The Conversation They Never Practiced
The Conversation They Never Practiced
Ten brilliant strategy consultants have just joined me in a beautiful meeting room outside Munich. They've cracked cases that saved their clients hundreds of millions. They can dismantle a competitor's strategy in twenty minutes flat. They are about to become partners.
And they are terrified.
Not of the strategy work. Not of the client pressure. Of selling.
I run this program for global consulting firms. Every cohort is the same — around ten people, flown in from offices around the world, each one a proven problem-solver. My job is to take away the fear and help them find joy in selling. I teach them relational sales frameworks. I bring in CEOs I know personally to roleplay with them. Real executives, real pushback, real stakes.
And it works. In that room, something shifts. They start to see that selling isn't performing — it's listening, asking questions, and solving. The energy changes.
Then they fly home. And that's it.
No more practice. No more feedback. The next CEO sitting across their table - or screen - will not only push back but decide whether to hire the firm or not. The consultants go from the most intensive learning experience of their careers back to figuring it out alone, in real conversations, with real revenue on the line.
This is the dirty secret of executive education: the learning ends exactly when the practicing should begin.
We spend enormous amounts of money teaching leaders what to do. We spend almost nothing helping them practice doing it.
The research on this is not ambiguous. A meta-analysis published in the Journal of Applied Psychology — the largest ever conducted on leadership training, covering 335 studies and over 26,000 participants — found that practice-based methods produce 78% larger effects on actual on-the-job behavior change than lectures and presentations (Lacerenza et al., 2017). Not 78% better test scores. 78% more change in what leaders actually do when they're back at work.
The authors were blunt in their conclusion: if you can only choose one delivery method, choose practice.
A second meta-analysis, also in the Journal of Applied Psychology, examined the specific training model that roleplay is built on: observe a behavior, practice it yourself, get feedback. Across 117 studies, this loop produced effect sizes above 1.0 — a full standard deviation of improvement. More importantly, the skills held up over time. Knowledge from lectures decayed. Skills from practice didn't (Taylor, Russ-Eft & Chan, 2005).
This isn't new science. We've known this for decades. We just haven't acted on it — because practice, at scale, has been logistically impossible in executive education.
Here's what a typical two-day leadership program looks like from a practice standpoint:
A participant gets maybe 15 to 20 minutes of actual practice time. The rest is lectures, case discussions, group exercises, and coffee breaks. The ratio of theory to practice is inverted from what every piece of learning science recommends.
And those 15 minutes? It happens once. One attempt at a difficult conversation. One round of feedback. Then it's over.
The research on skill acquisition is clear: competence requires repeated practice with feedback across varied conditions. A single roleplay in a leadership program is the equivalent of playing one tennis match and expecting to compete at Wimbledon.
Meanwhile, on the sales training side, this problem has already been solved. AI-powered roleplay platforms have raised hundreds of millions in venture capital. Mursion. Mindtickle. Second Nature. Hyperbound. Every one of them targets the same buyer: sales teams.
That leaves the most consequential conversations in business — the ones between leaders and their people, between partners and their clients, between executives and their boards — almost entirely unserved.
What changed my thinking was a study published in Science in September 2024.
Researchers at MIT and Cornell ran an experiment with over 2,000 participants who held deeply entrenched beliefs — the kind of beliefs that decades of psychology research said were essentially immune to evidence. They had each person articulate their specific reasoning, then engaged them in personalized dialogue with an AI that directly addressed their individual arguments.
The result: a 20% reduction in those deeply held beliefs, sustained for at least two months. The effect worked across every category tested. It worked even for people whose beliefs were central to their identity. And when a professional fact-checker evaluated the AI's claims, 99.2% were accurate (Costello, Pennycook & Rand, 2024).
The mechanism that made it work wasn't the AI itself. It was the personalization. Generic counterarguments had failed for years. What succeeded was a dialogue that addressed the specific evidence each person brought to the table.
That is exactly what practice-based leadership development requires. A leader preparing for a difficult performance review doesn't need a generic lecture on feedback. They need to practice that specific conversation, with an AI that responds the way their specific team member is likely to respond — defensive, deflecting, emotional, or silent. And they need to do it five times, not once.
If you run leadership development programs and you're still relying on a single roleplay session with actors or peers, consider what's happening after your participants leave the room.
They face the actual conversation — the termination, the restructuring announcement, the board pushback, the sales pitch to a skeptical client — with one attempt under their belt. In most cases, zero.
The technology to change this now exists. AI-powered roleplay lets every participant practice difficult conversations as many times as they need, with varied personas, with immediate feedback, at any hour, from anywhere. The same quality of experience whether they're in London, Tampa, or Singapore.
This isn't about replacing what happens in the room. The cohort experience, the human facilitation, the expert instruction — all of that matters. This is about what happens after the room. The weeks and months during which skills either develop through practice or atrophy through disuse.
The evidence says practice is what works. The question for L&D leaders is simple: Are you going to keep hoping that 15 minutes is enough?
I'm building this at RolePlays.ai — working with institutions and corporations to bring AI-powered roleplay to executive education. If you're exploring this for your programs, I'd welcome the conversation.
References
Lacerenza, C. N., Reyes, D. L., Marlow, S. L., Joseph, D. L., & Salas, E. (2017). Leadership training design, delivery, and implementation: A meta-analysis. Journal of Applied Psychology, 102(12), 1686–1718.
Taylor, P. J., Russ-Eft, D. F., & Chan, D. W. L. (2005). A meta-analytic review of behavior modeling training. Journal of Applied Psychology, 90(4), 692–709.
Costello, T. H., Pennycook, G., & Rand, D. G. (2024). Durably reducing conspiracy beliefs through dialogues with AI. Science, 385(6714), eadq1814.