What does the professional need before taking the job?
Keeping every detail against handing someone something they will actually read.
Context
The intake conversation holds everything: what the customer said, what was asked, what was confirmed, the photo, the hesitation. The tempting move is to forward all of it, on the grounds that more context cannot hurt. But a pro deciding whether to quote is doing that between jobs, on a phone, with about fifteen seconds of attention. A transcript is not a work order.
What was on the table
Send the full transcript.
Nothing is lost and nothing is read. It also quietly exposes offhand things the customer said to an assistant, not to a contractor.
Send only the structured fields.
Scannable, and it drops the one or two details that did not fit a field but change how someone shows up.
- Chosen
A structured summary, the attachments, and an explicit list of what is missing.
The decision
The pro gets the structured request, the photos as photos, and a short operational summary. Anything the system does not know is named rather than left blank, so the pro can see the shape of the gap before deciding to quote. The photo travels as the file, not as the assistant's written description of it — the pro reads the fitting themselves.
The principleWhat comes out of a conversation should be a job, not a recording of the conversation.
What it cost
Summarisation can drop something that mattered, and the summary is one more thing that has to be right. It also means the pro cannot audit how a value was reached without asking.
Evidence
- Photos have two destinations — one for the model to read during the conversation, one for the pro to look at — and only the second travels on the request.
- The summary names missing information explicitly instead of omitting the field.
- Structured fields are typed per service, so a plumbing request and an appliance request do not share a shape.
Enforced in the product code, not only described here.
What would change my mind
If pros keep asking questions the transcript would have answered, the summary is too lossy and the answer is a collapsed "everything else they said" section under it. What would not change is that the transcript is never the primary artefact.