When AI Coaching Falls Short: What New Research Means for Executive Development
By Scott Serviss | COO Berman Leadership Development
The question reshaping executive development right now isn’t whether AI can coach. It clearly can — to a degree. The real question is whether it’s good enough, and whether the cost and scalability advantages justify the tradeoff in outcomes.
For HR leaders, L&D professionals, and CHROs under constant pressure to democratize development while managing budgets, this feels like a genuinely difficult call. A new randomized controlled trial published in Human Resource Development International by de Haan, Terblanche, and Nowack (2026) offers some of the most rigorous evidence yet — and the findings deserve serious attention.
What the Research Actually Found
This wasn’t a survey or a case study. It was a three-arm RCT with 114 leaders, randomly assigned to:
-
Human coaching
-
AI chatbot coaching
-
A waitlist control group
Participants were tracked across multiple validated outcome measures: goal attainment, self-efficacy, perceived stress, wellbeing, and coaching effectiveness. Every participant — including dropouts — completed all questionnaires, eliminating one of the most persistent biases in coaching research.
The headline finding: Human coaches produced significant improvements across virtually every outcome measure, with effect sizes in the mid-to-high range. The AI chatbot group, by contrast, did not score significantly differently from the control group on most variables. That’s not a marginal gap. That’s a substantial one.
Perhaps equally telling: attrition in the AI group ran roughly ten times higher than in the human coaching group — even with regular reminders and researcher support. Leaders didn’t disengage from coaching. They disengaged from AI coaching. Several who dropped out of the AI group specifically requested a human coach instead.
So Is AI Coaching Good Enough?
The honest answer: it depends entirely on who you’re coaching and what you need them to accomplish.
The research does suggest contexts where AI coaching may provide real value — particularly for earlier-career employees and individual contributors who:
-
Have limited access to human coaches
-
Are working on discrete, bounded goals
-
May benefit from the availability and consistency a digital platform provides
In this space, AI coaching can be a meaningful democratizing tool.
But senior leaders are a different proposition entirely.
The complexity of the executive role — navigating board dynamics, leading through organizational ambiguity, managing high-stakes interpersonal conflict, building judgment that can only be developed in dialogue — requires something AI systems are not yet designed to provide. The de Haan study authors put it plainly: even the most advanced chatbots remain too superficial and forgetful, unable to track what’s truly relevant in a relationship or emulate the quality of human engagement over time.
The Co-Regulation Factor
One of the most important concepts in this research is co-regulation — the dynamic through which coach and coachee actively influence each other, building momentum, trust, and leverage toward meaningful outcomes. This mutual attunement is what makes coaching uniquely powerful compared to other development interventions.
Co-regulation depends on a human relationship. It requires:
-
Reading subtle cues
-
Holding shared memories across sessions
-
Responding to what’s actually happening in the room — including what isn’t being said
These are not features an AI platform can replicate. They are the mechanism of change.
The research also confirmed that readiness factors — a coachee’s initial hope, self-efficacy, cognitive hardiness, and lower perceived stress — significantly predict coaching outcomes. Good coaches don’t just respond to these factors; they actively develop them over the course of an engagement. This is the virtuous circle that separates transformative coaching from a series of useful conversations.
What This Means for Your Organization
For CHROs and talent executives making investment decisions, the implications are practical:
Where AI coaching tools can responsibly extend your reach:
-
Onboarding programs
-
Goal-setting support for high-potential contributors
-
Structured reflection tools for managers early in their development journey
In these contexts, the scalability and cost advantages are real, and the outcome tradeoff may be acceptable.
Where human coaching remains essential:
Your C-suite. Your executive team. Your high-stakes succession candidates.
The evidence argues clearly for human coaching at the senior level. The return on investment from executive coaching is well-documented, the effect sizes are meaningful, and the thing that makes it work — the human relationship — cannot be approximated by a well-designed chatbot.
The question was never whether AI is better. At senior levels, the question of whether it’s good enough now has a rigorous answer: not yet.
At Berman Leadership Development, we work at exactly this intersection — evidence-based coaching for leaders where the stakes are high and the complexity is real. We believe AI tools have an important role to play in making development more accessible across organizations. And we believe the human coaching relationship remains irreplaceable at the top.