Cresta, the unified customer experience (CX) AI platform, has announced the launch of Synthetic Customers, a new capability designed to create realistic customer personas derived directly from an enterprise’s historical conversation data.
The tool aims to bridge the gap between theoretical customer profiles and real-world behavior by mining existing call, chat, and email logs. These AI-generated personas are intended to power more accurate simulations for testing AI agents and training human staff, while providing the business with deeper insights into customer patterns.
According to Cresta, traditional customer personas are often built from siloed sources such as surveys, CRM records, and support tickets. The company argues that these methods can result in outdated or incomplete profiles that reflect how a business thinks customers behave, rather than how they interact in real time. Synthetic Customers are designed to evolve alongside the customer base, providing evidence-based personas that reflect current market trends.
Ping Wu, CEO of Cresta, said:
“The data that enterprises need to build accurate customer personas for better testing, training, and decision-making is right in front of them, at their finger tips. It lives in every call, chat, or email conversation. With Synthetic Customers, enterprises can finally put all that data to work to truly understand their customers, prepare for how they actually behave, and serve them better across every channel.”
The technology works by extracting language patterns, specific actions, and emotional cues—such as frustration, impatience, and unpredictability—from historical data. To ensure the realism of these simulations, Cresta utilizes a blind evaluation methodology to determine whether humans or AI can distinguish the synthetic interactions from genuine customer conversations.
Synthetic Customers unlock new ways to understand, test, and prepare for customer interactions across the business:
- AI agent simulated testing: Validate AI agents against Synthetic Customers that reflect how real customers actually behave, before a single live interaction happens, and continue to optimize throughout the AI agent lifecycle. This capability dramatically expands testing coverage, surfacing more edge cases to ensure AI agents are even more reliable. Synthetic Customers integrate directly into Cresta’s AI Agent automated testing suite, giving teams a faster, more realistic path to launch confidence.
- Human agent training: With Synthetic Customers, agents can prepare for what actually happens on the job. Synthetic Customers enable dynamic role play training to onboard agents faster, develop specific skills in precise situational contexts, and facilitate practice exercises that mirror real customer interactions to improve sales, service, and retention.
- Simulated customer interviews: Anticipate how your customer base might respond to a business update, like a product change, new policy, or messaging shift, before it goes live.
- Customer insights: Understand how different customer personas and segments behave and what they’re likely to need next, based on real historical patterns. Surface actionable AI recommendations to improve the customer experience before issues repeat.
Beyond training and testing, the company suggests the technology could have broader applications.
Wu added:
“What we’re announcing today is just the start; accurate AI representations of customers can be used for everything from incident planning to market research. The use cases for Synthetic Customers are endless.”
Safety first
As the CX industry rushes to deploy autonomous AI agents, the ability to stress-test these bots against realistic, unpredictable synthetic users before they go live is a critical safety net. Cresta’s technology essentially turns a company’s vast archives of data into a proactive laboratory for CX innovation, potentially shortening the feedback loop between identifying a customer trend and preparing the workforce to handle it.