Just weeks after unveiling Synthetic Customers, Cresta is extending the technology into human agent training. The company's new Training Simulator uses AI-generated customers to create realistic practice conversations, allowing agents to rehearse difficult interactions before speaking to real customers.
What the product does
Training Simulator builds its simulated customers from a company's own conversation history, the same underlying approach behind Synthetic Customers, rather than following scripted role-play paths. These AI-generated customers push back on weak answers, express frustration when responses feel generic and adapt dynamically to whatever the agent actually says. Training leads can assemble scenarios from prompts or generate them automatically from past interactions, and simulated conversations are graded against the same AI-driven quality criteria used to assess live calls. In practice, this lets agents repeatedly rehearse sales pitches, support queries and retention saves until they've demonstrated the behaviours they'll be judged on once they're handling real customers.
Ping Wu, chief executive of Cresta, said the aim is to make practice for high-stakes conversations available on demand rather than rationed through a manager's schedule, arguing that difficult calls agents once rehearsed rarely can now be run "anytime, as many times as it takes to get it right."
Why it matters
As organisations deploy both human and AI agents side by side, the need for safe environments to train staff and test AI systems before they reach customers is becoming increasingly pressing. Training Simulator shows Cresta's synthetic customer technology evolving beyond analytics and AI testing into workforce enablement, positioning simulated customers as a foundational capability for both human and AI workforces rather than a single-purpose tool. The shift also blurs traditional organisational boundaries, as technologies originally designed to test AI agents increasingly become tools for workforce training and operational readiness. It arrives at a time when businesses are already grappling with new questions of ownership and responsibilities due to AI technologies.
From testing to training
Cresta’s announcement at the end of May demonstrated that synthetic customers could represent the next evolution of CX analytics, enabling organisations to model and predict customer behaviour more effectively. Training Simulator suggests Cresta is already broadening that vision, using the same simulated customers not only to test AI agents but also to train the humans who work alongside them. As contact centres become increasingly hybrid environments of humans and AI, simulated customers may become critical infrastructure, provided they can demonstrate a meaningful impact on customer experience and business outcomes.