Copyright: Yedda Stancil

The Consent Problem We’re Not Talking About

Notes from the 2026 ICF Virginia Coaching Summit

Last week, I went to the first ICF Virginia Coaching Summit at the Science Museum of Virginia in Richmond. The theme was “The Human Advantage in a Changing World.” Throughout the day, one question kept coming up:

How does AI affect our work as coaches?

The speakers shared their views. In her session about staying human in a digital world, Yedda Stancil said the real risk isn’t AI replacing coaches. The real risk is coaching becoming formulaic, with presence turning into performance and depth replaced by efficiency. She also talked about the inner work that AI can’t do, and emphasized telling clients when AI is used and getting their consent.

I want to focus on consent, because I don’t think we’ve taken it as seriously as we believe we have.

What clients are actually agreeing to

The usual advice is that if you use AI in your practice, you should tell your clients, get their consent, and remove any identifying details.

But when a coach says, “I use AI to help with my notes,” what does the client actually hear? Most likely, they think it’s something like spell-check. In reality, that phrase could mean any of the following:

• A local transcription tool that processes audio on the coach’s own device and deletes the file afterward.
• A cloud transcription service that stores the recording on a vendor’s servers for thirty days.
• A general-purpose AI assistant where the transcript is retained and may be used to train future models.
• A note-taking integration that cross-references the transcript with the ICF competencies.

These options aren’t the same. Each one handles client privacy, storage, and access in its own way, and agreeing to one doesn’t mean agreeing to all of them.

Most coaches I know aren’t sure what their own tool actually does, and they can’t easily explain it to a client. We’re asking clients to agree to something we don’t fully understand ourselves.

Why this matters

Confidentiality covers all information from a coaching engagement, whether spoken or written. It’s at the heart of coaching. It’s not just an ethics rule; it’s what makes honest conversation possible. People share things with a coach they wouldn’t say anywhere else, and that trust is fragile.

At the summit, someone brought up the trustworthiness formula from The Trusted Advisor by Galford, Maister, and Green:

trustworthiness equals credibility, reliability, and intimacy, divided by self-orientation.

All of these are hurt when a client finds out their words were handled in ways they didn’t expect.

If a client finds out six months later that transcripts of their sessions are stored in a vendor’s training system, the disclosure they signed at the start won’t help. It might have been technically accurate, but it still wasn’t real consent.

This isn’t just my concern. The ICF Code of Ethics, updated for 2025, makes the standard clear. Standard 2.5, created with AI in mind, requires coaches to meet their ethical and legal duties to clients when using any technology, including AI. This covers confidentiality, security, and privacy. It’s not just about telling clients that AI is involved; coaches must ensure the tool doesn’t compromise client confidentiality, which means understanding what the tool does with the recording. The standard is already there. The real question is whether we’re meeting it.

But that requirement shows a deeper problem. How do you actually enforce understanding? You can’t check what’s in a coach’s mind. There’s no test to prove I know how my transcription vendor handles audio, or that I caught a change hidden in the latest terms-of-service update. A standard based on something invisible can only be enforced after the fact, when a client complains or a breach is discovered. By then, the harm is already done.

That’s why this isn’t just about coaches trying harder. We can and should hold ourselves to Standard 2.5. But the solution also needs to involve the tools themselves. Certifying tools wouldn’t fix everything — vendors update products and terms can change, and ICF doesn’t have the same authority as a law like HIPAA. Still, having a clear standard that a tool either meets or doesn’t is more practical than expecting each coach to keep up with every vendor’s terms of service alone.

What better consent would look like

I’m not against using AI. I just believe real consent requires much more honesty than most current practices provide.

Tell your client which tool you use, what happens to the recording, where the data is stored, how long it’s kept, and whether it’s used to train models. Be specific about the tool and its terms.

Then ask the real question: now that you know all this, are you still comfortable with it? Most clients aren’t technical, and asking “Any questions?” rarely gets a real answer. Asking questions like “Does it bother you that the audio is stored on a server outside the country?” or “Are you okay with this transcript possibly being used to train future models?” helps clients share what they really think, rather than just telling you what they think you want to hear.

Some clients will say yes once they understand. Others will say no. The important thing is to ask. If you’ve never had a client decline, it’s worth thinking about why.

Clients should also be able to change their minds, even about past sessions. If data is already in a vendor’s training system and can’t be removed, the client needs to know that before they agree for the first time, not after.

Closing the gap

If we want to keep the human advantage, we need to take real responsibility for what our tools do with client information. A signed form alone isn’t enough.

This is also an area where ICF can lead, and is beginning to do so. The AI in Coaching Task Force, the 2024 AI Coaching Framework and Standards, and the member-only Acceptable Use of AI and Client Protection Guidelines are solid foundations. The framework even speaks directly to developers, covering technical issues like encryption, data handling, and monitoring training data. What’s missing is enforcement. Coaches need guidelines that are specific enough to follow, and ideally a certification path for AI tools that clearly meet the standards — clear rules on data retention, training use, storage, and security — so coaches aren’t left to read every vendor’s terms of service on their own. The profession sets standards for credentialing people. It can do the same for the tools we trust with client information.

Yedda was right: the real danger isn’t being replaced. It’s the slow shift toward shallower, more formulaic coaching. Performative consent is part of that shift, and it’s happening in the most important part of our work.

None of this is easy. Vendor terms are hard to read, and the technology keeps changing. Most of us didn’t expect to deal with this when we became coaches. I work with AI as a researcher, and even I think explaining these tools in plain language could take a whole semester. But if we can’t explain what our tools do, we can’t really ask clients to agree to them. We have to know our tools and be willing to accept a no.

We owe it to our clients to be honest about where their words go.

References

1. International Coaching Federation. ICF Code of Ethics (effective April 1, 2025). Standard 2.5 on technology and AI.
2. International Coaching Federation. AI Coaching Framework and Standards (November 2024). International Coaching Federation. AI in Coaching Task Force.
3. Galford, R., Maister, D., & Green, C. (2000). The Trusted Advisor. Free Press.
4. Stancil, Y. “Coaching in the Age of AI: Staying Human in a Digital World.” Session presented at the 2026 ICF Virginia Coaching Summit, Science Museum of Virginia, Richmond, April 28, 2026.