The useful role of AI for business coaches is not to coach your clients for you. It is to clear the preparation, documentation and follow-up work that sits around the conversation.
That line matters. Your client is paying for your judgement, your questions and your ability to notice what they cannot see alone. Handing that work to ChatGPT would make the service worse. Handing ChatGPT the job of turning a transcript into an organised action list? Much more sensible.
Where AI fits in a coaching business
- You ask the next question. AI sorts intake answers into themes.
- You challenge a client's assumption. AI summarises a consented call transcript.
- You choose the coaching direction. AI drafts follow-up from your notes.
- You read emotion and context. AI finds repeated blockers across sessions.
- You provide accountability and care. AI formats a client resource.
Use those boundaries. AI handles the pile. You handle the person.
Workflow 1: turn intake answers into a useful session brief
Most intake forms create more reading, not better preparation. The answers land in a form or spreadsheet, then you scan them five minutes before the call and hope you spot the important bit.
A better workflow:
- Save approved intake responses in one place.
- Ask Claude or ChatGPT to group the answers under goals, current blockers, previous attempts and open questions.
- Make it quote the client's own words instead of rewriting everything into coaching language.
- Review the brief and write the three questions you want to ask yourself.
The AI prepares the evidence. You decide what matters.
Workflow 2: build a pre-call memory without rereading every note
Before a repeat session, give your AI assistant the last approved summary and current action list. Ask for:
- commitments made in the previous session
- actions completed, delayed or still unknown
- decisions that should not be reopened without a reason
- questions left unanswered
- anything the client asked you to bring next time
Do not ask it to decide what the client “needs”. Ask it to retrieve what already happened. That gives you a clean memory prompt without pretending a model understands the whole relationship.
If you use Claude Projects or a custom GPT, keep one controlled workspace per client. Never let one client's information drift into another client's context.
Workflow 3: turn a session transcript into accountable follow-up
This is the first build I would recommend to most coaches.
Record only with consent. Send the transcript through the same structure every time:
Using only this transcript, create a follow-up with: decisions made, actions owned by the client, actions owned by me, due dates explicitly stated, and questions still open. Quote the transcript for any commitment. Mark missing owners or dates as NEEDS REVIEW. Do not invent them.
That last instruction saves a lot of trouble. A polished invented deadline is still an invented deadline.
Review the draft, add the nuance that only you know, then send it from your normal client system. If you want the fuller build logic, my guide to turning AI meeting notes into useful work shows what to do after the summary exists.
Workflow 4: create client resources from your method
You probably explain the same core ideas in different ways every week. AI can help turn those explanations into a worksheet, reflection prompt or decision guide without starting from an empty page.
The source should be yours: a transcript where you explained the concept well, an existing framework, or notes from a real coaching session with identifying details removed.
Ask AI to preserve:
- the order in which you teach the idea
- phrases you use repeatedly
- the questions that create the insight
- any warnings or exceptions
Then edit it like a coach, not a content machine. If the resource could have come from any coach on the internet, it is not finished.
Workflow 5: turn repeated client questions into content
Your best content topics are often hiding in questions clients already ask.
Once a month, collect de-identified questions from session notes. Ask AI to group them by pattern and show the exact wording people use. Pick one question yourself. Record your real answer. Then use AI to repurpose that answer into a blog outline, email draft or social post.
The order is important: your answer first, AI production second. Asking AI to invent your opinion is how coaching content turns beige.
For a practical method, see the interview method for AI content.
A small AI stack for coaches
You do not need twelve subscriptions.
- Claude or ChatGPT: organise source material, retrieve decisions and draft from your method
- Fathom or your existing call recorder: create a consented transcript
- Google Drive or Notion: hold approved client notes and resources
- Zapier, Make or n8n: move information only after the manual version is reliable
Start with one assistant and one job. A reliable follow-up workflow is worth more than a folder full of “ultimate coaching prompts”.
What not to automate
Do not automate sensitive feedback, difficult boundary conversations, performance judgements or anything a client could reasonably believe came directly from you.
AI can draft the admin around a relationship. It should not impersonate the relationship.
Your first 30-minute build
Take one old, de-identified session transcript. Create a follow-up using the prompt above. Compare it with the follow-up you sent manually.
Look for three things:
- What did the AI miss?
- What did it invent or overstate?
- Which part genuinely saved you time?
Turn those answers into permanent instructions. That is how a useful coaching assistant gets built: test one real example, review what happened and change the instructions.
If you want a room where you can build that workflow with help, read my guide to choosing the best AI community for women or see inside Wright Mode.