Jev makes ~38K turn-level forecasts across 2,029 calls using only structured inputs — turns, tool calls, workflow stages, and timing — at a 118 ms median, without touching caller names, addresses, or health data.
Calls have a shape. Jev never heard a word. It saw only turns, tool calls, workflow stages and timing, and still made 38,012 turn-level forecasts across 2,029 calls. That's about 19 per call, at a 118 ms median. Structure-only input means no caller names, addresses or health
The author shares a Jev usage insight: it seemed tricky at first, but clicked when framed as 'not letting the AI write text.' Using inquiry handling as an example, the conventional approach dumps everything on GPT or Claude to analyze, while Jev takes a different approach.
The prototype uses AI to estimate hesitation or interest from cursor back-and-forth movements or lingering before clicking, observing customer demeanor and supplementing explanations or suggesting comparisons.
The author built an e-commerce real-time customer service demo using Jev and gpt-live-1, which instantly suggests recommended products based on the user's speech during a conversation, delivering a great experience.
