Structured AI fields: intent, objections, next step

The structured fields are a fixed set the model fills from the recorded conversation: lead intent, timeframe, price band, geography, language, objections, next step, and flags for hot lead, Spanish-speaking agent needed, do-not-call signal and wrong person. On a real production call, this looked like intent buyer, geography southwest with an area from Naples to Tampa, no stated price band or timeframe, two objections quoted from the conversation, and a next step of a 5PM callback the same day. The buyer's criteria (water access, three beds, two baths) landed in the summary text, not in a dedicated field.

What the fields are not: a probability of closing, a grade of operator quality, or consent evidence. The hot flag is a signal to review, not a grade. Use the fields to decide what deserves your ears and your follow-up time.

A field can disagree with the operator's note or the summary. Resolve the difference against the recording rather than picking the convenient version, and keep examples of repeated inconsistencies for support to investigate.

The fields earn their keep at scale. A manager picking calls to review can look for hot flags, wrong-person calls or do-not-call signals instead of opening every recording. A covering teammate inherits a lead with the next step already written instead of a stack of unexplained audio.

Write your review questions before you open anything. Did the operator state the purpose. Was the next step actually agreed. Does the summary preserve a refusal rather than converting it into interest. The fields tell you where to look; the audio tells you what happened.

Where a field seems wrong, record the call reference and describe the discrepancy rather than overwriting it silently. Repeated examples help investigate model behavior. Structured fields are included on Pro and Team, with no per-call charge.