How AI call answering books jobs into ServiceTitan

"Books into ServiceTitan" is on a lot of feature lists. It sounds like a checkbox. It is not. A ServiceTitan booking is six or seven linked records, and getting any one of them wrong creates work for your dispatcher, your tech, or your bookkeeper. This post explains what the booking actually consists of, where AI answering tends to get it wrong, how to set rules that prevent most of it, and what to look at in your first week.

We describe how Contour does this; much of it applies to other products too.

What a ServiceTitan booking actually is

When a CSR books a call in ServiceTitan, they are creating or linking these things:

Customer. The billing entity. A person or a business. Has a name, phone numbers, email, and a billing address. May already exist from a previous job.

Location. The service address where the tech shows up. One customer can have several locations (a landlord with three rentals, a homeowner with a lake house). The location is where the equipment history, gate codes, and "dog in backyard" notes live.

Job. The actual work order. It has a job type (for example "Plumbing - Service Call" or "HVAC - No Heat"), a business unit, a priority, a summary, and one or more appointments.

Business unit (BU). How ServiceTitan segments your company. Most shops split by trade (Plumbing, HVAC, Electrical) and often by service versus install, or residential versus commercial. The BU drives which techs can be assigned, which price book applies, how revenue is reported, and sometimes which dispatcher sees it.

Priority. Usually Normal, High, Urgent, or Emergency. Drives where the job sits on the dispatch board and whether it triggers alerts.

Appointment. The time window, and the arrival window the customer was told.

Notes and summary. The free-text description of what the customer said, what the CSR observed, and anything the tech needs to know before arriving.

Campaign and lead source. Where the call came from (Google LSA, website, referral), used for marketing attribution.

A booking that gets the customer, location, and appointment right but puts the job on the wrong BU or sets priority to Normal on a burst pipe is technically "booked into ServiceTitan." It is also a mess your dispatcher has to clean up at 7 AM.

What Contour writes on each booking

When Contour books a call, it creates or matches the customer, creates or matches the location, and creates the job with these fields set:

• Job type, chosen from the list you approve during setup, based on what the caller described.
• Business unit, according to the mapping you define (for example, anything about water heaters, drains, leaks, or fixtures goes to Plumbing Service; anything about furnaces, AC, or thermostats goes to HVAC Service).
• Priority, from the triage outcome. Water everywhere, no heat below your temperature threshold, sewage backup, and similar get Urgent or Emergency. Routine requests get Normal.
• Appointment, in a slot allowed by your booking rules for that priority and BU.
• Summary and notes, a plain-English call summary: what the caller said in their own words, where in the house, how long it has been happening, whether they found the shutoff, any access notes, and a link to the recording.

Within about a minute of hang-up you or the on-call tech also get a text with the name, address, issue, urgency, booked time, and recording link, so you do not have to open ServiceTitan to know what happened. Urgent calls say URGENT first.

What goes wrong, and why

These are the failure modes we design around. If you are evaluating any AI answering product, ask how it handles each one.

Duplicate customers

The most common problem. A homeowner calls, gives their name as "Mike" instead of "Michael," or calls from their spouse's cell phone, and the system creates a brand-new customer record instead of matching the existing one. Now you have two Michaels at the same address, equipment history split across them, and a membership attached to the wrong one.

Contour matches on phone number first, then on address, before creating a new customer. When the match is ambiguous (same address, different name and phone), it books under the existing location and flags the possible duplicate in the job notes so your dispatcher can merge if needed. It errs toward attaching to an existing record rather than creating a new one.

Wrong business unit

A caller says "the thing in the basement is leaking." Is that a water heater (plumbing) or a condensate line on the furnace (HVAC)? A human CSR would ask a follow-up. An AI that does not ask will guess, and the guess lands the job on the wrong board, in front of the wrong techs, with the wrong price book.

The fix is partly in the conversation (Contour asks clarifying questions when the trade is ambiguous) and partly in your rules (you define a default BU for genuinely ambiguous calls, and whether ambiguous calls should be booked at all or sent to you for a decision).

Wrong priority

Under-triage is dangerous: a sewage backup booked as Normal and scheduled for Thursday. Over-triage is expensive: a dripping faucet booked as Emergency, your on-call tech rolled out of bed, and a $150 ticket.

Contour's priority comes from triage rules you set, including the local weather thresholds. The caller's words matter ("water everywhere" is not the same as "a little drip"), but so does context (no heat at 18°F versus no heat at 50°F). Our emergency triage guide walks through writing these.

Over-booking

Most ServiceTitan shops have capacity rules the dispatcher holds in their head: four service calls per tech per day, keep one slot open for emergencies, no installs on Saturdays. An AI that books into any open-looking slot will fill the emergency buffer with tune-ups by Tuesday.

You tell Contour which slots are bookable, how many jobs per window, and which windows are held for urgent calls only. It books inside those rules. If no slot fits, it takes the information, tells the caller someone will confirm a time, and texts you to decide, rather than inventing a slot.

Out-of-area bookings

A caller forty miles outside your territory books a service call and your tech drives an hour and a half for a $200 job. Contour confirms the service address is inside the service area you define before booking. If it is outside, it tells the caller politely and does not create the job.

Thin notes

A job that says "customer has an issue, please call" is useless to the tech. Contour writes the summary in the caller's own words with the specifics the tech needs. Look at the notes on your first few bookings and tell us what is missing; the format is adjustable.

How to set up your rules

Setup takes an afternoon. You will spend most of it on these decisions.

1. Job types. Export your active job type list from ServiceTitan. Trim it to the ones a phone caller could reasonably trigger. "Water Heater Install" probably should not be bookable from a cold call; "Water Heater - Service" should. Contour will only use the types you approve.

2. Business unit mapping. For each trade you run, write down the keywords and situations that map to each BU. Decide your default for ambiguous calls. If you have residential and commercial BUs, decide how the AI should tell them apart (usually by asking "Is this a home or a business?").

3. Priority rules. List the situations that are Urgent or Emergency. Set the outdoor temperature below which no-heat is urgent and above which no-cool is urgent. List the situations that always transfer to a human (gas smell is the standard one). Everything else is Normal.

4. Capacity and windows. Which arrival windows exist, how many jobs per window per BU, which windows are held for emergencies, and what days are off limits.

5. Service area. Zip codes, or a radius, or a list of towns. Be specific about the edges.

6. Approved language. What the AI may say about pricing (often just "there is a diagnostic fee of X, and the tech will quote the repair on site" or nothing at all), your membership program, and your guarantees. It says nothing you have not approved.

7. Who gets texted. You, the on-call tech, or a rotation. Urgent and Normal can go to different people.

Our ServiceTitan connection guide covers the technical side (API access, which ServiceTitan permissions are needed, and how to test). For the content side, the knowledge base guide covers what to write down about your shop.

What to check in the first week

Treat week one as a shakedown. Go into ServiceTitan every morning and look at the jobs that were booked overnight. Specifically:

Check	What you are looking for	Fix if wrong
Customer records	Any new customer who already existed	Tighten matching rules; merge the duplicate
Business unit	Any job on a BU that does not match the described problem	Adjust the keyword mapping or add a clarifying question
Priority	Any Urgent that should have been Normal, or vice versa	Adjust triage rules or temperature thresholds
Appointment	Any slot that violated your capacity or holdback rules	Adjust window rules
Service area	Any address outside your territory	Tighten the area definition
Notes	Anything the tech would have wanted to know	Tell us what to add to the summary template
Recording vs. booking	Listen to two or three recordings and compare to what was booked	Flag any mismatch

Also ask your dispatcher and your on-call tech, directly, what annoyed them. They will know within three days whether the bookings are usable.

A realistic first-week expectation

Expect to adjust rules. Every shop has a job type naming convention, a BU structure, and scheduling habits a little different from everyone else's, and no setup questionnaire captures all of it. The goal of week one is not zero errors; it is finding the two or three rules that need tightening.

After that, the daily check becomes a glance at the morning texts and an occasional listen to a recording. The recordings are the safety net: anything that looks odd in ServiceTitan can be checked against what the caller actually said.

To see a booking end to end, including the ServiceTitan job record and the text that follows it, book a demo on your own shop's setup. Details on how Contour works and current plans on the pricing page.