useful second pair of eyes for the arithmetic and the gaps. A prompt that keeps it in that lane:
Here is my client's pricing plan for next season. Check the
arithmetic, list any assumptions the figures in the file don't
support, and list questions a lender would ask. Do not recommend
a strategy. Number each point and quote the row it refers to.Illustrative output for a guest-house owner's plan:
1. Row 14: November occupancy is 85%, the same as August (row 11).
Nothing in the file supports a winter level that high.
2. Row 22: the revenue total adds breakfast income twice; the
correct total is 4,620 lower.
3. Question: what happens if the price rise loses the two regular
corporate bookings in rows 30-31?
4. Consider introducing dynamic pricing to maximise revenue in
peak periods.Points 1 to 3 are good preparation. Point 4 is advice the prompt ruled out, and it's generic enough to fit any hotel on earth. The coach's reply to the client passed on the double-counting error straight away, because the client needs that fact before sending the plan to anyone, and turned point 1 into a question for the next session: "What made you pick 85% for November?" Whether the plan is ambitious or reckless is the coach's call, and it rests on three sessions of hearing how last winter actually went, which no model has.
Where coaches overreach
- AI feedback sent as your feedback. Asking AI to critique a client's business plan is fine as preparation. Sending its critique under your name, unedited, is where clients notice the generic tone and the absence of anything you know about them.
- Automated social messages. AI-written connection requests and "saw your post" messages are easy to spot, and prospective clients judge your coaching by them.
- Posts that fake vulnerability. An AI-written story about "the moment everything changed" in your business reads as exactly that. Write your own stories; let AI tidy them.
- A chatbot that talks as if it's you. Some coaches are building AI versions of their method for clients. That can work within limits, but it's a separate decision with its own risks, covered in whether coaches should build an AI version of their method.
Telling clients what you use
Clients are generally comfortable with a coach using AI for notes and admin, provided they hear it from you. One paragraph in your welcome pack covers it:
With your agreement, I record our sessions and use AI tools to transcribe them and help me prepare. That lets me listen properly rather than take notes, and means I don't forget what you've told me. Recordings stay private, and everything I send you is written or checked by me. If you'd rather I didn't record, just say, and I'll take notes by hand.
How much to disclose in other contexts is covered in should you tell customers you use AI?
A week in a coaching practice that gets the balance right
Consider a solo business coach with fourteen one-to-one clients, most of them owners of small hospitality businesses: tour operators, campsites and guest houses, plus a monthly group session. Before AI, a typical week ran to about 14 hours of sessions and 11 hours of everything around them: 4 on notes and recaps, 2 on prep, 2 on workbooks and materials, 3 on admin.
After a couple of months with a transcription tool, a chat assistant and the routines above (illustrative figures):
- Notes and recaps: 4 hours down to about 2. Transcription and first drafts are automatic; the coach spends ten minutes per client on the edit.
- Prep: 2 hours, unchanged, but better spent. Reading a brief takes five minutes; the rest goes on thinking about the client.
- Materials: 2 hours down to 1.
- Admin: 3 hours down to about 1.5.
Roughly 3.5 hours saved a week. The coach decided in advance where they'd go: a short personal voice note to every client midweek, which takes a little over two hours, and one more discovery call. So the week isn't much shorter. It's the same length, with more of it spent on the part clients would miss if it vanished, which is the right trade for a business whose product is attention.
The cost side, at list prices: Otter Pro at $8.33 a month billed annually for transcription, plus ChatGPT Plus or Claude Pro at $20 a month, comes to under $30 a month, or about $2 for each of the 14 or so hours saved in a month. The expensive part isn't the software. It's the ten minutes of editing per client, which is exactly the part not to cut.
Further reads
- Are AI Tools Built for Coaches Worth Paying For Over ChatGPT? — Whether coach-specific AI tools beat a general assistant.
- How Freelance Writers Use AI Without Losing Their Voice — More techniques for keeping AI drafts in your own voice.
- Are AI Meeting Note-Takers Safe for Client Calls? — Check a note taker before it records a client session.
- How Much Time Can AI Save a Solo Consultant Each Week? — How much time AI realistically saves a one-person practice.
- How to Draft Customer Email Replies With AI That Sound Like You — Drafting replies that sound like you, for any client email.
- How to Keep Customer Data Private When Your Team Uses AI — Keep client confidences private when AI is in the workflow.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI and Anthropic plan pages (ChatGPT Plus, Claude Pro, two-seat minimums on business plans); Otter.ai pricing page. Coaching examples and outputs are illustrative.