The AI wrote the plan. The plan was fine. Nobody internalized it.

That is, more or less, the finding of a study published in PLOS One this past March, and it is the single most commercially useful piece of coaching research to land this year.

Researchers ran pre- and post-interviews with 15 Human Resources leaders from South Korean IT companies who used AI-mediated coaching, then analyzed how those people experienced the three classic components of the working alliance: emotional bond, goal agreement, task assignment.

The findings indicate that AI coaching effectively accelerates goal clarification and provides structured task suggestions, yet it does not generate a coaching presence.

And then the sentence worth printing on a business card: in its absence, interactions were described as transactional and emotionally unresponsive; relational engagement weakened, goal agreement was less co-constructed, and task intentions were weakly internalized.

The bot is good at the part clients think matters

Notice what the AI won on. Speed to clarity. Structure. A tidy list of next actions.

That is exactly the part of coaching that clients believe they're paying for. Ask most buyers what a coach does and they'll describe goal-setting and accountability steps. Both of those are now available for twenty dollars a month at three in the morning.

What the AI lost on is the part clients can't see until they've lived it. Coaching presence is described in the literature as a foundational competence that lets coaches build effective relationships through mindful, attentive, responsive interaction: staying conscious, open, flexible, grounded and engaged, with empathy, emotional self-regulation, comfort with uncertainty, and the intentional use of silence.

Silence. The one thing a chatbot structurally cannot do.

The detail that should worry the junior end of the market

The study found the reaction split by experience level. Coaching-educated and certified participants identified relational deficits early and judged AI more critically, whereas novices initially valued efficiency but later reported a lack of partnership and motivation.

Read that twice. The people who had never been coached liked it first. Then the motivation drained out.

If your offer is essentially a structured goal framework delivered competently, you are competing in the exact category where the novice buyer is happiest with a bot — for a while.

Don't oversell this

The honest version of this story includes the counter-evidence, because clients will find it.

AI coaching does move goals. Research has shown goal-based AI coaching chatbots can produce outcomes comparable to human coaching in terms of goal attainment, and that they can establish a working alliance with coachees. A recent longitudinal randomized experiment found students in the AI-based chatbot condition achieved significantly stronger increases in goal attainment over time compared to students in both control conditions.

The same researchers flagged the limit: while AI-based coaches may be effective in structuring goal-directed behavior, they may be less capable of fostering a strong working alliance compared to human coaches.

So the argument is not "AI coaching doesn't work." It's narrower and sharper than that.

The line for your next sales call

The PLOS One authors land on a conclusion a coach can quote directly: procedural assistance without relational depth struggles to sustain a transformative working alliance.

Their own recommendation isn't replacement either. They point toward emotionally responsive AI coaching platforms with empathic attunement features, adaptive conversational feedback, and hybrid human-AI models.

Which means the coaches most exposed right now are the ones selling plans. The ones selling the thing that makes a human keep a plan have a research paper on their side, and should start saying so out loud.