The estimator records a voice memo and snaps photos on site. An AI workflow turns them into a structured scope of work with line items mapped to the company's price book. The estimator reviews and adjusts it before sending.
Tools used
CompanyCam (photos + notes)
OpenAI API or Claude (vision + structuring)
Make or n8n
Jobber, JobTread or QuickBooks Online (estimate creation)
Expected outcome · Typical goal
Typical goal: estimates go out the same day as the walkthrough, not days later, and scopes come out consistent no matter which estimator wrote them.
Setup notes
Export the company's price book / service catalog as CSV first. The model maps to it and never invents prices.
Output a draft only. A person always approves prices and quantities.
Start with the 3-5 most common job types and build out from there.
Store the 'notes to estimate' pairs so the prompt can be improved with real examples.
Prompt / agent config
Replace the {placeholders} with your own details.
You are an estimating assistant for {Company}. Using the walkthrough transcript and photo descriptions below, produce: (1) a plain-English scope of work (bullets, customer-facing), (2) a line-item list as JSON [{"catalog_code":..., "description":..., "qty":..., "unit":...}] using ONLY codes from the price book provided, (3) a list of questions or measurements the estimator still needs. If something isn't in the price book, put it under 'needs_custom_pricing' instead of guessing. Do not output prices.
Price book: {{csv}}
Transcript: {{transcript}}
Photos: {{photo_notes}}
Last reviewed October 15, 2026. Tool names and features change; check vendor docs before you build.
After an estimate goes out, an automated, personal-sounding follow-up runs over two to three weeks: a check-in, an answer to common objections, a gentle deadline, and a 'should I close this out?' message. Replies stop the sequence and alert the estimator.
Every new lead, whether a website form, Google Local Services Ads message, Angi, Thumbtack or Facebook, gets a personal-sounding reply within a minute, two or three qualifying questions, and a booking link. Qualified leads are pushed to the CRM and flagged for a callback.
Home Services · Contractors & Remodelers · HVAC · Plumbing · Roofing · medium complexity
Techs already take job photos. Each week the best few, tagged with town and job type, become short Google Business Profile updates ('New heat pump install in Huntersville'). AI writes the captions and the owner approves them in a batch.
HVAC · Plumbing · Roofing · Contractors & Remodelers · Home Services · medium complexity
One practical AI use case a month, for Charlotte-area businesses.