Voice AI platform provider Krisp has introduced two new AI-powered capabilities designed to help contact centers strengthen governance across the voice channel, combining real-time fraud detection with automated conversation analysis.
The company is expanding its Call Center AI platform with Voice Security and Speech Analytics, tools aimed at addressing two longstanding operational challenges: protecting live customer interactions from emerging AI-enabled threats and increasing quality assurance coverage beyond traditional manual reviews.
According to Krisp, the new capabilities enable organizations to monitor every customer interaction rather than relying on sampling approaches that typically review only a small percentage of calls.
"The voice channel carries a company's most sensitive moments, and for thirty years it ran on one assumption: if the voice sounds human, it is human. That assumption is now a liability," said Harry Folloder, Chief Commercial Officer at Krisp.
Responding to AI-Driven Fraud
As GenAI makes voice cloning and social engineering attacks more accessible, contact centers are becoming an increasingly important line of defense. The voice channel is often where customer identities are verified, sensitive account information is accessed and high-value financial transactions are authorized.
Randy Layman, Chief Technology Officer at AVOXI, told CX Today in a recent roundtable discussion that customers are seeing AI used to drive more attacks on call centers. “AI is making it so much easier and the volume is just increasing,” Layman said. “We need to really start looking at authentication, not as a black and white yes and no, but as a continuum of probability.”
Ron Zayas, Chief Executive Officer at Ironwall by Incogni, added:
“My biggest concern is the next generation of AI that will be able to combine both the volume that you're already seeing along with the information that's out there and then being able to make it in a realistic conversation again at volume.”
Krisp cites industry research indicating that deepfake fraud attempts have increased significantly over the past three years, growing from 0.1 percent to 6.5 percent of all detected fraud attempts in three years, a 2,137 percent increase, while studies suggest human agents struggle to reliably distinguish synthetic voices from genuine callers, only identifying synthetic voices approximately 60 percent of the time. The company also points to growing financial losses linked to AI-enabled business email compromise and voice impersonation attacks, with GenAI-enabled fraud losses projected to reach $40BN in the U.S. by 2027, up from $12.3BN in 2023.
Voice Security is designed to detect these risks during live conversations through three capabilities:
