How to use ChatGPT (and Claude) in your coaching practice

Helping thoughtful, experienced coaches use AI in ways that deepen their practice and preserve their judgement.

My sense is that most executive and leadership coaches have probably opened ChatGPT, asked it a few questions, and closed tehm again unsure of how they might add real value to their practice. The hype says AI will transform coaching, or replace it entirely. I’m not so sure. I think the reality is much more nuanced and useful, especially when we think about how AI can help us become better coaches and help our clients reach better outcomes – rather than simply replace us.

I wrote this article; to cover the practical ways AI tools like ChatGPT and Claude can support a coaching practice, the ethical questions you need to answer before client material goes anywhere near them, and where the tools genuinely differ.

One principle underneath everything that follows. AI should deepen what makes you an exceptional coach, not dilute it. The judgment, presence and professional wisdom you’ve spent years developing are the point. If a use of AI erodes those, it’s the wrong use, however clever it looks.

About me: I’ve been an executive coach since 2013. I work with founders, CEOs and senior technology leaders in high-growth technology businesses, the investment industry and progressive corporates. For the past two years, I’ve been experimenting with AI tools in my own practice, working out what’s genuinely useful and what’s hype. That experimentation became AI-Augmented Coaching, a live course for coaches. Cohort 5 starts Friday 26th September – find out more here.

What can ChatGPT actually do for a coach?

Four uses stand out for a working coach.

• Transcription and session summaries. Recording a session (with consent), transcribing it, and having AI produce a structured summary for both you and your client.
• Reflective practice. Using AI as a thinking partner on your own coaching. Not a supervisor, and not a replacement for one, but a tireless second pair of eyes on what you did in a session and what you missed.
• Preparation and synthesis. Pulling together what you know about a client before a session, or making sense of a 360 report, an assessment and six months of session notes as one picture rather than five documents.
• Thinking things through. The unglamorous everyday use. Talking through a contracting dilemma, stress-testing how you’d open a difficult conversation, working out your rates, or getting a first reaction to a proposal before you send it.

What I’ve not put on the list is using AI to do the coaching for you. I’m talking here about AI chatbots, which aren’t something that I’m necessarily against (I’ve not had much experience of using them as it doesn’t lean towards my preference) but they’re also not something that I’m going to focus on in this article. Rather, I’m going to focus on how you can use AI to augment the coaching experience for both you and your client. The work happens in the relationship, and no model has a relationship with your client. You do.

How do you use ChatGPT to transcribe and summarise coaching sessions?

The workflow is relatively straightforward, even for a non-technical coach. Record the session using Zoom, or a notetaker such as Fathom or Granola, having agreed it with your client first. Take the transcript and ask ChatGPT or Claude to summarise it in a structure that serves you, for example themes explored, commitments made, questions to carry forward.

Two things make the difference between a generic summary and a useful one.

First, tell the AI who you are and what you need. “Summarise this” produces mush. “You’re supporting an executive coach. Summarise this session for my own records, drawing out the client’s core tension, the moments the energy shifted, and what was left unresolved” produces something you’ll actually reread.

Second, treat the summary as a draft, not a record. Read it against your own memory of the session. The AI wasn’t in the room. It can’t feel what you felt at minute forty three, and sometimes it confidently gets emphasis wrong. This is a good example of what is meant by keeping the human in the loop.

Before you run a single client transcript through any AI tool, read the section on client data below. Consent and settings come first, not after.

Can ChatGPT be a reflective partner for your coaching?

This is the use that changes coaches’ practice most, and the one almost nobody starts with.

A reflective partner is an AI you’ve set up, in ChatGPT or Claude, to help you examine your own coaching. You give it standing context about who you are as a coach, the frameworks you work with, and the edges you’re developing. Then, after a session, you bring it the transcript or your notes and ask the questions a good supervisor might. Where did I lead rather than follow? What did I avoid? Whose agenda was that section serving?

Done well, it’s confronting in a useful way. The AI has no stake in flattering you, provided you’ve told it not to. It will point out that you interrupted the client three times in the first ten minutes, or that you moved to problem-solving the moment emotion showed up. You then take those observations to your own reflection, your peers, or your supervisor.

Done badly, it’s an echo chamber. Models default to being agreeable, so a reflective partner has to be deliberately built to challenge you. That setup, the standing instructions and context you give it, matters far more than any individual prompt.

To be clear about the boundary, this doesn’t replace supervision. A supervisor holds ethical accountability, sees your blind spots across time, and is in relationship with you. The AI is a rehearsal space and a mirror, available at 11pm on a Tuesday when the session is still fresh.

What about ChatGPT prompts for coaching?

Search for this topic and you’ll find endless libraries of copy-and-paste prompts. “50 ChatGPT prompts for coaches.” I’d encourage you to skip them.

Not because prompts don’t matter, but because collecting them teaches you nothing. AI fluency is conversational. The skill is describing what you actually need, in your own words, with enough context for the model to be useful, then pushing back when the first answer is mediocre. A coach already has this skill. It’s contracting, clean questions and iteration, pointed at a machine instead of a person.

A more useful habit than any prompt library is a simple structure you adapt every time. Who you are, what the material is, what you want back, and in what form. From there, refine in dialogue. If the summary is too long, say so. If the reflection is too gentle, ask it to be harder on you. The conversation is the method.

Is it safe to put client information into ChatGPT or Claude?

This is the most important question in this article, and it deserves a serious answer rather than a reassuring one.

The short version is that safety depends on the plan you’re on, the settings you’ve chosen, and the consent you’ve obtained, and all three are your responsibility.

On free consumer plans, your conversations may be used to train models, and that’s not an acceptable basis for handling client material. On paid plans, with the right settings enabled, major providers do not train on your data and your conversations are not accessible to other users. The detail varies by tool and changes over time, so check the current data policy of whatever you use rather than relying on an article, this one included.

Settings are necessary but not sufficient. You also need informed consent. That means your client knows you use AI tools, knows what for, and has genuinely agreed, ideally in your contracting and in writing. Many coaches now include an AI clause in their coaching agreements covering recording, transcription, AI-assisted reflection, and the client’s right to say no to any of it.

And a habit that costs little and protects a lot, anonymise where you can. For much reflective work, the AI doesn’t need your client’s name, employer or identifying detail to be useful.

If your instinct is that this all sounds like effort, that instinct is right, and it’s the same effort the profession has always applied to confidentiality. AI doesn’t lower that bar. It extends it.

Claude vs ChatGPT for coaching, which should you use?

Either. The overlap between them is far larger than the difference, and the skills transfer completely. But since almost nobody answers this question properly for coaches, here is the practical distinction as I’ve experienced it.

ChatGPT is the more familiar tool for most people, and it has the larger ecosystem of add-ons and integrations around it. If your clients are likely to have used any AI tool, it’s this one, which can make conversations about your methods easier.

Claude tends to be favoured by coaches for the quality of its writing and reflection. For long transcripts and nuanced reflective work, many practitioners, myself included, find its responses read less like output and more like thinking. Its Projects feature is a tidy way to keep standing context in one place, which is exactly what the reflective work above needs.

Both offer paid plans at around £20 a month, and both are dramatically more capable on paid plans than free ones. A perfectly sensible approach is to trial both for a month, run the same session transcript and the same reflective questions through each, and notice which one you keep opening. My course deliberately works in both, because the fluency you’re building isn’t loyalty to a product. Tools will change. The practice of working with them well is what lasts.

Where AI shouldn’t reach

A short section, but it belongs in any honest answer to this article’s question.

Don’t use AI to do the coaching. Don’t paste a client’s message into a chatbot and send back its reply. Don’t let a summary replace your own reflection rather than prompt it. And watch for cognitive offloading, the slow habit of letting the tool think first. A discipline I hold to, and teach, is brain-first, AI second. Do your own thinking, then bring in the AI to challenge and extend it.

The coaches getting real value from these tools aren’t the ones using them most. They’re the ones who’ve decided, deliberately, where AI belongs in their practice and where it doesn’t.

How do you learn this properly?

You can go a long way on your own with the uses above, taken in order. Start with transcription and summaries, then build a reflective partner, and sort your ethics and contracting as you go rather than afterwards.

If you’d rather do that work alongside other coaches, with structure and someone who has already made the mistakes, that’s what AI-Augmented Coaching is for. It’s a live online course for experienced executive and leadership coaches, hands-on in both ChatGPT and Claude, covering everything in this article in depth, including a full session on ethics, confidentiality and contracting. No coding, no prompt libraries, and no pretence that AI replaces what makes you a uniquely human coach.