Quote-in-five-minutes: how dispatchers actually do it
A live look at the AI quoting flow we built for HVAC and plumbing crews.
A homeowner calls an HVAC company at 7:43 pm with a broken unit. The dispatcher who quotes them by 7:48 pm wins the job. The one who calls back tomorrow morning loses to the competitor who didn't. Here's the AI flow that lets a single dispatcher quote in five minutes — and how it actually works on the call.
The five-minute flow
- 0:00–0:30 — AI greets, captures address, problem, system type.
- 0:30–2:00 — AI pulls service history (if existing customer), warranty status, and equipment specs from the CRM.
- 2:00–3:30 — AI asks 3–4 diagnostic questions ("Is the outdoor unit running? Any unusual noises?") and matches against a decision tree of common failure modes.
- 3:30–4:30 — AI generates a price range based on the likely diagnosis: "$220–$340 for a capacitor, $1,800–$2,400 for a compressor."
- 4:30–5:00 — Dispatcher takes the handoff with the full context, confirms the price range, and books the truck.
The dispatcher's job is no longer gathering information. It's closing the booking.
What makes this work
- Decision tree, not LLM guesswork. The pricing comes from your actual price book, matched to symptoms. AI does the matching; it doesn't invent prices.
- The handoff is loud. When AI passes to the dispatcher, the screen pops with: customer, address, problem, history, suggested quote. No "let me catch up."
- One screen, three actions. Confirm price, book the slot, send the SMS. The dispatcher never types the address.
What dispatchers told us
After 90 days of running this with three HVAC companies:
- Average call time dropped from 11 minutes to 4. Most of the saved time was the dispatcher typing things AI already had.
- Booking rate went up 22% — faster quote = more wins.
- Dispatcher headcount stayed the same — but they handled 40% more call volume.
The pattern repeats in plumbing, electrical, and garage door service. Anywhere price ranges can be quoted from symptoms, this flow works.
What to build first
Don't try to build the AI. Start with the decision tree. List your top 20 service calls and the price range for each. Pull from last 90 days of invoices.
If you can't write the tree, AI can't quote from it. The price book is the product. Once that's clean, plugging it into an AI front end is a two-week build.