The qualifier agent
The multi-step AI run that turns an inbound message into a reply, a profile, and a decision.
Every inbound message triggers the qualifier agent — a sequence of AI steps, not a single chatbot call. Each step has one job, which keeps replies fast and decisions auditable.
The steps
Reply
The agent writes the response the lead sees, using your listings, your knowledge wiki, and your voice. For property questions it can search your imported inventory to answer with real listings.
Extract the profile
It reads the latest exchange and updates the lead's structured profile — budget, area, timeline — recording only what the lead actually said.
Update qualification
It re-evaluates the qualification stage and route target in light of the new information.
Hand off (only if needed)
If the lead needs a person, it writes a handoff summary and escalates. Otherwise this step is skipped. See Handoffs & escalation.
Fast vs. queued
How the run is triggered depends on the channel:
- WhatsApp and Gmail messages are queued as debounced jobs (a short quiet period lets a lead finish their thought across several messages), then processed by a background worker.
- Web-chat messages run the agent inline — the visitor is holding an open connection, so replies stream back live.
Testing it safely
The Playground runs the exact same agent — prompts, extraction, your data — without sending anything to a real lead. Use it whenever you change your knowledge base, voice, or behavior settings to see the effect before it goes live.