Analyzing 10,000 Automated Healthcare Voice Calls

The article discusses insights gained from analyzing 10,000 automated healthcare voice calls processed by the Loquent AI platform. Key findings include the dominance of 4 call patterns, the importance of sub-2-second response latency, and the impact of time-of-day on caller patience.

Details

The article discusses insights gained from analyzing 10,000 automated healthcare voice calls processed by the Loquent AI platform over 6 months. The calls came from 14 clinics in Canada and covered a range of intents like appointment scheduling, confirmations, and insurance questions. Key findings include: 1) 73% of calls followed just 4 core patterns, suggesting the need to focus on perfecting those flows rather than trying to handle every possible conversation; 2) callers tolerate up to 1.8 seconds of response latency before drop-off rates increase significantly, requiring careful optimization of the speech-to-text, language model, and text-to-speech pipeline; 3) morning callers are 2.3x more patient than afternoon callers, leading the team to dynamically adjust the system's behavior based on time of day; and 4) even with an accurate first response, callers often request a human transfer due to robotic-sounding follow-up sentences, prompting the team to focus on more natural conversational patterns.