AI scheduling and dispatch optimization: what's real

What AI dispatch means in practice

AI dispatch is a broad label, and vendors use it for very different things. Some features are mature, well-understood optimization. Others are forecasts with wide error bars. A few are mostly marketing. Before you evaluate any product, separate the features into what they actually do.

Route optimization and rule-based assignment are the parts most likely to pay off quickly, because the math is well understood and the data is mostly addresses and times you already have.

Why drive time matters so much

Every mile a service truck drives costs money in fuel, wear and technician time. The IRS standard mileage rate for business use was 70 cents a mile in 2025, 72.5 cents for January to June 2026 and 76 cents for July to December 2026^[1]. Those rates are a reasonable proxy for vehicle cost per mile, before you add the technician's wage for the time spent driving.

How trucks are driven matters too. The Department of Energy and EPA note that idling can use a quarter to a half gallon of fuel an hour, and that aggressive driving can lower gas mileage by roughly 15% to 30% at highway speeds and 10% to 40% in stop-and-go traffic^[2]. Better routes reduce both miles and the rushing that comes from impossible schedules.

Example: a fictional five-technician company saves an average of 15 miles per technician per day after cleaning up its routes. Over 21 working days at 72.5 cents a mile, that is about $1,142 a month in vehicle cost, before counting the technician hours freed for another call or an earlier finish.

The data an optimizer needs

1. Geocoded service addresses: every customer location pinned correctly, including rural routes and new subdivisions.
2. Real job durations by type, from actual start and finish times, not the default your software was set up with years ago.
3. Technician skills and certifications, such as refrigerant handling, licensed electrical work, pesticide application or torsion springs.
4. Start and end points for each technician: the shop, or home if trucks go home.
5. Arrival windows promised to customers, and which ones are fixed.
6. Priority rules: emergencies, maintenance plan members, warranty callbacks, commercial service agreements.
7. Break and shift rules, overtime limits and on-call rotation.
8. Parts and truck stock, if certain jobs need parts only some trucks carry.
9. Weather constraints for outdoor trades; roofers, for example, usually do not work during inclement weather^[3].

Where optimization falls short

• It only knows the constraints you give it. The customer who must have the same technician, the gate that only opens after 10 a.m., the dog that needs a phone call first: none of that exists unless someone recorded it.
• It optimizes what you tell it to. Minimizing drive time can send your best diagnostician to five easy tune-ups while a hard repair waits.
• Emergencies break plans. A no-heat call in January or a storm day reorders everything, and dispatchers make better trade-offs in those moments than any rule set.
• Technician trust is fragile. If the tool keeps sending people across town for no visible reason, they will ignore it.
• Fairness matters. Revenue-based assignment can starve newer technicians of good jobs and slow their development.

Dispatching is skilled work. US building equipment contractors employed 14,490 dispatchers in May 2025^[4], and dispatchers outside police, fire and ambulance earned a median of $50,340 a year^[5]. Good tools make those people faster and less stressed; they do not replace the judgment calls that keep customers happy.

How the dispatcher's day changes

When a tool takes over the routine ordering of stops, the dispatcher's job shifts rather than disappears. Less time goes on dragging jobs around a board; more goes on the decisions that need context and judgment.

• Handling exceptions: the customer who calls to say they will be late, the job that overruns, the part that is not on the truck.
• Protecting promises: making sure a re-optimized route does not break an arrival window someone was told on the phone.
• Inserting emergencies: deciding which routine job moves when a no-heat call comes in, and calling that customer.
• Coaching technicians: following up on jobs that consistently run long, which also improves the duration data the tool uses.
• Keeping the rules current: adding new constraints whenever an override reveals one.

Explain this shift before the pilot starts. A dispatcher who thinks the tool is there to replace them will find reasons it fails; one who sees it as relief from the tedious part of the board will help it succeed.

Piloting a dispatch tool

1. 01

   Measure the baseline

   For four weeks, record miles or drive hours per technician per day, jobs per technician per day, late arrivals and overtime.

2. 02

   Clean the data

   Fix addresses, update job durations from actuals and record technician skills and start points.

3. 03

   Run it in suggestion mode

   Let the tool propose routes and assignments while dispatchers decide. Note every time they override it and why.

4. 04

   Turn overrides into rules

   Most overrides reveal a missing constraint. Add it, then see whether suggestions improve.

5. 05

   Compare after a month

   Look at drive time, jobs per day, late arrivals and technician feedback against the baseline, in a similar season.

6. 06

   Expand carefully

   Automate the routine parts of the board first, such as ordering stops, and keep assignment of complex jobs with dispatchers.

Setup mistakes that make tools look bad

• Leaving default job durations in place, so every route looks perfect on screen and runs late in practice.
• Treating every arrival window as fixed, which leaves the optimizer no room to improve anything.
• Forgetting technician start points, so routes assume everyone begins at the shop when half the team drives from home.
• Switching to automatic mode on the first day, before dispatchers have seen and corrected suggestions for a few weeks.
• Measuring success by miles alone, while late arrivals and overtime quietly rise.

Each of these is easy to fix once spotted. The pilot structure above is designed to surface them in the first month, while the tool is still in suggestion mode and no customer promise depends on it.

Questions to ask a vendor

1. Which features are optimization on rules I set, and which are predictions? How accurate are the predictions on companies like mine?
2. What data do you need from my field service software, and how is it synced?
3. Can dispatchers override any suggestion, and does the tool learn from overrides or just log them?
4. How does it handle emergencies inserted mid-day?
5. What happens when a technician's status is not updated?
6. Can I see the reason for each assignment in plain language?

Example: what a pilot reveals

Example: a fictional pest control company with seven technicians pilots route optimization in suggestion mode. In week one, dispatchers override about a third of suggestions. The reasons cluster: several commercial accounts require visits before opening hours, two technicians are not licensed for a specific treatment, and one rural area has addresses that geocode to the wrong road. After the company adds those constraints and fixes the addresses, overrides drop to a handful a day. Drive time falls, and technicians say the days feel less rushed. The company keeps assignment of termite inspections with the senior dispatcher, because those jobs depend on who can sell the treatment plan.

Booking rules and dispatch rules are two sides of the same coin; see connecting AI phone answering to your scheduling software for the booking side.