How Our AI Conducts Signal Sessions
The coordinated AI agents behind every Savo conversation
When a participant is in a Signal Session, they're talking to Savo — an AI interviewer that feels like a single, coherent conversational presence. Behind the scenes, several specialized agents work together on every turn.
The agents behind the conversation
The safety check is the first thing that processes every participant utterance. Before anything else happens, it checks the input for safety — off-topic content, sensitive personal disclosures, distress signals. If everything is clear, the conversation proceeds. If not, it intervenes appropriately — redirecting, acknowledging, or pausing — before any other step.
The interviewer generates Savo's response. It draws on the session context and recent conversation history to produce a natural reply that sounds like a skilled interviewer — curious, responsive, and unscripted.
The signal monitor runs in parallel, analyzing the conversation for evidence quality. It evaluates how much useful signal exists on each topic and what's still needed — informing where the conversation goes next.
The coordinator keeps the turn flowing, brings the pieces together, and manages timing. When a session has run its course, it closes gracefully.
What this means for the conversation
From the participant's perspective, the conversation feels natural and responsive. Topics don't feel mechanically scripted — the session adapts based on what's been covered. Savo probes deeper where evidence is thin, moves on naturally when a topic is saturated, and respects the participant's time.
The session doesn't follow a rigid checklist. It follows a research-informed facilitation plan, executed adaptively in real time.