AI work order dispatch: suggest the technician, approve it
Most FM help desks in the Kingdom still assign jobs the same way: a coordinator reads the ticket, remembers who is free, and sends a WhatsApp message. It works until the portfolio grows, a coordinator goes on leave, or a client disputes why an urgent job waited two hours. Dispatch software is meant to fix that, but much of what is sold today promises "zero dispatcher involvement". This guide explains what a dispatch suggestion should check, how it should rank, and why, for a Saudi FM company, the supervisor should stay in charge.
What is AI work order dispatch?
Dispatch is the step between a ticket arriving and a person being sent. AI dispatch does the shortlisting for the coordinator: it reads the ticket (trade, building, priority), compares it with what the company knows about its people and contracts, and proposes the best match with a short explanation.
The value is not speed alone. A wrong assignment is expensive: in Aquant's 2025 field-service benchmark, a failed first visit added two more visits on average and stretched resolution by 14 days, and the best performers' first-time fix rate was 86% against 53% for the weakest. Sending the right trade, with the right parts, to the right site the first time is where most of the savings sit.
What should the software check before suggesting anyone?
Good dispatch works in two stages: gates first, ranking second. A gate is a hard rule; if a person fails it, they are not suggested at all, however close they are.
| Check | Gate or ranking | What it means in a Saudi FM company |
|---|---|---|
| Trade | Gate | The ticket needs HVAC, electrical, plumbing, lifts… and the person is qualified for it |
| Documents | Gate | Personal licences valid; for a subcontractor, company documents valid (CR, VAT, GOSI, Nitaqat, Zakat, Chamber) |
| Contract scope | Gate | The client's contract covers this trade and building, and the subcontractor is approved for this client at an agreed rate |
| Availability | Gate | On shift or on call, not on leave |
| Working-hour limits | Gate | The job will not take them past 8 hours a day or 48 a week, or 6 and 36 in Ramadan for Muslim staff |
| Site access | Gate or flag | The building needs an access permit or a specific escort |
| SLA time left | Ranking | How close the client's response and resolution deadlines are |
| Travel at this hour | Ranking | Realistic drive time from the current job, not straight-line distance |
| Workload | Ranking | Jobs already open and hours already booked today |
| Track record | Ranking | First-time fix and repeat visits on this asset type or building |
The working-hour gate is the one most software ignores. The Saudi Labour Law caps actual working time at 8 hours a day or 48 a week, reduced for Muslim workers in Ramadan to 6 a day or 36 a week (Article 98, as published by the Ministry of Human Resources and Social Development). A dispatcher who keeps sending the same reliable technician will break those limits without noticing; a gate will not.
The documents gate matters just as much. A subcontractor whose GOSI certificate or commercial registration has expired is a compliance problem the moment they walk on site. We cover the full document set in our guide to managing maintenance subcontractors.
How should it rank the people who pass?
Among eligible people, ranking is a trade-off, and the trade-offs are local.
- SLA first. A P1 job with 40 minutes left on its response clock outranks everything. The clocks themselves should come from the client's contract, not a generic priority list. See how to set SLA response and resolution times.
- Travel at this hour. "Nearest" changes during the day. TomTom's 2025 traffic index puts the average 10 km trip in Riyadh at 16 min 48 s, but 24 min 12 s in the evening rush; in Jeddah, 15 min against 20 min 24 s. A technician 8 km away across the city at 5 pm may arrive later than one 12 km away on an open road.
- Workload. Spreading jobs evenly protects quality and keeps people within their hours. Saudi government O&M guidance already asks work-control supervisors to check daily whether any person or team is overloaded and whether jobs are grouped by area.
- Track record, used carefully. Past first-time fix on this asset type is useful, but small samples mislead. Treat it as a tie-breaker, not the main factor.
Why should a supervisor approve each suggestion?
Fully automatic dispatch sounds efficient. In practice, four things argue for a human approving each pick.
- The data is never complete. A missing trade tag or an expired document that was never uploaded produces a confident, wrong suggestion. In Salesforce's 2026 State of Field Service survey, 61% of organisations said field workers have limited access to the data they need.
- Some context is not in any system. A resident who asked for a female technician, a VIP floor, a technician who has just finished a difficult job, a family emergency. The supervisor knows; the software does not.
- People stop checking. When a system is usually right, people approve without reading. Regulators call this automation bias; the answer is to show the reason for each suggestion so the approver has something to check.
- Responsibility stays with you. SDAIA's AI Ethics Principles ask for "human oversight, governance, and proper management… across the entire AI System Lifecycle" and for systems whose decisions are "susceptible to intervention from authorized users". These are national guidance rather than law, but they describe how Saudi clients increasingly expect AI to be run.
There is a practical benefit too. The same Salesforce survey found 40% of organisations struggle to tell whether their AI is working. If every approval and every override is logged with its reason, you can answer that question with numbers: how often supervisors accepted the suggestion, and why they changed it when they did.
Where do rules-only and AI-only dispatch go wrong?
| Approach | Typical failure |
|---|---|
| Rules only (round-robin, fixed zones, "nearest available") | Rules drift out of date: leavers stay on the rota, expired documents are never re-checked, Ramadan hours are forgotten. Fixed zones ignore real traffic. Round-robin overloads some people and idles others. |
| AI only (no approval) | Confident wrong picks from incomplete data, reasons nobody can explain to the client, missing context, possible assignments past hour limits or to a non-compliant firm, and no clear owner when it goes wrong. |
| Gates, then ranking, then approval | Slower by a few seconds per job; in exchange the coordinator checks one reasoned suggestion instead of building the shortlist from memory. |
When is your data ready to switch AI dispatch on?
A suggestion can only be as good as the records behind it. Before turning it on, check:
- Every technician and subcontractor has their trades recorded, using one shared list.
- Subcontractor company documents are uploaded with expiry dates.
- Each client contract lists the trades and buildings it covers, and which subcontractors are approved for it.
- Shifts and on-call rotas are in the system, including Ramadan hours.
- Buildings have locations, so travel can be estimated.
- SLA targets are set per client, not one table for everyone.
If two or three of these are missing, fix the data first. Switching on early teaches the team that the suggestions are wrong, and they will ignore them after it is fixed.
How Mubaney handles dispatch
Mubaney's AI dispatch proposes the best qualified person for each new ticket, with a reason, choosing only from the people the company's own records allow: trade, documents, contract and availability. A supervisor approves the suggestion before anyone is sent. It is switched on for each company once the data behind it is in place. Jobs then reach in-house technicians or subcontractors as a secure link, with no app and no account. Read more about what a CAFM system does and about AI in facilities management.
FAQ
What is AI work order dispatch?
It is software that suggests which technician or subcontractor should take a new maintenance job. It checks eligibility (trade, documents, contract, availability, working hours), ranks the eligible people by SLA urgency, travel and workload, and proposes one person with a reason for a supervisor to approve.
Should AI assign maintenance jobs automatically?
For most Saudi FM companies, no. Records are rarely complete, some context never reaches the system, and responsibility stays with the company. A supervisor approving each reasoned suggestion keeps the speed and removes most of the risk.
How does dispatch software respect Saudi working-hour limits?
It should treat hours as a hard rule: no one is suggested if the job would take them past 8 hours a day or 48 a week, or 6 and 36 in Ramadan for Muslim staff, as set by Article 98 of the Labour Law. That needs shifts and booked hours recorded in the system.
Can AI dispatch include subcontractors?
Yes, if their trades, company documents and approved clients are recorded. A subcontractor with an expired CR or GOSI certificate, or one not approved for that client's contract, should be excluded before ranking.
What data do I need before using AI dispatch?
Trades for every person, subcontractor documents with expiry dates, contracts that list trades and buildings, shifts and on-call rotas, building locations and per-client SLA targets. Without them, the suggestions will be confidently wrong.
Sources
- SDAIA: AI Ethics Principles
- HRSD Knowledge Centre: actual working hours (Labour Law, Articles 98–107)
- TomTom Traffic Index 2025: Riyadh
- TomTom Traffic Index 2025: Jeddah
- Aquant 2025 Field Service Benchmark Report (Industrial Machinery Digest coverage)
- Salesforce State of Field Service (Agile Brand Guide coverage, July 2026)
- NIST AI Risk Management Framework 1.0
- EXPRO: work-control scheduling checklist (Arabic)