The Coaching Laboratory

The coaching industry's encounter with generative AI remains largely uncharted territory. Practitioners like BeHati Hart, who has transitioned from government service to solo practice, are conducting real-time experiments that reveal both the possibilities and constraints of algorithmic augmentation in human services.

Ms Hart's approach resembles the iterative methodology she learned in government design work. She began modestly: recording client sessions, using AI to identify recurring themes, then gradually expanding the tool's role. A critical juncture came when one client expressed willingness to continue the experiment beyond the standard five-session engagement. This created the conditions for developing what she calls the "Betty Method"—a nine-step framework co-designed with AI by analysing patterns across multiple clients.

The progression is revealing. Initially, AI served as a research assistant, aggregating data and surfacing patterns. Then it became a creative collaborator, helping to systematise those patterns into a replicable methodology. Now, as a custom GPT named "Betty," it functions as a client-facing tool available outside standard business hours. Each phase expanded the AI's autonomy while raising new questions about quality control and professional boundaries.

Ms Hart's ongoing adjustments to Betty demonstrate the experimental nature of this work. She describes implementing "bumper rails" in response to hallucinations and redirecting the system when it ventures toward therapeutic advice beyond her coaching credentials. This trial-and-error process—identifying failure modes, adjusting parameters, retesting—mirrors product development cycles in software engineering.

The experiment operates at multiple levels. On the individual level, Ms Hart is testing whether AI can help manage information overload for a practitioner with ADHD. On the client level, she is exploring whether asynchronous AI access improves outcomes. On the business level, she is investigating whether custom GPTs enable sustainable solo practice scaling. Each layer generates different data about where algorithmic tools add value versus where they introduce new complications.

What makes this experiment particularly instructive is its transparency about limitations. Ms Hart acknowledges that Betty "would need to have some engineers behind that" to operate fully independently. She recognises the tool requires continuous human oversight. She distinguishes clearly between coaching and therapy, attempting to embed those boundaries in the system itself. These acknowledgements reflect a practitioner grappling with real constraints rather than marketing an idealised solution.

The broader significance lies in methodology. Unlike corporate AI deployments with extensive testing infrastructure, solo practitioners are conducting experiments with limited resources and immediate client exposure. They must simultaneously refine their tools and deliver client value—an approach that accelerates learning but increases risk.

Ms Hart's work essentially asks: can AI extend professional expertise without diluting it? Can asynchronous algorithmic guidance supplement synchronous human interaction? Where precisely does pattern recognition end and judgment begin? These questions have no predetermined answers. They require empirical investigation.

The coaching industry may prove an ideal testbed. It is large enough to matter economically but informal enough to permit experimentation. It involves tacit knowledge that resists codification, making it a challenging case for AI augmentation. And it deals with subjective outcomes—client satisfaction, behaviour change, goal achievement—that test whether algorithmic tools can handle ambiguity.

The results of these experiments will emerge slowly. For now, practitioners like Ms Hart are generating the data points that will eventually inform broader understanding of how AI reshapes professional services.

What we’re into this week

Aaron: 

Behati’s interview made me think of this experiment by Robert Caruso, a senior Cistrix engineer, who ran an Atari 2600 versus ChatGPT chess match a few months ago in which the Atari won! It was a total upset; no one saw that one coming. I’m always so glad to see people playing with AI and their own interests and practices. These are great tools, but always need open world testing. 

Scott: 

I didn’t listen to this interview until yesterday, but I grabbed it right on the heels of listening to Ezra Klein’s interview of one of my favorite artists, Brian Eno. Eno has such a wonderful orientation to AI and all forms of generative art, and, in thinking about BeHati’s work, this quote from the Eno interview floated to my mind: 

“It wouldn’t be a bad idea if everything that was generated in that way, 50 percent of all the profits from it immediately go back to society in some way.

It would be saying, “We are not claiming to be the geniuses here. We’re claiming to be the people who know how to corral it all, how to put it together, how to make it available to you. But thank you for all the stuff you’ve written, Ezra and Brian and everybody else. And with your permission, we’ll redistribute this to society.” 

I just think that experimenters like BeHati and Brian Eno should get something for their work. 

Ana: 

Here here! Knowing that AI companies get to make profit off of work but we don’t see any of that profit really rubs me the wrong way.  What I’m into this week is actually a book, called The Fraud by Zadie Smith. It’s topical because it’s about a guy in the 1800s who persuaded thousands of people in England that he was a long lost aristocrat. The story is epic and truly amazing, but it’s topical here because so often I think that’s what AI is doing: convincing us that it is what it isn’t, to gain our trust and attention. The Fraud is just a proxy for and a concentration of how society fools itself already, which also seems to be in the same orbit as the greater discussion here. 

CX Research—Documented

BeHati Hart drives her business through fearless experimentation and continuously talking to her clients. She's developed that skill through years of experience, but if you're just starting out or looking for a packaged CX research to bring to your leadership, get the HCD Discovery Guide. 

The Discovery Guide was developed alongside thousands of professionals working in healthcare, veteran support, education, and operations. When you open this Guide, you'll see: 

1. The Why and How of customer research.

2. Step-by-step, modular instruction that you can dip in and out of easily. 

3. Plain language for working professionals. 

You don't need a full UX or CX team to get to know your customers: you need the HCD Discovery Guide. Grab the ebook version today!

 

Credits: 

Host: Aaron Meyers

Guest: BeHati Hart

Producer: Ana Monroe

Article: Fellow, Ishmael Interactive 

Artwork: Acrobatic Dance 1911 Ernst Ludwig Kirchner, via the National Gallery of Art