Tomorrow, October 15, the International Coaching Federation's Mentor Coaching Community of Practice is running a webinar called "Use AI Ethically in Mentor Coaching." The speakers include Susan Caesar, ICF's Director of Artificial Intelligence, and Kristin Kelly, ICF's Director of Ethics, Compliance and Culture, along with mentor coaches Lerae Gidyk, MCC, and Carly Anderson, MCC.
It's a one-hour community session, not a new rule. But the topics ICF chose to put front and center are telling. The listing says participants will look at "how confidential information is handled, when consent may be needed, what should be disclosed, and how transparent agreements can support trust and responsible practice."
That last phrase is the one I'd act on, whether or not you attend. If you're using AI anywhere near your clients' words, it belongs in your coaching agreement.
Why this is coming up now
The ICF listing says it directly: AI is already influencing "the tools, workflows, recordings, transcripts, learning processes, and professional conversations that support coach development." That matches what a lot of coaches are doing quietly. Recording sessions and running the transcript through a summarizer. Pasting notes into a chat tool to spot themes. Using AI to draft a recap email.
None of that is necessarily wrong. ICF's session is framed around where AI-supported tools may be useful for reflection, preparation, feedback, transcripts, documentation and pattern recognition. It's also framed around where "human judgment, presence, partnership, and discernment remain essential."
The problem is when a client doesn't know any of it is happening. A client who shared something painful in session three probably assumed it stayed between the two of you. If that transcript went through a third-party AI tool, they deserve to know before it did, not after.
The session also ties responsible AI use to ICF Core Competencies, including Demonstrates Ethical Practice, Establishes and Maintains Agreements, and Cultivates Trust and Safety. In other words, this isn't a tech question sitting off to the side of coaching. It's part of the agreement and the trust that coaching rests on.
What to put in your agreement
I'm not a lawyer, and you should check what applies where you and your clients live. But in plain terms, an AI clause in a coaching agreement can be short. It should answer four questions.
1. Do you record sessions, and why? Say whether you record, what the recording is for, and that the client can say no without it affecting the coaching.
2. Do any AI tools touch the client's words? Name the kinds of use, for example "I use an AI tool to transcribe and summarize my session notes." You don't need to list every app by brand, but you should be specific enough that a client understands what happens.
3. Where does it go and how long does it stay? Say roughly where the data is stored, who can access it, and when you delete it. If you don't know the answer for a tool you use, that's a sign to find out before using it with clients.
4. What will you never use AI for? This is the part that builds trust. For example: you won't feed session content into a tool that trains on it, you won't let AI write your feedback on a client's growth, and you won't share anything identifying.
Then give the client a simple choice. Something like: "You can opt out of recording and AI note tools at any time. I'll take handwritten notes instead."
A hypothetical example
Picture a leadership coach who's been recording sessions and using a summarizer for six months without mentioning it. A client asks, mid-engagement, "Do you record these?"
The coach can handle that moment in two ways. They can explain, apologize, and offer to stop. Or they can have had a line in the agreement from day one that the client signed, and simply point to it.
The second coach isn't more ethical because of the paperwork. They're more ethical because they asked first. The agreement just makes that visible.
Where AI does help, and where it doesn't
I don't want this to read as "don't use AI." Some of the uses ICF lists are reasonable ones. Preparing for a session. Pulling together your own notes so you remember what a client said last month. Spotting a pattern across your own reflections. Drafting an admin email.
What I'd keep away from AI, as a default:
- Deciding what a session meant. A summary tool will tell you what was said. It won't tell you what mattered. Your notes should be yours.
- Feedback on a client's progress. If a client's going to read it, they're paying for your read, not a model's.
- Anything sent in your voice without you reading it. Recaps, check-ins, follow-ups. Read them first, every time.
We wrote a while back about the ninety-second voice memo after a session. That's still the habit I'd protect. Your first impression of the hour, in your own words, is worth more than any automated summary, and it doesn't raise a single consent question.
How CoachAssist thinks about this
We build tools for coaches, so we have an obvious interest in AI being part of your practice. That's exactly why consent should come first. A tool that clients don't know about is a risk to your relationship with them, however useful it is to you. Tell people what you use, give them a way out, and the tool becomes something they trust instead of something they find out about.
My take
ICF running a session on this tells you where the profession's heading. The questions in that listing, about consent, disclosure and transparent agreements, are going to come up with clients more and more, whether through a direct question or a nervous pause when you mention recording.
You don't need to wait for an official rule to get ahead of it. Add four short answers to your agreement this week. Mention it at the start of your next new engagement. And for existing clients, bring it up yourself before they have to ask. That conversation takes five minutes, and it's a lot easier to have before someone finds out than after.