Home Services Dispatch: Where AI Cuts Drive Time and Where It Costs You the Job

A shop can answer every call that comes in and still lose the day on the board. Home services dispatch, the decision about which technician takes which job in what order, sets how many jobs get finished, how much of the day goes behind a windshield, and whether the 4:30 p.m. call gets booked today or pushed to Thursday. Most owners spent two years fixing intake and left dispatch to whoever holds the marker.

Two bills accrue while you leave it alone. Trucks burn miles: the IRS set the 2026 business standard mileage rate at 76 cents per mile effective July 1, up from 72.5 cents in the first half of the year, a workable proxy for what fuel, maintenance, and depreciation cost. Technicians are also paid to drive. Under 29 CFR 785.38, travel from job site to job site during the workday "must be counted as hours worked." At the 2025 median of $29.33 an hour for HVAC mechanics, per Bureau of Labor Statistics data, that is payroll spent on pavement.

Routing software is sold on the first bill and quiet about the second. Both are worth attacking, and the arithmetic below shows what each is worth.

Table of Contents

1. What Home Services Dispatch Automation Decides, and What the Dispatcher Keeps
2. The Four Constraints Any Routing Engine Has to Respect
3. What Drive Time Is Worth: Worked Math for a Six-Truck Shop
4. The Arrival Window Is the Customer-Facing Half of Dispatch
5. Five Failure Modes and the Early Signals
6. A 90-Day Rollout for a Small Fleet
7. Frequently Asked Questions
8. Ready to Make Your Dispatch Board Earn Its Keep?

What Home Services Dispatch Automation Decides, and What the Dispatcher Keeps

Home services dispatch automation makes two decisions: assignment, which technician gets a job, and sequence, where it lands in their day. Everything else sold under the name wraps notification and reporting around those two, and they are not equally automatable across the four situations a dispatcher faces.

The morning build. Sequencing tomorrow's booked work, usually the afternoon before. Best candidate for software, because the inputs are known and nobody is standing at the dispatcher's desk.

The intraday insert. A no-cooling call arrives at 11 a.m. and has to go somewhere. Highest value, hardest problem: it depends on which job can absorb a delay without a phone call you would rather not make.

The reshuffle. A job scoped at 90 minutes runs three hours. Software shows the cascade well and picks who gets bumped badly.

The last call of the day. Whether to take the 4:30 job is a question about a person's evening.

Automate the first, let software propose on the second and third, leave the fourth with a person. AI is good at predicting job duration from job type, equipment age, and the individual technician, at ranking assignments by cost in miles and minutes, at flagging by 10 a.m. the day about to fall apart, and at drafting the delay notification so it goes out at all.

It should not decide emergency triage, because no heat in January for an elderly customer is a judgment call rather than a priority score, nor which technician goes to the customer who was unhappy last visit, nor who absorbs the overtime. AI enhances human judgment, never replaces people: the dispatcher shifts from building the board to auditing a proposed one. Process first, technology second, because rules that live in one person's head get reproduced faster, not fixed.

The Four Constraints Any Routing Engine Has to Respect

A router that only minimizes miles hands you schedules your shop cannot run. Four constraints have to exist as data before optimization means anything, and most shops are missing at least two. A 2024 Technical University of Denmark dissertation on the dynamic technician routing and scheduling problem, built on data from Denmark's largest infrastructure provider, models skill substitution, technician home locations, and task urgency as core constraints, not refinements.

Run each as a yes-or-no test on your system today:

1. Skill and certification match. Can you pull a current list of who is cleared for a heat pump changeover, a gas line, or a panel swap, out of your software, without asking a manager? If it lives in memory, the router sends the wrong person.
2. Parts on the truck. Does the system know what is stocked on each truck, or assume all six are identical? Routing that ignores inventory trades drive time for return trips.
3. Time windows as structured data. Is the window a field the software reads, or a sentence in the job notes? Notes are invisible to an optimizer.
4. Start and end location per technician. Take-home trucks move the optimal assignment substantially, and a router that assumes everyone leaves from the shop is wrong every morning.

Fail two or more and your first project is a data project, not an AI project. Unglamorous, and where the return comes from.

What Drive Time Is Worth: Worked Math for a Six-Truck Shop

Routing improvements usually pay back, but the cash and the capacity arrive separately and only one is automatic. Consider a hypothetical six-truck HVAC and plumbing shop, 250 working days a year; every assumption below is illustrative.

The baseline. Each technician drives 85 miles and 2.5 hours on an average day: across 1,500 truck-days, 127,500 miles and 3,750 drive hours a year. At 76 cents a mile, vehicle cost is $96,900. If 70% of those hours are compensable inter-job travel, 2,625 hours at $29.33 plus a 25% burden, roughly $36.66 loaded, is $96,200 in payroll. Call it $193,000 a year spent moving trucks.

The improvement. Verizon Connect's 2026 Fleet Technology Trends Report, a survey of nearly 900 fleet professionals, reports average decreases of 11% to 19% in fuel, accident, labor, and maintenance expenses among technology users. That is a vendor surveying its own market, so take the low end: 12%.

The cash half. 15,300 fewer miles, worth $11,628. This arrives whether or not anyone changes behavior.

The capacity half. 315 fewer paid drive hours, about 13 minutes a day per technician. That adds slack, not a job, and becomes money only if the dispatcher deliberately books into it. Assume half converts, 157 hours, and a completed call consumes 1.8 hours including travel: roughly 88 extra calls. At a $410 average ticket with 30% materials cost, contribution is $287 a call, about $25,100.

Against that. A routing module at $100 per technician per month is $7,200 a year, plus perhaps $9,000 once to clean skill and service-area data, write rules, and train. The questions you ask before signing matter more than the monthly price.

Benefit near $36,700 against $7,200 recurring leaves about $2,460 a month, so setup pays back in under four months. Bank only the fuel and never sell into the recovered time, and the same project takes roughly two years. Two thirds of the return depends on a decision nobody makes inside the software.

The Arrival Window Is the Customer-Facing Half of Dispatch

Every routing gain has a customer-facing price, because a tighter arrival window makes customers happier and routes harder. Decide which you are optimizing before you buy anything.

The customer evidence is consistent. J.D. Power's U.S. Telecom In-Home Service Technician Study, fielded across late 2017 and early 2018 with 3,744 responses, found satisfaction 49 points higher with a one-hour window than a two-hour one and 104 points higher than a four-hour one, 12% of technicians arriving outside their window, and satisfaction 138 points higher when the customer was contacted before arrival. Different trade, same house call. A Housecall Pro survey of 1,040 U.S. homeowners in October 2025 found 59% expect text updates during service, 35% name late arrivals as a frustration, and 53% are comfortable with AI handling an initial inquiry.

The cheap win follows. Narrowing a four-hour window to two roughly halves the slack your router works with, costing the flexibility you just paid for. Notification with a live ETA costs nothing in flexibility and moves satisfaction nearly as much, so do that first. Then narrow windows only on job types you can predict: seasonal maintenance, filter changes, scheduled installs. Keep the wide window for diagnostics, where you honestly do not know how long the job before yours will run.

Five Failure Modes and the Early Signals

Most routing projects fail quietly, in ways the monthly report reads as success. Each shows an early signal inside eight weeks.

1. Miles fall while payroll rises. Signal: fuel down, overtime up, because compressed routes push jobs later and inter-job travel is paid either way. Chart the two together or you will congratulate yourself on the wrong number.
2. Duration estimates copied from the price book. Signal: the board is accurate at 8 a.m. and wrong by 2 p.m. Price book times describe billing, not clocks. Log actual on-site duration for 60 days before trusting any optimizer.
3. Skill data that is aspirational. Signal: callbacks clustering on specific job types, because somebody marked a technician qualified for work they can do slowly. Audit the matrix against who finished those jobs first time.
4. Overrides nobody logs. Signal: no one can say whether the software helps. Log every override with a one-word reason and review monthly. Overrides that beat the model are knowledge to feed back in; ones that lose are a training conversation.
5. Optimizing a service area you should not serve. Signal: the router keeps sending someone 40 minutes out for a $180 call. That is a zone and pricing decision, worth making first.

The fourth decides whether the project survives. A forecast or a route your people quietly ignore is worse than none, because you are paying for it and steering by it.

A 90-Day Rollout for a Small Fleet

Roll this out in the shoulder season, in three stages that each end with a number you did not have before.

Days 1 to 30, measure. Pull miles per truck per day, drive hours, jobs completed per technician per day, and the share of appointments finishing inside their window. Write down the rules the whiteboard already follows, including the unwritten ones about who gets which neighborhood. Run the four-constraint test and fix what fails.

Days 31 to 60, run in parallel. Let the software propose tomorrow's board while your dispatcher builds theirs the usual way. Log every disagreement and, the next day, who was right. Six weeks of that log is the most useful document the project produces, and it names the constraints still missing from the data.

Days 61 to 90, hand over the morning build only. Keep intraday inserts and reshuffles with a person. Publish four numbers weekly: miles per completed job, jobs per technician per day, share of appointments met inside the window, and override rate. Those four say within two months whether it worked, and they make an implementation budget defensible when you expand it.

If intake is the weaker link, fix that first: dispatch optimization multiplies capacity you already have, and does not answer the phone at 5:40 on a Friday.

Frequently Asked Questions

Ready to Make Your Dispatch Board Earn Its Keep?

Dispatch is a data and process problem before it is a software purchase: measure the miles and the paid hours, get skills and windows into structured fields, automate the morning build, keep the judgment calls. ANtreliX is a strategy-first AI and business intelligence consultancy in North Myrtle Beach, South Carolina, working with home services and field operations companies locally and remotely. See our case studies, or Talk to ANtreliX.