AI for dispatch in HVAC: what it does and what stays with your dispatcher

It is 2:40 on a hot afternoon. A no-cool call comes in from a customer with a baby in the house, your nearest tech is forty minutes out finishing an install, and the tech who is free doesn't carry the right certification for that unit. The dispatcher has maybe a minute to decide. That minute is where AI for dispatch in HVAC either helps or gets in the way.

This guide covers what AI can do on a dispatch board, what ServiceTitan's own tools already handle, where a separate agent fits, and what stays with your dispatcher.

What AI for dispatch in HVAC means

AI for dispatch is software that looks at open calls, technician skills, locations and the day's schedule, then suggests (or makes) the assignment a dispatcher would otherwise work out by hand. The dispatcher keeps the final say. The useful versions do three jobs:

• Match. Pair a call with a technician who has the skills and is close enough.
• Watch. Flag a job running late before the next customer notices.
• Confirm. Draft tomorrow's confirmations and notice which ones didn't answer.

Why it matters when technicians are scarce

The Bureau of Labor Statistics projects employment of heating, air conditioning and refrigeration mechanics and installers to grow 11% from 2025 to 2035, much faster than average, with about 40,600 openings a year. It also notes that technicians may be on call for emergencies and work overtime or irregular hours in peak heating and cooling seasons. If you can't hire your way out of a busy week, the hours your current technicians spend driving the wrong direction or waiting on a reassignment are the cheapest ones to win back.

What ServiceTitan already does

If you run ServiceTitan, start with what you already own. ServiceTitan's documentation says Dispatch Pro analyzes job requirements, technician skills, drive time and capacity to assign jobs. Its playbook recommends letting it assign at least half of your jobs, turning on Auto Mode so it re-optimizes on its own, setting up zones, and keeping skills broad and job types narrow.

A ServiceTitan blog post adds that dispatchers can set a maximum drive time between appointments (it uses 30 minutes as an example), and that a job value predictor estimates each job's revenue to help prioritize. Treat the blog's revenue claims as marketing, and measure your own numbers before and after.

The limits matter as much as the features. The Auto Mode documentation says managed technicians are required, that it won't assign multi-technician or multi-day jobs, and that it skips technicians on on-call or time-off shifts. Dispatchers still handle dispatching, holding, completing and confirming jobs, and they can pause it for any technician. That leaves a lot of judgment on the board.

Where a separate agent fits

An agent doesn't replace an optimizer. It covers what the optimizer doesn't touch.

Job	Who owns it today	What an agent adds	Person approves
Suggest a tech for an emergency call	Dispatcher, or Dispatch Pro if on ServiceTitan	A ranked shortlist with the reason (skill, distance, current job)	Dispatcher
Flag a job running late	Dispatcher watching the board	A heads-up when a job passes its expected window, with the next customer named	Dispatcher
Confirm tomorrow's board	CSR, by hand	Drafts confirmation texts, lists who hasn't answered	CSR
Multi-day or two-tech installs	Dispatcher	A checklist of crew, equipment and permits to confirm	Dispatcher
End-of-day summary	Nobody, usually	A list of jobs not closed out and calls not booked	Owner reads

The last row is the one owners ask about most. These are illustrations of the pattern, not a client story.

A day with an agent on the board

The 2:40 call arrives. The agent reads the call notes, sees "no cooling" and a unit type, and checks who is free and certified. It shows the dispatcher two options with drive times and what each tech is doing now. The dispatcher picks one and sends the job. Nothing reached the customer until a person acted.

Later, an install runs long. The agent notices it has passed its window and that the 4 p.m. customer is next on that tech's route. It tells the dispatcher, who calls the customer or moves the job.

At 5:30 it lists tomorrow's unconfirmed jobs. A CSR works that list.

How to start

1. Write down your dispatch rules. Which techs hold which certifications, what counts as an emergency, how far you'll send someone. Most shops keep this in a dispatcher's head.
2. Use the optimizer you already pay for. If you're on ServiceTitan, turn on Dispatch Pro's recommended settings first and watch for a few weeks.
3. Add one agent for one gap. Late-job alerts or confirmations are low risk because a person still acts.
4. Keep a person approving. Read human-in-the-loop AI for how approval works.

According to the Census Bureau's Business Trends and Outlook Survey, 19.8% of businesses used AI as of May 2026, and the smallest firms, with four or fewer employees, were under 20%. You would not be early, and you would not be behind. Pick one job and do it properly.

What AI should not do on dispatch

• Decide for a customer without a safe option. A no-heat call in winter, or an elderly customer with no cooling in a heat wave, gets a person's judgment.
• Promise arrival times it can't keep. A person sets the window.
• Override a technician's safety call. If a tech won't go on a roof in lightning, the board adjusts.
• Pick favorites. The dispatcher still owns fairness across the crew.

Where Cortex fits

Cortex builds AI agents that work inside the software you already run, with a dispatcher or owner approving what matters. We are not a ServiceTitan partner, and when your software's own optimizer does the job, we'll say so. See how dispatch and scheduling fits in a service company and what we do for HVAC companies. If you're deciding where to begin, the AI Operations Audit for HVAC is free and takes about 30 minutes.