Sooner or later, every sales leader we talk to asks the same question. People like practicing with AI, they say. But do they actually sell better afterwards?
It is a fair question. Practice that feels good and practice that changes the number are not the same thing, and most of what has been written about AI role-play so far is about the first. Now there is a real answer to the second.
What the study did
Four researchers at the University of Houston — Johannes Habel, Michael Ahearne, Seshadri Tirunillai and Amy Vandaveer Novak — looked at almost 2,000 salespeople at a large industrial company in the United States. Some of them practiced with an AI customer that pushed back and gave feedback. The others did not.
The part that matters is what was measured. Not simulation scores, not satisfaction surveys, not how many sessions people completed. The researchers tracked real job performance over time, which is the only measure a sales leader actually cares about.
What it found
On average, performance rose by about 2%.
That sounds small. Across a whole sales force it is a lot: two percent on every quota, every territory, every quarter.
Two percent is the baseline. It is what AI role-play in general can do, before anyone builds the practice around your company.
The baseline is a standard scenario
So what happens when the practice is built around your company instead of a standard scenario? When the customer on the other side raises the objection your team hears every week, and the feedback measures the things your own best people are judged by?
That is what we built Simpanion for. Your team practices your own cases against your own criteria. The counterpart comes from several specialized models, a case can be lost, and it rarely plays out the same way twice. That is the kind of practice we believe moves performance much further than the baseline.
What this means if you run a sales team
- AI role-play works, measured on real performance rather than on how it feels. The question is no longer whether to practice this way.
- The effect in the study is what AI role-play in general delivers. It is the baseline to build on, not the ceiling.
- Practice compounds. A small lift on every conversation, every quarter, is what the business is judged on at the end of the year.
- Practice that mirrors your own cases, your own criteria and your own difficult counterparts is where we believe the larger gains are.
Simpanion — train the human edge. Your Simulation Companion lets your team train whenever they want and gives instant, structured feedback. Built and hosted in Germany.