There is a moment in sales training that most programs quietly skip.
A rep knows the product. They have read the messaging. They may have completed the onboarding course, passed the quiz, and sat through a few coaching sessions.
Then a customer says something unexpected.
“We already have a provider.”
“Your competitor is cheaper.”
“We don’t have the budget.”
“Just send me the information.”
And suddenly the rep is no longer recalling training. They are trying to think on their feet.
That is where sales training gets real.
The problem is that practicing this part of selling is surprisingly difficult. A manager can role-play with a rep, but managers have forecasts to review, deals to inspect, one-on-ones to run, and a team to manage. Asking them to become a practice partner for every difficult conversation does not scale.
This is one reason AI roleplay has become interesting to sales organizations. Practis describes the problem directly: managers do not have time to run roleplays with every rep, while new hires can end up learning on real customer calls. Its AI Roleplay product is designed to give reps a private environment for practicing conversations, receiving feedback, and repeating difficult scenarios without waiting for a manager.
But the more interesting question is not whether AI can role-play.
It is whether repeated practice can solve a problem that traditional sales training has struggled with for years: getting knowledge to show up when the conversation becomes uncomfortable.
The manager is not the bottleneck. The practice model is.
I have seen this problem described as a coaching-capacity issue, and that is partly true.
A manager might have ten, fifteen, or more reps. Even if each rep needs only one meaningful role-play session a week, the hours add up quickly.
And there is another problem.
Not every rep wants to raise their hand and say, “I need to practice handling price objections.”
Some people are uncomfortable role-playing with their manager. Others know they are weak in a particular area but avoid exposing that weakness. Experienced sellers can be especially resistant when role-play feels like a classroom exercise rather than preparation for something they actually have to do.
That creates an awkward situation.
The people who need the most repetitions may be the least likely to ask for them.
AI changes the economics of that practice.
With an AI roleplay system, the rep does not need to wait for a manager, another rep, or a scheduled training session. Practis, for example, lets salespeople practice scenarios on demand and receive feedback after the session.
That does not make the manager unnecessary.
It changes what the manager spends time doing.
Instead of spending 30 minutes playing the customer so a rep can rehearse the same objection five times, the manager can spend that time discussing why the rep is struggling, reviewing performance, or working through a more complex coaching issue.
That distinction matters.
Practice is different from training
One of the biggest mistakes in sales enablement is treating exposure to information as evidence of skill.
A rep can watch a video about discovery.
They can read a document about objection handling.
They can complete a quiz.
None of those activities necessarily proves that they can handle a buyer who interrupts them halfway through an answer.
Sales training research has found evidence that the way practice is structured affects transfer to the job. One field study comparing spaced and massed sales training found that spaced practice produced better transfer quality and higher self-reported sales competence than massed practice.
There is also older research specifically examining sales training cycle time that found properly preparing trainees before role-play could improve training efficiency and initial revenue-generating potential.
The lesson is not that “more role-play fixes sales.”
It is more specific:
Practice has to be designed as part of the learning process.
That means the question for a sales leader shouldn’t be:
“Do we have sales training?”
It should be:
“Where does a rep actually get to try the behavior?”
The difficult part is not the first attempt
Imagine a new AE is preparing for an enterprise discovery call.
The manager says:
“Let’s practice.”
The manager becomes the buyer.
The rep gives an opening.
The manager raises an objection.
The rep responds.
They discuss it.
Then they move on.
That can be useful.
But something is missing.
The rep may need to hear the same objection several times.
They may need to answer it badly once.
Then try again.
Then discover that their first response was too defensive.
Then try a different approach.
Then encounter a completely different version of the same objection.
That is where practice starts to resemble what happens in a real conversation.
Practis customer evidence provides an interesting example. In one customer story, a learner described repetition as helping them work through their words so that when a customer asked a question, the response was already available to them. The same case-study collection describes organizations using regular role-play to improve customer-facing performance.
That is a much more useful description of practice than simply saying a rep “completed training.”
The objective is not memorizing a perfect sentence.
It is reducing the amount of cognitive effort required to respond effectively.
The best practice is not necessarily the most realistic practice
There is a temptation in AI roleplay to jump straight into the hardest possible simulation.
That can be a mistake.
If a rep does not know the basic messaging, throwing them into an aggressive, unpredictable buyer conversation may create frustration rather than learning.
Practis has built its training approach around a progression it calls “Script-to-Scrimmage.” The first stage focuses on practicing core language and messaging. The second moves into more open-ended conversations with AI personas that push back and adapt to the rep’s responses.
That sequencing makes sense.
Think about sports training.
A baseball player does not start by facing the hardest pitcher in the league.
They work on the movement.
Then timing.
Then increasingly difficult pitches.
Then game situations.
Sales conversations have the same progression.
A rep might first practice:
“How do I explain our value proposition?”
Then:
“How do I respond when the buyer asks about price?”
Then:
“What if the buyer says they already have a competitor?”
Then:
“What if they interrupt me and challenge the business case?”
The difficulty can increase as the rep becomes more comfortable.
Practis Practice Sets are designed around this type of structured progression, allowing teams to organize scenarios into practice paths and assign them based on role, tenure, skill level, or performance gaps.
That is an important distinction between having an AI chatbot and having a sales practice system.
What should managers actually do?
This is where I would be careful about the phrase “AI replaces role-play.”
It does not have to.
A better model is:
AI handles repetition.
Managers handle judgment.
A rep can practice the same price objection ten times with an AI customer.
The manager does not need to listen to all ten.
But if the rep repeatedly responds by discounting too quickly, that is worth a conversation with the manager.
The manager might ask:
“Why did you go straight to discounting?”
“What did you hear in the buyer’s question?”
“What would happen if you asked one more question first?”
That is coaching.
The practice created the evidence for the coaching conversation.
Practis describes its manager workflow in similar terms, giving managers visibility into practice performance and skill gaps so coaching can become more targeted
This is potentially more valuable than simply saving manager hours.
It can change the quality of the coaching conversation.
Instead of:
“How are you feeling about objection handling?”
The manager can ask:
“I noticed you handled the budget objection well, but when the buyer challenged the competitor comparison, you moved into a product explanation. What happened there?”
That is a much better coaching conversation.
There is a catch: bad practice can scale too
This is the part that gets missed in enthusiastic discussions about AI sales training.
Making practice available 24/7 is not enough.
If the scenario is poorly designed, the rep can practice the wrong behavior repeatedly.
If the scoring is weak, the rep can receive misleading feedback.
If the buyer personas are unrealistic, the rep may become good at beating the simulation rather than having better conversations.
Practis itself makes this point in its sales training material: the technology is only part of the problem, and scenario quality, content, manager buy-in, and ongoing monitoring matter significantly.
That should be the standard for evaluating any AI roleplay system.
The question isn’t:
“Can the AI talk to my reps?”
The question is:
“Does the practice reflect the conversations my reps actually have?”
That means using real objections.
Real messaging.
Real buyer concerns.
Real competitive situations.
Real mistakes.
And ideally, real evidence from the field.
A better way to introduce AI roleplay
If I were designing a program for a sales organization, I would not start with 100 scenarios.
I would start with the five conversations that are costing the team the most.
For example:
- The first 30 seconds of a cold call
- “We’re happy with our current provider”
- Price objection
- Discovery when the buyer gives short answers
- Asking for the next step
Then I would look at what actually happens.
Which scenario produces the weakest performance?
Where do reps repeatedly get stuck?
Which objections produce inconsistent responses?
Which behaviors improve after several attempts?
That gives the enablement team something much more valuable than completion data.
It gives them a map of where the organization needs practice.
From there, the practice program can expand.
Practis supports this kind of structured approach through Practice Sets, where scenarios can be organized into defined practice paths and assigned to specific people or teams.
The goal isn’t to make every rep spend hours talking to an AI.
The goal is to give each rep enough targeted repetition around the conversations that matter.
What about experienced salespeople?
This is where the argument gets more interesting.
A senior rep probably does not need to practice:
“What is our company?”
They may need to practice:
“How do I challenge a CFO who says the project is too expensive?”
Or:
“How do I respond when procurement demands a 20% discount?”
Or:
“How do I recover when an executive tells me the business case isn’t compelling?”
The more experienced the seller, the more useful the scenarios can become.
Practice stops being about learning the script and becomes preparation for difficult situations.
That is also why AI roleplay can potentially work as an ongoing development tool rather than only a new-hire tool. Practis describes use cases spanning SDRs, AEs, account managers, new-hire onboarding, objection handling, ongoing skill development, and manager coaching.
So, can AI solve the manager role-play problem?
Partly.
But I would frame the result differently.
AI does not solve the need for sales coaching.
It solves one of the reasons coaching is difficult to scale: the sheer amount of repetition required to build conversational skill.
That distinction is important.
A manager should not have to spend their best coaching hours pretending to be the same skeptical buyer for the sixth time.
A rep should not have to wait until next Tuesday’s one-on-one to practice a conversation they are struggling with today.
And a new hire should not have to discover whether they can handle an objection for the first time in front of a real customer.
That is the gap AI roleplay is most interesting for.
Practis calls its approach “practice in private, perform in public.” The idea is simple enough: create a place where reps can make mistakes before those mistakes become customer experiences.
The real test, however, is not how impressive the AI sounds.
It is what happens afterward.
Do reps respond better?
Do managers coach more specifically?
Do new hires become ready sooner?
Do difficult conversations become less intimidating?
Can leaders see where skill gaps actually exist?
Those are the questions worth measuring.
And that is ultimately where AI sales roleplay has to earn its place.
Not in the demo.
In the conversation that happens after the demo.
A practical starting point
If your managers do not have time to role-play with every rep, don’t try to create more manager role-play sessions.
Start by identifying the five conversations your team most needs to get better at.
Turn those into realistic scenarios.
Give reps repeated opportunities to practice them.
Use performance data to identify where they struggle.
Then have managers spend their limited coaching time on the patterns that practice reveals.
That creates a much healthier relationship between technology and coaching.
The AI provides the repetitions.
The manager provides the judgment.
The rep gets the practice.
And, ideally, the customer gets a better conversation.
That is a far more useful way to think about AI sales training than simply asking whether AI can “replace role-play.”
It is really asking a different question:
What would happen if every salesperson could practice the hardest conversation of their week before having it with a customer?