ASAPP’s Update to GenerativeAgent Increases Transparency, Trust, and Control at Scale

ASAPP, the leading provider of AI-powered contact center software, is announcing new enhancements to GenerativeAgent—ASAPP’s AI customer service agent. Designed to improve accuracy, control, and trust in AI-automated conversations, ASAPP’s latest iteration of GenerativeAgent brings the precision and predictability to AI in the contact center that invites confidence at scale.

GenerativeAgent is engineered to handle complex, multi-turn conversations through both voice and text chat. The solution integrates with a company’s historical data, resolving customer interactions guided by enterprise-grade safety, security, and control.

“We believe that enterprise AI, yes, of course, should be powerful—but it also needs to be controllable, testable, auditable, and ever improving,” said Devidas Desai, senior vice president of product management at ASAPP. “We are bringing precision and predictability to the AI in the contact center…we believe if our customers are not deploying with control, they are not ready to deploy at all.”

To drive greater AI agent control, transparency, and trust, ASAPP is introducing three new features: Human-in-the-Loop Agent (HILA) with Approver Mode, conversation monitoring and fine tuning, and testing and simulation.

Human-in-the-Loop Agent (HILA) with Approver Mode helps develop confidence in AI-automated responses, allowing a human user to review and approve AI outputs in real time or asynchronously. With the agent learning over time, Approver Mode drives faster resolutions and better outcomes, according to ASAPP.

“Approver Mode is a mode where…some customers, until they have confidence in the generative agent running by itself, they would want to monitor every response that it's coming up with and then approve it or change it,” explained Desai. “This is also a way for your generative agent to get educated and essentially get fine-tuned, because these responses—when they are edited by a human-in-the-loop—it's basically making your generative agent better, and it's learning with respect to what are appropriate responses.”

With new conversation monitoring and fine-tuning capabilities, GenerativeAgent delivers full visibility into AI interactions, paired with intuitive observability tools. Users can flag anomalies, track patterns, and enforce compliance with configurable guardrails at scale.

Finally, new testing and simulation functionalities allow enterprises to safely test AI behavior in simulated environments before release updates are pushed to production. This way, users can ensure that new features don’t degrade GenerativeAgent's performance, or surface unexpected behaviors for the end user.

Testing and simulation “really make our customers and our external users very comfortable with respect to making changes to the GenerativeAgent and testing it before anything hits their customers,” said Desai.

Fundamentally, “With conversational AI, it is not very uncommon for customers to feel like, ‘Okay, I have a solution, but it's more like a black box. I don't know how exactly it's doing what it is doing,’” Desai explained. “Something that ASAPP does, I would say very differently as compared to some of our competitors, [is] we have given safety and security the number one priority with everything that we do…We want to make sure that your agent is doing three things: it's providing accurate responses, it is detecting threats real time and mitigating it, and thirdly, it's handling your customers information in a secure manner.”

To learn more about GenerativeAgent, please visit https://www.asapp.com/.