Coaches build workbooks and programmes with AI by starting from their own method, not a blank prompt. Record yourself explaining your framework, have AI turn it into a method map, then let it draft the outline, exercises and reflection prompts in your voice. Edit every page, pilot with a few clients, and expect about 20 hours for a six-week programme.
What AI does badly is the part clients pay you for. Ask it to "create a six-week confidence programme" and you get the same wheel-of-life exercise, values list and SMART goals worksheet as every other coach who typed that sentence. It also slips in claims you can't support, such as exercises "proven to reduce anxiety". Treat it as a fast production assistant working from your material, and the result sounds like you.
Get the method out of your head before any drafting
Most coaches carry their method as instinct: the questions they always ask, the stage where clients get stuck, the reframe that usually lands. None of that is written down, so it can't go into a prompt. Fix that first.
- Record two 20-minute explanations. Talk through your approach as if briefing a new associate coach: the stages, what you do at each, the mistakes clients make, the phrases you use. A phone voice memo is fine.
- Transcribe them with any note-taker or your phone's built-in transcription.
- Ask AI to extract a method map, using the prompt below.
- Correct the map by hand. This is where the thinking happens. Expect to rewrite a third of it.
Below are two transcripts of me explaining my coaching approach.
Extract a method map with these headings, using my words wherever possible:
- The problem my clients arrive with (as they describe it)
- The stages I take them through, in order, with what changes at each
- The tools, questions or exercises I mention by name
- Where clients typically get stuck, and what I do about it
- Beliefs I hold about how change happens
- Words and phrases I use often, and any I say I avoid
Don't add stages, tools or ideas I didn't mention. Mark anything unclear
with [ASK ME].An illustrative excerpt of what comes back for a coach who works with agency managers:
Stages I take them through:
1. "Stop being the best doer" - noticing where they still do the work
themselves instead of leading it.
2. Setting expectations - agreeing what good looks like with each
team member. [ASK ME: you mention "the one-page agreement" but
don't describe it]
3. Feedback that lands - weekly, specific, small.
4. Managing upwards - negotiating deadlines with account directors.
Words I avoid: "empower", "journey".The fixes are the valuable part. The coach realises stage 4 usually comes before stage 3 in practice, because new managers hit deadline conflicts in their first fortnight. The [ASK ME] on "the one-page agreement" turns out to be the most distinctive tool in the whole method, and it had never been written down. Both corrections would have been lost if the AI had been asked to write a programme straight away.
The "don't add" instruction and the [ASK ME] markers are what stop the map turning into a generic coaching textbook. Every [ASK ME] is a gap in how you explain your own work, which is useful to know before you try to sell it. If you are weighing whether to go further and put the method itself into an AI assistant for clients, read whether coaches should build an AI version of their method first.
Pick the programme's shape before writing a single exercise
The same method can become very different products. Choose the format first, because it decides how much between-session work the materials must carry.
| Format | Suits | What the workbook has to do |
|---|---|---|
| Six-week small group | A method with a clear sequence; clients who benefit from peers | Pre-work for each session, group discussion prompts, one exercise per week |
| Twelve-week 1:1 with workbook | Premium clients who want structure between calls | Deeper exercises, space for reflection, a progress review at weeks 4, 8 and 12 |
| Self-paced workbook | A lower price point or a lead-in to coaching | Everything: explanations, examples, exercises and self-checks, with no coach in the room |
| One-day intensive | A single decision or transition | A short pre-work pack and an action plan template to leave with |
Once you've chosen, ask AI to propose a session-by-session outline from your method map, then argue with it. Its first outline will usually put stages in textbook order. Yours may start with the stuck point, because that is what clients recognise. The self-paced format is closest to a course; if you head that way, how course creators build a course with AI covers the module-and-lesson side in more depth.
An exercise menu: what AI drafts well, and what it gets wrong
Coaching exercises fall into types, and AI's usefulness varies a lot between them.
| Exercise type | AI's first draft | What you add |
|---|---|---|
| Reflection prompts | Good, if given your stage and tone | Cut the ones that are too leading; keep five, not fifteen |
| Scaling questions (0 to 10) | Good | The follow-up question you always ask after the number |
| Case scenarios | Good structure, bland detail | Real-feeling detail from your client work, fully disguised |
| Values or strengths sorts | Generic word lists | Your own list, or a licensed tool you're entitled to use |
| Habit or experiment trackers | Good layouts | What counts as a "done" day in your method |
| Role-play scripts | Useful for difficult conversations | The realistic pushback your clients actually face |
| Scored self-assessments | Looks convincing, isn't validated | Present as a reflection tool, never as a measurement |
The last row is where coaches get into trouble. AI will happily produce a 20-item "leadership confidence index" with scoring bands. It has never been tested on anyone, and calling the result a score implies it measures something. If you use one, frame it as "a starting point for our conversation", and never paste a licensed psychometric instrument into an AI tool to "adapt" it.
The reframing is mostly in the wording around the items. Before: "Your Leadership Confidence Score. Add up your answers. 0-15: low confidence. 16-30: developing. 31-40: confident leader." After: "Rate each statement from 1 to 4 as it feels this week. There's no pass mark and no total. Circle the two statements you rated lowest and bring them to our first session." The items can stay exactly as they were; what changes is that the page no longer tells a newly promoted manager they have "low confidence" on the strength of an untested questionnaire.
Case scenarios show the "bland detail" row in action. Asked for a scenario about a manager whose best designer keeps missing internal deadlines, a first draft reads, illustratively: "You are a new manager at a busy agency. One of your team members has been missing deadlines. You aren't sure how to raise it. What should you do?" The structure is fine; nothing in it would make an agency manager recognise their own week. The coach's rewrite uses details from real client work, disguised by changing the agency type, the role and the numbers: "You manage four people at a social media agency. Your strongest content lead has delivered three of the last five client calendars a day late, always with a good reason. The account director has started checking the work herself. You haven't said anything, because the work is excellent when it arrives." Now the exercise has the actual tension, excellent work and missed deadlines, that makes it hard. Keep names out of scenarios altogether, or use roles, so no client can see themselves in the page.
Prompting for exercises that sound like you
Voice comes from examples. Before generating exercises, collect three short samples of your own writing: an email to a client, a social post, a paragraph from a talk. Save them with your method map in a Project (ChatGPT or Claude) or a Gem (becoming a skill from November 2026) in Gemini, so every draft starts from the same material.
Using my method map and the three voice samples saved in this project,
draft one exercise for week [n]: [stage name].
Purpose: [what the client should notice or decide by the end]
Format: [reflection prompts / scenario / tracker / role-play]
Constraints:
- Plain English, reading age about 12, second person ("you")
- One short paragraph of explanation, then the exercise itself
- Time to complete: [15] minutes
- No claims about outcomes, research or psychology
- Avoid these words: journey, unlock, empower, transformation, authentic self
End with one question the client brings to our next session.For week 2 of the agency-manager programme, an illustrative first draft:
Setting Expectations That Stick
Clear expectations are the foundation of a high-performing team. In
this exercise you'll reflect on how you currently communicate
expectations and identify opportunities for growth.
1. Think of a time when a team member misunderstood what you wanted.
What happened?
2. On a scale of 1-10, how clear are your expectations?
3. What would change if your team always knew what good looks like?
Question for next session: What's one expectation you'll clarify?What the coach fixes: the opening paragraph is filler ("foundation of a high-performing team" could open any management worksheet), so it is replaced with one sentence about the one-page agreement. Question 2 gets the coach's usual follow-up ("What would make it one point higher?"). Question 3 is too abstract and becomes "Pick one account. Write down what 'done' means for the next deliverable, then ask your team member to do the same without looking. Compare." That last rewrite is the method in action, and no prompt would have produced it.
The banned-words line does more than it looks. Coaching content drifts towards the same dozen words, and a list of your own keeps drafts recognisably yours. Building a banned-phrase list explains how to grow it over time.
To check the voice is holding, paste one finished exercise next to your three samples and ask: "List any sentences in the exercise that don't sound like the writer of the samples, and say why." An illustrative answer flags "Take a moment to honour how far you've come" (the samples never use "honour" and are more direct), "Lean into your strengths" (a stock phrase the samples avoid), and a closing line with an exclamation mark. It also flags "Compare the two lists", which is exactly how the coach writes; that one stays. The check takes two minutes per exercise and catches drift before twelve exercises have all picked up the same phrases.
Generate one exercise at a time, not the whole programme at once. Bulk generation produces exercises that repeat each other's structure, and you'll spend longer untangling them than you saved.
From drafts to a workbook people will actually fill in
Layout matters more than coaches expect: a workbook that looks like a wall of text doesn't get used. A simple, repeatable page pattern works best: title, one-paragraph explanation, the exercise with generous space to write, and the question for next session.
- Google Docs or Word: the easiest to update after the pilot. Gemini is built into Docs on Workspace Business Standard and above, and Copilot works inside Word if you have a Microsoft 365 Copilot licence, for tightening wording in place.
- Canva: better-looking pages and templates. On the free plan its Magic Write text tool draws on a monthly AI allowance, and paid plans are more generous, but for a workbook you will mostly paste in text you've already edited.
- Fillable PDF: useful for self-paced workbooks so clients can type answers. Build it last, once the wording has stopped changing.
Before you lay anything out, run the text through a readability check. Checking AI content for plain English has a quick method; aim for short sentences and one idea per paragraph, especially in pre-work people read on their phones.
Pilot with three clients before you sell it
Offer the programme free or at a reduced fee to three existing clients who fit the target group. Ask them to note, on each exercise, how long it took and where they got stuck. At the end, ask three questions:
- Which exercise changed how you thought about something, and which felt like homework?
- Where did you not understand what you were being asked to do?
- What did you skip, and why?
A before and after from an illustrative pilot. The original pre-work for week 4 read: "Reflect on your relationship with senior stakeholders and journal about any patterns you notice." Two of three pilot clients skipped it, one saying they didn't know what to write. The rewrite: "List the last three times an account director changed a deadline on you. For each, write what you said in the moment and what you wish you had said. Ten minutes." All three completed it, and it gave the session its opening. Specific beats reflective when people are busy.
Paste the anonymised answers into your assistant and ask it to group the feedback by exercise and flag any exercise two or more people struggled with. Then rewrite those by hand. AI is good at sorting feedback and poor at deciding which criticism to act on.
An illustrative grouping from the pilot: "Week 3 feedback tracker: two of three found it unclear; one said 'I didn't know if a quick comment in the corridor counted'. Suggestion: simplify the tracker to a yes/no each day." The first half is useful. The suggestion misreads the problem: the tracker wasn't too complicated, it didn't define what counted as feedback. The coach's fix was one line at the top ("Count it if you named one specific thing the person did and what it caused"), with the tracker unchanged. Simplifying it would have removed the part the third client said was most useful.
Claims, credentials and copyright
Three things to settle before the programme goes on sale.
Claims. Remove anything that implies clinical or measured outcomes unless you can point to the evidence. "Clients often find" is honest; "reduces stress by 40%" is a claim someone can ask you to prove. The sales page needs the same treatment. An AI-drafted line such as "a proven six-week system that turns new managers into confident leaders" becomes "a six-week programme built from my one-to-one work with newly promoted agency managers", plus, once you have them and with permission, a sentence from a pilot client in their own words. Keep the programme inside coaching: if an exercise could surface trauma or a mental-health crisis, add a line about when to seek other support, and know your own referral route.
Copyright. In many jurisdictions copyright protects only human authorship, so text generated entirely by AI may not be protected, while your own writing, arrangement and substantial edits generally are. The more of the final workbook that is your editing and your method, the stronger your position. If you plan to license the programme to other coaches, take advice from an intellectual-property adviser, and see who owns AI-generated content for the general picture.
Other people's material. Don't ask AI to "rewrite" an exercise from someone else's book or certification. Rewording doesn't make it yours, and a client who has read the book will notice.
Hours and costs for a six-week programme for agency managers
Consider an illustrative leadership coach turning a 1:1 method into a six-week small-group programme for newly promoted managers at marketing agencies, with a 30-page workbook.
| Stage | Coach's hours |
|---|---|
| Record, transcribe and correct the method map | 3 |
| Choose format and argue the outline into shape | 2 |
| Draft and edit 12 exercises (about 30 minutes each) | 6 |
| Explanatory text, session plans and group prompts | 3 |
| Layout in Docs or Canva | 3 |
| Pilot debrief and revisions | 4 |
| Total | about 21 hours |
Tool cost for the build month is one chat assistant at about $20 a month, plus Canva if you already use it. Nearly all the time saving is in first drafts; editing, layout and the pilot take about as long as they would without AI, and that is where the quality comes from. In this sketch, the exercise the pilot group rated most useful was a role-play for pushing back on a client's unrealistic deadline, drafted by AI but rebuilt around a situation the coach had seen agency managers face many times.
The rebuild shows where the value sits. In the AI draft, the account director's lines were reasonable from the start: "I understand the pressure on your team. Let's find a solution together." Nobody needs practice for that conversation. The coach replaced them with the pushback agency managers actually meet: "The client's already been told Friday. Can't your team just pull a late one this once?", followed by "I'll mention to the MD that you weren't keen." The manager's side of the script kept the AI's structure (acknowledge, state the constraint, offer two options), which was sound. Realistic resistance from the coach and a clean structure from the draft is the division of labour that works across the whole workbook.
Further reads
- AI Session Notes and Follow-Ups for Coaches: A Weekly Workflow — Your session notes show which exercises clients actually need.
- How Business Coaches Use AI Without Losing the Personal Touch — Keep programmes built with AI from feeling mass-produced.
- Are AI Tools Built for Coaches Worth Paying For Over ChatGPT? — Whether coach-specific platforms justify their fee over a general assistant.
- Can You Sell an Online Course Made With AI? — What to check before selling a programme you drafted with AI.
- Turning One Online Course Into Emails and Posts With AI — Market the finished programme from the material you already wrote.
- How to Spot Bias and Stereotypes in AI-Written Content — Catch assumptions in AI-drafted case scenarios before clients do.
- Creating Home Exercise Programmes With AI for Physio Patients — A physio-led workflow for AI-assisted home exercise programmes: shorthand to patient instructions, pain rules, delivery tools and adherence messages.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Canva help pages on Magic Write and AI usage allowances; Google Workspace plan information on Gemini in Docs; published copyright office guidance on works containing AI-generated material.