Tuesday, July 25, 2023

Can Artificial Intelligence Deliver Transformational Coaching? ~ Increasingly, the answer is yes

We are pleased to share an article entitled “Can Artificial Intelligence Deliver Transformational Coaching? ~ Increasingly, the answer is yes” written by Joe Dunn & Julian Humphreys


In 2021, the International Coaching Federation (ICF) confidently asserted that artificial intelligence (AI) will soon allow coaches to “offload low value coaching activities such as brief questions, assessment, and journaling” to computers, leaving human coaches with “the high-value work of complex, transformational coaching.” A lot can change in two years, and despite many coaches still believing that there is something inherently human to transformational coaching, it’s a position increasingly difficult to defend. In this article, we address the question, ‘can AI deliver transformational coaching,’ concluding that the answer is “increasingly, yes,” with the breadth and depth of that experience improving rapidly over time.


What do we do when we coach? We draw on our training, our practice, our life experience. We listen carefully to the client, absorbing their state into ours. We allow their words, tone of voice, facial expressions to combine with what we know of their history, their current situation, their personal patterns and with our own rich amalgam of experience and knowledge, and from that combination, make the next best move.

But how does that move come into being? Do we choose it, or does it simply arise, driven by something beyond us and beyond the client in the same way that the next line of a poem simply appears to the poet? Are these crucial moments in coaching so intrinsically human that no machine will ever be able to replicate them?

The line that separates intrinsic human ability from machines has been moving rapidly. For a long time, it was unimaginable that a computer could beat a human at chess. Now, not only do computers trounce the very best human chess players, they also beat them at more creative games, like Go, sometimes producing moves that are shockingly original.

The idea that machines could produce art or poetry would have been ridiculed a few years ago. Now that they do, and in prodigious quantity, the argument is that it’s just not very good – it lacks some “human” quality of genuine creativity. Exactly what that quality is remains uncertain. We know it when we see it, apparently, except that it has to come from human experience.

The argument can appear quite circular: a machine cannot produce human-like art because human-like art can only be produced by a human.

The same argument can be made against truly transformational AI coaching. Transformational coaching can only be done by a human because it depends on some magical quality that is essentially human and can never be replicated, regardless of how powerful the machines become.

AI, after all, is simply tossing together patterns of expression based on a predictive analysis of billions of already existing patterns of expression. It is not experiencing the client’s emotional state, it has no insight, and it knows no meaning from which it constructs the coaching interaction.

But how different is that process from what occurs inside a human coach? How is the “arising” that happens inside us as we coach so different from a sophisticated prediction of the “right” action based on an enormous, automated pattern-matching exercise? What is the human really doing that an AI can’t?

Whether an AI can ever deliver transformational coaching hinges, we believe, on the answer to that question. An AI can “know” a client as well as any human can, if not better. It can remember every detail of every previous session, every important insight, every metaphor that cracked open a space in the client’s perception. It can know the client’s heart rate, diet, amount of sleep and blood sugar level. It never gets bored or tired. It benefits from the knowledge and experience of thousands of coaches, millions of blog posts, and every book ever published.

So, what is it missing?


Perhaps the missing piece is the felt experience of the relationship.

Coaching exchanges often have an energy; a fluidity that arises from the trust established between coach and client, a product, at least in part, of how the coach is – who they are, what they sound like, how they hold themselves.

Two arguments suggest that an AI coach will never be able to form a similarly close relationship with human clients.


An AI coach expresses itself though text, synthetic audio, video and animation – all of which emphasize the artificiality of the connection and have less nuance (at least for now) than seeing and hearing another human being.

The interface constraints on human-machine interaction are being reduced very rapidly though. Lil Miquela is a 19-year-old robot living in L.A. that appears in YouTube videos, on Instagram (with 2.9 million followers) and elsewhere. She still has the slightly plastic look of an avatar (perfect for Instagram!), but only slightly. It won’t be long before Lil Miquela, along with her brothers and sisters, will be talking to us, just as a human might, over Zoom, drawing on the full extent of her AI-enabled wisdom.

Given that the interface constraints on AI coaching will rapidly improve, will we be able to overcome the strangeness of relating to a new type of entity and build close, trusting relationships with a non-human system?


An AI coach that is “human-like” but not fully human feels weird to us on some primitive level. Developments in human-computer interaction suggest we can easily overcome this bias, however.

Joseph Weisenbaum, the creator of the first (and very primitive) AI therapist, Eliza, was so concerned about the illusion of care that Eliza provided that he wrote, “What I had not realized is that extremely short exposures to a relatively simple computer program could induce powerful delusional thinking in quite normal people.”

The delusion Weisenbaum was referring to was that the computer actually cares, and there is lots of evidence to suggest that this delusion remains. For example, Replika is an AI-enabled chatbot “virtual friend,” designed to be “The AI companion who cares. Always here to listen and talk. Always on your side.” It has more than 10 million users, and is rated 4.5 out of 5 with 200,000-plus ratings.

Evidence also strongly suggests that the “weirdness” of new technologies quickly subsides when they show themselves to be widely useful. There was a period when walking down the street having a conversation with a small microphone dangling in front of your face was socially odd. A few decades before, the idea of gathering the family around a large box in the living room to watch images was felt to be strange. We adapt to interfaces that work.


Foundational new technologies in their initial phases often look barely functional. Using a browser in 1995 was a slow, unrewarding experience. The predictions that vast sections of commerce, communication and entertainment would be upended by this clunky technology seemed ridiculously overblown. Yet here we are.

Consequently, it’s important that we ask questions about the impact of new technologies based on their fundamental power, not their current, nascent implementations.

We predict that AI’s impact on coaching will follow the trajectory of most disruptive technologies. AI coaches will not be as good as the “real thing” initially, but they will be far, far cheaper. A service that is currently only accessible to those with corporate support will become increasingly available to pretty much everybody, and over time AI coaches will improve to the point where human coaches will be an elite luxury for those with moneyto burn.


How do we feel about these predictions? In a word: queasy, especially as AI coaches will inevitably be the creation of organizations, very likely corporations, and the history of corporations as moral actors is a dubious one. Whose interests will an AI coach be aligned with? And how will we know about and correct for misalignments?

Nevertheless, the path forward is pretty clear: the technologies will continue to improve incredibly quickly. Humans will become comfortable working with, interfacing with and, yes, having relationships with AI entities.

AI will deliver transformational coaching unless transformational coaching is by definition coaching delivered by a human – which, of course, raises the question of what it means, exactly,to be a human.

One way or another, we’re going to find out.

Tell us what you think about this article.

Let’s continue the conversation by connecting with your colleagues on our Facebook page