Most "AI dispatch" asks you to trust a score you can't see. The TradesWork AI dispatcher works the other way around. A rules-based planner builds the board from things you can check (skills, certifications, shifts, promised windows, parts on the truck, job value), shows you every change with a one-line reason, and waits for you to apply it. The AI part is narrow on purpose: it explains the plan in plain English and turns a dispatcher's note into rules.
The four goals
Owners set the goal once in Planner settings. Each preset changes how the planner weighs driving, job value, changes to the current board, moved windows and even workloads.
| Goal | What it optimizes | Use it when |
|---|---|---|
| Balanced (default) | Less driving, even workloads, the valuable jobs covered | Most days |
| Least driving | Shortest routes first; workloads may be uneven | Fuel-heavy territories, spread-out service areas |
| Most revenue | Highest-value and member jobs first, even if routes get longer | Peak season, when demand outruns capacity |
| Keep every promise | Never moves a customer's arrival window; fewest reassignments | Commercial accounts, shops rebuilding trust after late arrivals |
Two things hold under every goal. Emergency and high-priority jobs carry a heavy penalty for being left off the board, and members carry an extra one. And in re-plan mode, the cost of changing someone's day triples, so a re-plan fixes the problem with the fewest moves rather than reshuffling everyone.
What the planner respects
The planner starts from your current board, not a blank page. It first frees work that can't stay where it is, then places waiting and freed work, then improves.
Never broken:
- A technician whose certification for a required skill has expired can't take that job.
- Time off, shifts and non-job events (training, meetings, vehicle service) block time.
- An apprentice marked "needs a mentor" is only planned onto a crew with someone marked "can mentor".
- Work already under way comes first on that technician's route.
- Promised arrival windows hold, unless you tick "allow window changes" for that plan.
Weighed, not absolute:
- Zone and business unit fit.
- Continuity on multi-day jobs: later visits prefer the crew already on the job.
- Parts: if the job lists parts and the technician's truck doesn't carry them, the stop is marked as a parts run and counted. A truck with no stock on file counts as "unknown", not "missing".
- Job value, priority and membership.
- Daily capacity caps per technician.
Crews (jobs that need two or more people at once) are planned as one unit: every member gets the same block, and crews move only as a whole crew.
Re-planning the rest of today
A job runs long, a van breaks down, a no-heat call comes in at 9:40. Re-plan is built for that moment.
- 1
Pick what happened
A job is running long (pick the appointment and how many more minutes, 5 to 600), a technician called out, an emergency came in, or just re-check the rest of today.
- 2
It starts from now
Nothing is scheduled before the current time. Trucks start from their live position, their current job, or home base. Windows already missed become a cost, not a block.
- 3
Fewest moves
Changing someone's day costs three times as much as in a day plan, so the fix touches as few customers as possible.
- 4
Apply or let it apply
Review the changes, or, with auto mode on, re-plans that keep every window and unassign nobody apply by themselves.
Each changed job gets a reason line, such as the job that's running long and by how much, so the technicians and CSRs hear a consistent story.
Notes in plain English
Dispatchers know things the system doesn't. The note box takes up to 1,000 characters of plain English, and AI turns it into five kinds of rules:
| You write | Rule it becomes |
|---|---|
| "Don't touch Maria's day" | Lock that technician's day |
| "Sam is at training until noon, no new work" | Give that technician no new work |
| "Keep Luis near the shop today" | Keep their jobs within N minutes of home base (20 if you don't say) |
| "Give #1043 to Dana if you can" | Prefer that technician for that job |
| "Leave #1051 where it is" | Lock that job |
After planning, you see what was applied ("Keep Luis within 20 min of base") and anything it couldn't express. It never invents technicians or jobs; names must match people on the board. If AI isn't available for your role or company, the note is ignored and the screen tells you to use the options instead.
Reading a plan before you apply it
Every plan comes with before and after numbers:
- Drive minutes across every route (blank if no stop has coordinates).
- Utilization: booked minutes over working minutes.
- Revenue at risk: the value of promised jobs left unassigned or past their window.
- Windows kept out of windows promised.
- Unassigned jobs and parts runs.
Below that, each change is labelled Schedule, Assign, Reassign, New window or Unassign, with before and after. Jobs the plan couldn't place are listed with the reason. Constraints it couldn't honor show as warnings, for example a "stay near base" rule for someone with no home base on file.
Drive times inside the planner start as straight-line estimates. When a routing provider such as openrouteservice1 is connected, the totals are refined with real road times for plans up to 400 legs, and the plan tells you which method it used.
How it compares with Dispatch Pro
ServiceTitan's equivalent is Dispatch Pro. Its help center says auto mode reshuffles the board every 10 minutes during business hours, never sends a technician over the configured maximum drive time, and only works with managed technicians.2 Its marketing page says it can prioritize drive time, average ticket size, or both, and cites drive-time reductions of 6% and 10% for select customers.3
The design choice in TradesWork is different. Continuous reshuffling suits a big call center with a dedicated dispatch team. In a 5 to 30 tech shop, a board that keeps changing on its own can be hard for technicians and customers to follow. TradesWork proposes, you approve, and the only automatic path is the narrowest one: a re-plan that keeps every promise and drops nobody.
Getting good plans
The planner is only as good as what it knows. Before your first real plan:
- Give every technician skills, certifications with expiry dates, shifts and a home base.
- Set default durations and required skills on job types.
- Mark apprentices "needs a mentor" and seniors "can mentor".
- Put stock on trucks if you want the parts check to mean something.
- Run Plan the day for tomorrow every afternoon for a week without applying. Compare its after numbers to how the day actually went.
If the apply rate stays low, the problem is usually missing setup, not the goal.
Footnotes
-
openrouteservice, routing and directions API. https://openrouteservice.org/ ↩
-
ServiceTitan help center, "Dispatch Pro overview." https://help.servicetitan.com/docs/dispatch-pro-overview ↩
-
ServiceTitan, Dispatch Pro product page. https://www.servicetitan.com/features/pro/dispatch ↩
