Key points
- Dispatch automation is only as good as the job data, technician profiles and rules it is given.
- Check six things before you start: job data, technician profiles, written rules, an approval path, a baseline and data handling.
- Start read-only, with suggestions that a dispatcher approves, and compare the results with your baseline before you go further.
Why readiness matters
Automation does not fix a dispatch process. It repeats it, faster. If job records are incomplete, skills are out of date and the rules for priority and overtime live only in one dispatcher's head, an automated system will make confident-looking mistakes at scale.
Readiness is the work done before the tool is switched on. It is usually less exciting than the software, and it decides whether the project pays off.
The checklist
1. Job data
Every job should have a type, a site location, a priority, a time window and a note on access. The same kind of job should be named the same way every time. If "pump service", "pump svc" and "pump maintenance" are three separate job types, the system cannot learn what a pump job needs.
2. Technician profiles
Skills, certifications, working hours, vehicle stock and availability should be current. A profile that was accurate two years ago and never updated is a quiet source of wrong assignments, and of safety risk when a certification has lapsed.
3. Written rules
Write down the rules dispatchers apply by habit: how priority is set, which customers have service commitments and what they are, how far a technician can travel in a day, when overtime is allowed, and who can override a decision. If a rule is not written down, it cannot be checked. Procedures for the work itself can be captured from video; see SOP365.
4. An approval path
Decide who approves a suggested assignment, how quickly they must respond, and what is recorded when they change it. Human approval is not a formality. It is how the business keeps control of commitments made to customers.
5. A baseline
Before you change anything, record four to eight weeks of current performance: first-time fix rate, average time from call to arrival, overtime hours, and cost per job. Without a baseline, you cannot tell whether the change helped.
6. Customer and data handling
Know which customer and site details will be shared with which tools, where they are stored, and what the contract says about the AI providers used. For Canadian businesses, check this against the privacy law that applies to your customers, including Quebec's rules where they apply.
What to avoid
- Starting with full automatic assignment. Begin with suggestions that a dispatcher accepts or changes.
- Skipping the baseline. A project with no starting numbers will end in an opinion, not a decision.
- Overrides with no reason. If dispatchers change suggestions and nobody records why, the system cannot learn, and nobody can see which rules are wrong.
A 30-day start
Pick one region or one job type. Run suggestions alongside your current process for 30 days, record what was accepted, changed and why, and compare first-time fix rate and overtime against the baseline. If the results hold, extend to the next region. If they do not, the checklist shows you where to fix the process first.
Our Dispatch product is designed for this kind of staged start, and the pilot process sets out how we measure one.