Most coaches and consultants I talk to aren't short on clients. They're short on the hours around the clients. The coaching itself, the hour where you're present with someone, is the part you love and the part that pays. It's everything stuck to it that grinds you down. Writing up the session before you forget half of it. Building the next programme block. Chasing intake answers. Putting together a progress summary the client can actually feel proud of.
That's the work AI is genuinely good at. Not the coaching. The admin scaffolding around it.
I want to be precise about where the line sits, because this is people work and the failure modes are real. A wrong figure in a marketing email is embarrassing. A wrong inference about someone's health, motivation, or readiness to change is a different category of mistake. So this is a guide to using Claude for the mechanical layer of a coaching or consulting practice, with the judgement layer kept firmly in your hands. I'll go through it in the order a client actually moves: intake, sessions, programme, progress.
Intake: stop losing the good stuff in a Google Form
The intake form is where most practices quietly leak. Either it's too short and you walk into the first session blind, or it's a wall of forty questions and people abandon it halfway. And even when it's filled in, the answers sit in a spreadsheet row that nobody reads properly before the call.
Claude helps at two points. First, designing the form. If you describe your niche, your typical client, and what you actually need to know to do good work, it'll draft an intake that's sequenced properly. The ordering matters more than people think. You open with easy, factual questions to build momentum, then move to the harder reflective ones once someone's committed to filling it in. Ask "what does success look like in six months" as question two and people bounce. Ask it as question nine and they answer honestly.
A prompt that works: "I'm a [strength and habit] coach for [women over 40 returning to training after injury]. Draft an intake form, ordered from easy to reflective, that surfaces their history, their actual goal versus their stated goal, any red flags I should screen, and what's made past attempts fail. Keep it under 15 questions and tell me why each one earns its place."
Second, and this is the bit that saves your sanity, Claude turns a completed intake into a one-page client brief you read in the two minutes before the call. Paste the raw answers in and ask for a summary that flags the real goal, the likely obstacles, the things they said between the lines, and two or three questions you'd want to open the first session with. You walk in prepared instead of skimming a form while they're settling into the chair.
The screening line matters here. Claude can flag that someone mentioned a medical history worth asking about. It can't tell you whether to take them on, and it should never be making clinical calls. It surfaces; you decide. If you work in health and wellness specifically, a structured intake also gives you a cleaner paper trail for when you refer out, which you should be doing more readily than AI would ever prompt you to.
Session notes: write them in five minutes, not thirty
This is the single biggest time saver, and the one most people get slightly wrong.
The mistake is asking Claude to write the session for you. It can't. It wasn't there. What it's brilliant at is taking your rough, messy, post-session brain-dump and turning it into a structured, searchable note you'll actually understand in three weeks' time.
So right after a session, while it's fresh, talk or type for two minutes. No structure, no polish. "Sarah came in flat, work stress, missed two of her three planned sessions, we talked about why and it's not motivation it's that the gym slot clashes with the school run, reset the plan to home workouts three mornings, she lit up when we talked about the 5k, homework is to book the parkrun." Then ask Claude to structure it: what came up, what was decided, what changed in the plan, what the client committed to, what you want to follow up on next time.
A reliable format to ask for:
- Presenting state — how they showed up
- Key themes — what the session was actually about
- Decisions and adjustments — what changed
- Client commitments — their actions before next time
- Coach follow-ups — your actions, and threads to pick up
Two non-negotiables. One, you write the raw input. The note is only ever a tidy-up of your own observations, never an invention. If you didn't notice it, Claude shouldn't be adding it. Two, read the output before it's saved. You're checking that nothing's been smoothed into something you didn't mean. AI loves to make things sound more conclusive than they were. A client who's "exploring whether to change jobs" can come back as "has decided to change jobs," and in coaching that drift is the whole ballgame.
Keep these notes somewhere private and secure. They're sensitive personal data, sometimes special-category data if there's health information involved, and that carries real obligations under GDPR. A locked, access-controlled location, not a shared Drive folder named "clients."
Client programmes: the blank page is the enemy
Building a twelve-week programme from nothing is where a lot of evenings disappear. You know the methodology. You've run it dozens of times. But translating it into a structured, sequenced, client-ready document every single time is repetitive work that doesn't get easier with experience.
This is where Claude pulls its weight, on one condition: the intelligence comes from you. The pattern that works: give it your method, your principles, your sequencing logic, the arc you take people through. Then let it draft the structure for a specific client using their intake and their goals. You're not asking it to invent a coaching programme. You're asking it to apply your programme to this person and lay it out cleanly.
For a business consultant that might be a 90-day engagement plan with phases, milestones, and deliverables mapped to the client's stated objective. For a health coach, a progressive block with weekly focuses, habit targets that ramp sensibly, and check-in points. Ask it to flag where the plan assumes something you haven't confirmed, so you catch the gaps before the client does.
The honest limit: a generic programme generated cold, with no methodology behind it, will read like every other generic programme. Clients can feel that. The value is your framework. Claude just stops you rebuilding the scaffolding by hand every time, and it's a decent sparring partner when you want to pressure-test whether week six is doing too much.
Progress reports: show the client how far they've come
Clients forget their own progress. Week one they couldn't do a full push-up, or they were drowning in unbilled hours, and by week ten that's just normal life. A good progress report is one of the most powerful retention tools you have, because it makes invisible change visible. It's also a chore nobody enjoys writing.
If your session notes are structured, the report mostly writes itself. Feed Claude the run of notes and the original intake goals, and ask for a progress summary the client will actually read: where they started, what's shifted, the wins worth naming, the honest challenges, and what the next phase focuses on. Written warmly, in your voice, addressed to them.
The voice point is worth sitting with. The fastest way to make a client feel like a number is to send them something that reads like a system generated it. Give Claude a couple of your real past messages so it learns how you actually talk, then have it match that. And read every report before it goes out, because this is a document about a real person's effort and it has to be true. Inflated progress is worse than none.
The line, stated plainly
Here's the split I'd hold to. Claude drafts, structures, summarises, formats, and remembers. You diagnose, decide, frame, challenge, and hold the relationship. The moment a task involves judgement about a human being, it's yours, with AI as a clerk, never a colleague.
Two practical guardrails. Don't paste identifiable client data into tools without checking your settings and, ideally, telling clients you use AI for admin support. A short line in your contract covers it and most clients shrug. And insurance: standard professional indemnity may not cover an AI-related error, so it's worth a five-minute call to your provider rather than an assumption.
Where Skill Locker fits
Everything above you can run with good prompts and patience. I built the Health and Wellness coaching pillar on Skill Locker, and I should be upfront that this is my product, because I got tired of rewriting the same prompts. It's a set of Claude skills for exactly this flow: a structured intake builder, a session-notes skill that turns your brain-dump into a clean note, a programme builder, and a progress-report generator that holds your voice. They encode the sequencing and the guardrails so you don't have to remember them every time. Eight skills across the platform are free, so you can test the approach before paying for anything.
But the tools are the small part. The real shift is deciding, deliberately, which two hours a week you're handing to a machine so you can put them back into the hour that actually matters. Start with session notes. Do it for a fortnight. You'll feel the difference before you've built anything else.