AI Dispatch Won't Fix a Bad Dispatch Strategy

AI-assisted dispatch and route optimization solve different problems. ServiceTitan Dispatch Pro describes skill, location, performance and predicted-job-value recommendations. Housecall Pro Optimize by Drive Time reorders jobs inside one technician's day on its MAX plan; it does not move jobs between days or reassign technicians. A feature called routing is not proof of automated skill or margin assignment.

ServiceTitan reports a 14% average-ticket increase in select customer data measured over six months, with individual results varying. That is a vendor-reported result, not Level evidence, a guaranteed lift or gross profit. A vendor-reported result does not establish a universal outcome. Measure your own completed-job count, travel time and complete job costs on matched periods.

The question nobody asks before they buy: what are you optimizing for? Because AI dispatch will faithfully maximize whatever objective you hand it, and that objective may be implicit in settings. If drive time improves and margin is flat, test whether job mix and assignment changed. Pricing, callbacks, labor cost and demand can also explain the result; the route itself does not establish the cause.

What AI dispatch actually does well

Give it credit, this is genuine productivity:

• Matches work and technicians within the documented scope. ServiceTitan describes skills, performance, location and predicted job value. Do not assume every product sees every job, certification, stocked part or customer tier.
• Supports revised routes. Jobber's routing guide distinguishes legacy and new schedule workflows. Verify whether a change recalculates automatically or needs a dispatcher action.
• Tests a shorter travel order. Compare actual drive minutes and paid hours; a percentage of drive time is not automatically a percentage of the whole workday.
• Keeps dispatcher control. ServiceTitan offers automation or suggestions. Override learning is a product-specific capability to verify, not an assumption to make from an AI label.

For a labor-short contractor, that's exactly the kind of leverage you want, more completed work out of the same crew. (It's one pillar of the operating layer for the trades.)

Where it goes wrong: the objective function

Here's the part the FSM demo skips. An optimizer is only as good as the thing you tell it to maximize. Read the actual objective and constraints. Housecall Pro documents minimizing drive time; ServiceTitan also describes job-value and performance-based recommendations. Productivity proxies can change job mix, so test these possible consequences:

• Optimize for most jobs → the system may favor short calls when the objective and constraints permit it. Compare gross profit dollars and skill needs before assuming fewer long jobs hurt the result.
• Optimize for least drive time → the system clusters geographically and may route your senior tech to whatever's nearby, not whatever needs his skill. Cheap to get there, expensive in opportunity cost.
• Optimize for SLA / response time → great for contracts with penalties, but it can starve the discretionary work where your real margin lives.

Gross profit per technician hour is a useful operating check, alongside contract obligations, safety, skill limits, arrival windows and capacity. A revenue or ticket-value predictor is not a complete-cost margin calculation. FieldEdge describes job-level profitability tracking, so do not assume native financial visibility is universally absent. Test the cost basis, allocation and assignment scope before deciding whether a cross-system finance bridge is needed.

Strategy first: decide what you're maximizing

Before you let an algorithm run your board, make four calls a human has to make:

1. Define the objective. Gross profit per tech hour, not raw job count. If the system can't weight jobs by expected margin, you have to feed it a proxy, job type, ticket size, segment, that correlates with profit.
2. Protect the high-margin work. Verify complete expected costs, skill requirements and customer commitments. A replacement or high-ticket job is not automatically the highest-margin assignment.
3. Match skill to margin, not just to proximity. The closest tech isn't always the right tech. Keep your scarce senior hours on the work that needs them.
4. Decide where the human overrides. Big jobs, key accounts, judgment calls, the dispatcher (or owner) owns those. The AI runs the routine. (This is the same logic as the AI operating system for your money, automate the reflex, keep the judgment human.)

Those decisions give you a testable dispatch policy. Compare results against a retained baseline; they do not guarantee a margin increase.

The honest take

I'm not anti-software. Test the productivity result against your own baseline. I'm anti-buying the engine and skipping the strategy, which is what I see constantly. The vendors sell the routing horsepower because that's what demos well. Some vendors provide goal settings and financial reports. Your responsibility is to establish which costs and constraints they actually use, then reconcile the resulting assignments with the financial outcome.

Dispatch is the control center for field profitability. Treat it that way. The software is necessary; it is not sufficient. Decide what you're optimizing for first, then let the AI run flat out.

FAQ

Does AI dispatch software actually work?

It can support routing and assignment, but capabilities and outcomes vary. ServiceTitan documents skill and predicted-value recommendations, while Housecall Pro's MAX drive-time feature only reorders one technician's day. Retain the selected objective, constraints and baseline, then measure travel time, completed jobs and complete-cost margin. Vendor case studies do not establish a universal lift.

What should AI dispatch optimize for?

Gross profit per technician hour, not raw job count or pure drive-time minimization. An objective that counts jobs without checking cost can favor a short call even when its contribution is lower. Drive-time minimization also needs skill, contract and contribution constraints. Check native financial visibility and cost definitions first. If you use job type, ticket size or segment as a proxy, test its relationship to complete-cost profit rather than assuming revenue means margin.

Do I need to replace my FSM to get AI dispatch?

First inspect the available account feature. The products document different routing, assignment and financial functions, with plan and schedule-version limits. A standalone tool also needs verified access, identifiers and approved write scope. Compare that evidence before replacing the FSM.

Why did our margins not improve after we turned on AI routing?

Investigate before resetting the board. Compare job mix, actual travel, labor and material costs, pricing, callbacks and contract obligations on the same dates. An objective mismatch is one hypothesis; a flat margin does not identify it as the cause.

---

If you've got AI dispatch running but can't say whether it's optimizing for profit or just for a tidy board, that's the strategy gap, and it's costing you margin at full efficiency. Book a 15-minute call and we'll map your dispatch objective to gross profit per tech hour. The first profitability audit is free.