Introducing AI Roleplay: Practice Under Pressure Before the Customer Is Real

Every sales leader in home improvement knows the gap. A rep can recite the pitch in the office, and then a homeowner at a kitchen table says the quote is four thousand dollars higher than the other company's, and reciting is suddenly worth nothing. Handling pressure out loud is a different skill from knowing the answer, and the only place most reps get to practice it is in front of a real customer, at the cost of a real deal.

The traditional fix is roleplay in the weekly meeting: one rep performs, everyone else watches, and the trainer has time for maybe two rounds. It works, and it cannot possibly scale to every rep, every objection, every week.

So we built the practice partner that is always available. AI Roleplay has been live since July; today we are telling the full story. One team we work with rolled it out across a sales floor of more than two hundred reps. Sessions run about five minutes, so practice fits between appointments instead of waiting for the next training day.

Roleplay has always worked. The constraint was never the method, it was the logistics: a practice partner who knows your process, has the time, and will push back like a real customer is the scarcest resource in the building.

Open Practice in the Craft app and pick a scenario. Your phone starts a live call, and the voice on the other end is the customer: a homeowner who wants to think about it, a spouse who is not home, a price shopper with a competing quote. You answer out loud, in real time, the way you will have to at the door. When the call ends you get a scored recap against your company's sales process, step by step, with the moments worth another run.

Scenarios come in difficulty tiers, so a new rep starts against a friendly rookie-level customer and works up to the advanced one who has heard it all before.

The reason practice libraries go stale is that authoring is work. In Craft, a coach types one line, like "create a single decision-maker objection scenario," and Craft drafts the whole thing: the customer persona and backstory, the setup the rep sees, and a scorecard mapped to your sales process rather than a generic rubric. Edit anything, publish, and reps can practice it that afternoon. Scenarios can also start from the conversations Craft has already recorded, so the objection your team keeps hearing this month becomes this month's drill.

That scoring detail is the whole product in miniature. A rep who passes a Craft roleplay did not get good at roleplay in general; they executed your process, in your company's language, under pressure.

Single scenarios are for sharpening. For onboarding and standards, coaches sequence them into courses: clear each scenario in order, pass at the score you set, retry until you do. Assign a course to the team and the dashboard shows who is certified, who is practicing, and who needs attention before they walk into another living room.

New-hire ramp gets a definition managers can stand behind: nobody meets a real customer until they have closed the AI one.

Team readiness7 of 10 certified

Marcus RiveraCertified91%

Danielle FosterCertified84%

Tyler BrooksNeeds practice61%

Evan DelgadoNot started–

A large home improvement operator built their new-hire training on AI Roleplay. Before, their training manager played the customer in every practice round and graded it at the same time. In his words, doing both at once "takes us away from fully concentrating on it." Each trainee got half a scene partner and half a grader, and the rest of the class watched.

Now the assignment fits on a whiteboard: "Today's homework: intro, D personality." That evening every trainee runs the intro against an AI homeowner with a dominant, interrupting personality, built on the company's own sales process, scoring the exact behaviors the trainer chose. He reviews the attempts and scores the next morning, trainees repeat the scenario until they clear it, and nobody runs a real appointment until they have certified on the course.

His first class all struggled against the high-D customer who interrupts the process, demands a price, and marches the rep straight to the rooms. That is the point: the difficult customer exposed a skill gap the friendly scenarios never would. Trainees reran it, watched their scores climb, and kept practicing outside class. The trainer's review: "They loved it. They love practicing with it."

The high-D scenario also shows why scoring on intent matters. A rep who deliberately adapts the process to keep control of that customer is doing it right, and Craft scores the intent of each step rather than matching a script, so adapting well and losing control read as the different things they are.

New reps arrive at their first real appointment having already survived the hard conversations a dozen times, and the confidence shows up at the door with them. Every appointment run by a rep who was not ready is a lead your marketing budget paid for; certification puts a floor under that. Trainer and manager hours go to reviewing scores and coaching the moments that need it, because playing the customer all day is no longer the job. Veterans drill the one objection that cost them last week instead of sitting through training built for beginners. And managers get a readiness answer that used to be a gut feeling: certified, practicing, or needs attention.

Generic sales training rehearses generic selling. Craft roleplay is built from your playbook and grounded in your real conversations, and it is scored the same way your real appointments are scored. Practice and field performance stop being two separate worlds; the same process, the same scorecards, the same language run through both.

And because practice is a five-minute voice call instead of a scheduled event, it actually happens. The rep in the car outside their 2 PM can run the price objection one more time before knocking.

How fast can a coach get the first scenario live?

Minutes. Describe it in a line, review the draft, publish. The practice library grows as fast as you can describe what your reps should rehearse.

Does it work for contact center agents too?

Yes. The same practice loop covers booking calls: agents rehearse against an AI caller and get scored on your booking process, the same way reps rehearse the in-home appointment.