Case Study: Five9 integrates Deepgram Nova-2 for 2-4x more accurate alphanumeric transcription
Key results
The challenge
Five9's Intelligent Virtual Agent self-service channels struggled to accurately convert numeric and alphanumeric speech, such as order, account, and tracking numbers, into text. Poor recognition in these fields lowered self-service containment and pushed interactions to live agents.
The solution
Five9 integrated Deepgram's Nova-2 automatic speech recognition model as a speech-to-text option in Five9 IVA Studio 7, giving customers real-time transcription for self-service applications directly within Studio.
“Deepgram's real-time transcription accuracy enables improved call resolution and higher self-service containment rates without the need for a live agent, resulting in substantial cost savings for our customers.”
RDRichard DumasVP AI Product Strategy, Five9
The results, in context
In real-world tests, Deepgram was found to be 2-4 times more accurate than alternative speech-to-text options for transcribing alphanumeric inputs. After switching to Deepgram in Five9 IVA Studio, a major healthcare provider doubled its user authentication rates.