Every time I talk with learning and development leaders about AI-driven role-play simulations, the conversation eventually lands in the same place.
"So, are these training, or are they assessments?"
It's a fair question, but I think it's the wrong question.
AI-driven role-play simulations are quickly becoming one of the most talked-about applications of generative AI in workplace learning. Organizations are looking beyond content delivery and asking how AI can help people build skills through realistic practice. That shift has made simulations a hot topic, but I think we've framed the discussion too narrowly.
I've actually been having this conversation for years, long before generative AI entered the picture.
Years ago, we built a series of medical coding simulations for one of the largest healthcare organizations in the country. The goal was simple. Give billing coders realistic patient scenarios to work through instead of having them read about coding rules or answer multiple-choice questions.
Their concern was that the moment learners saw a score, they would focus on passing instead of learning. They wanted people to explore, make mistakes, and try again without worrying about how they performed. So we created what we called sandbox simulations. They looked and behaved exactly like the graded versions, but there was no score. The only objective was to practice.
It was exactly what they wanted.
A few months later, the conversation changed.
Now they wanted to know who was ready to work independently. They wanted to identify who needed coaching and who consistently demonstrated good judgment. They asked if the same simulations could also provide meaningful evaluation data.
Simulation Role-Play within the RemoteBridge platform providing realtime coaching as well as graded reporting like an assessment.
Nothing had changed about the simulations. The only thing that changed was where the learners were in their journey.
That experience has stayed with me because it illustrates a mistake we still make today. We tend to think of learning and assessment as two separate activities. First we teach. Then we measure.
I don't think that's how people actually develop expertise.
Think about learning to lead a difficult performance conversation. Handling an upset customer. Interacting with colleagues of a different background. Managing a sales objection.
Every response tells us something about the learner's capabilities.
At the same time, every response gives the learner another opportunity to improve.
That's what makes AI-driven role-play simulations so interesting. They don't simply automate role-play. They allow learning and assessment to happen at the same time.
There are absolutely times when learners need a safe environment to experiment without feeling evaluated. In those situations, removing the score encourages curiosity, confidence, and repetition. Learners focus on improving instead of performing.
There are other situations where organizations need evidence that someone is ready. Before a manager conducts a difficult employee conversation, before a salesperson handles a strategic account, or before a healthcare professional works independently, leaders need confidence that the necessary skills have been demonstrated.
The same simulation can support both objectives.
Early in the learning process, it becomes a safe place to practice.
Later, it becomes evidence of competence.
AI makes that transition much easier than it has ever been before. Every role-play can provide feedback to the learner while also providing meaningful insights to managers. Instead of relying on a single quiz score, organizations can begin looking at patterns over time. Is the learner improving? Are they asking better questions? Are they adapting when the conversation changes? Are they demonstrating empathy, sound judgment, and confidence?
Those are much more meaningful indicators of performance than whether someone selected the correct answer on a multiple-choice assessment.
My answer is the same today as it was when we built those medical coding simulations years ago. They're both.
More importantly, I don't think we should force organizations to choose between them.
For decades, we've designed learning first and assessment second because that's what the technology allowed. AI gives us another option. We can create learning experiences where design learning where every conversation helps someone improve, and every conversation also tells us something meaningful about their readiness.