The AI Plays the Customer. It Doesn't Run the Call.

Part 1 of the series “Practice Makes Progress: Inside Dynamic Syms.”

How SymTrain uses AI in dynamic simulation, and what we never let it decide. Practice that feels real, progress you can measure.

About one second passes between a trainee finishing a sentence and the simulated customer on the other end of the line answering back.

Inside that second, a lot happens: the customer’s mood updates, their memory is checked, the reply is written and spoken, and every word is screened before the trainee hears it. Meanwhile, as the conversation goes on, the trainee may see a short coaching tip appear on their screen to guide them.

Meanwhile, a second model has been watching the whole exchange, scoring what the trainee just said, and deciding whether to put a short tip on their screen. The customer never waits for it. The customer never hears it.

That is what happens on every call in a SymTrain Dynamic Sym. This post is about the idea behind it, because the idea is the part that matters: the AI plays the customer. It does not run the call. And the reason that matters isn’t the technology. It’s what the technology is for: a trainee who practices against a customer like this gets measurably better, call after call. The AI is how we make the practice real. The progress is the point.

What is a Dynamic Sym?

A scripted simulation has authored dialogue: the trainee practices against lines someone wrote. A Dynamic Sym replaces the script with a description.

An author describes a customer: who they are, what happened to them, how quickly they calm down, plus the concerns that customer will raise and what a successful call looks like. No dialogue is written. At runtime, the trainee is in a real-time voice conversation with that simulated customer, who improvises within the scenario: shares details when asked, gets frustrated when handled badly, raises their concerns one at a time, and when the trainee earns it, wraps up and says goodbye.

No two runs are the same, which means nobody can memorize their way through one.

Five Things You Can Count On

These are what make a Dynamic Sym a training tool rather than a chatbot with a voice. Each one is a commitment about how the customer behaves, and each one exists so that practice turns into measurable progress.

1. The customer’s mood has to be earned. A Dynamic Sym customer doesn’t calm down because a trainee said the right phrase, and doesn’t flip from furious to friendly in a single turn. Their mood moves the way a real person’s does, gradually and in response to what you actually do. The AI plays whatever mood the call has earned; it never decides on its own to be forgiving.
2. The customer keeps its story straight. Real customers volunteer details like an account number, a charge, a date. Ours do too, and those details stay true for the whole call. Ask about that charge ten minutes later, and it’s the same charge. Trainees never have to wonder whether the customer “forgot.”
3. The customer can’t hang up on a whim. Every call ends with a spoken goodbye, and every ending is explained in plain language. A trainee who is struggling but still trying gets redirected and coached, not cut off. And if something fails on our side, that call is never held against them.
4. The coach knows when to speak. A trainee who’s struggling gets short, specific guidance while the call is still happening, the kind of tip a good trainer would lean over and give. A trainee who’s doing well is left alone. The customer never hears any of it.
5. Every score shows its evidence. Calls are graded on your scorecard, and every score is backed by what the trainee actually said on that call, checked before it counts, never just asserted. Nothing is adjusted invisibly; reviewers can always see what was scored and why.

Who Controls What

For IT, security, and AI governance teams, the most useful thing to know about any AI product is where the AI’s authority stops. In a Dynamic Sym, the lines are clear.

• What the AI controls: the customer’s words. It improvises what the customer says, in character, inside the scenario it was given.
• What the AI doesn’t control: how the customer feels, whether a concern counts as resolved, when the call ends, or the final score. Those behave the same way on every call, and a technical problem on our side is never scored against a trainee.
• What you control: the scenario, the goal, the customer’s temperament, the scorecard calls are graded on, and the guardrails the customer stays inside. Nothing runs that you didn’t author.

What This Means for You

• If you’re a trainee, it means the hardest caller you’ll ever face can be the one you practice on, with a coach beside you and an ending that’s always explained.
• If you’re a coach or trainer, it means every practice call already has a coach that steps in when someone struggles and stays out of the way when they don’t, so your time goes where it’s actually needed.
• If you run quality, it means grades on your own scorecard, backed by what the trainee actually said and verified before they count, evidence you can calibrate against and put in front of a client.
• If you run a contact center, it means unlimited, realistic practice without a human role-play partner, for every new client program, without writing a script tree.
• And if you author simulations, it means you describe a customer and a goal, and the conversation writes itself differently every time, while you keep the controls that matter, including how strong the customer’s language gets.

Where This Series Goes Next

Over the coming months we’ll take each of those one at a time: what practice actually feels like for a trainee, how coaches catch the coachable moment in every practice call, what “scores with receipts” looks like in a real quality review, why practice doesn’t end at onboarding, and what authors control when there’s no script to write. And for IT and security teams, a full post on who controls what, the short version above, in depth.

Every one of those is the same idea from a different angle: the AI brings the realism, and the rules make it fair. That combination is what turns a convincing conversation into practice that produces measurable progress, and it’s what we’ll keep coming back to in this series.

Part 2 available now “Practice Before It’s Real”.