field_service_dispatch v1.0.0

name: field_service_dispatch description: Run field service dispatch day to day: intake triage by urgency and skill, matching jobs to technicians through hard skill and certification gates, route-aware scheduling on a single dispatch board, confirmation and on-my-way customer messages, midday recovery when the day breaks, and an end-of-day board review that catches dropped jobs. Use when dispatching or scheduling service technicians, triaging incoming service calls, assigning jobs to techs, building a dispatch board, handling a same-day emergency call, or reviewing a day or week of field service work for HVAC, plumbing, electrical, appliance, or similar trades.

Field Service Dispatch

Purpose

Give a dispatcher, an owner-operator, or an agent running the office a complete daily system for field service work: every call triaged the same way, every job matched to a tech who can actually close it, routes built around zones instead of call order, customers told what is happening before they have to ask, and a board that gets checked at the end of the day so no job quietly dies on it. Written for shops from one tech to a small crew, with or without dispatch software. Where software exists, this is the method the software should be made to follow.

Workflow

1. Set up the board once. One board is the single source of truth: unassigned jobs by priority, one timeline per tech, each job's zone, window, skill, cert, and part needs on the job itself. A job that lives in a text thread, a sticky note, or someone's memory does not get routed and does not get reviewed. Keep a current skills and certification matrix for every tech in the same place as the board, and a zones map with the shop's own travel minutes between zones.
2. Triage every call at intake. Collect the 10 dispatchable fields in references/triage-and-matching.md before the job reaches the board, run the safety check first, try or rule out one approved remote fix where it is safe to do so, then assign P1 to P4 from the triage table by signals, not by how loud the customer is. Score a single call fast with python3 scripts/dispatch.py triage --safety --json, --no-service, --vulnerable, --business-down, --agreement-same-day, and --sla-hours <n> as they apply.
3. Match through gates, then score. A job goes only to a tech who holds the required skill and a current certification for it and has enough shift left. Among the techs who pass, rank by zone fit, part on truck, site familiarity, and current load, in that order. Full rules in references/triage-and-matching.md. Never solve a coverage gap by sending an unqualified tech. Surface the gap while there is still time to use a contractor, a partner shop, or an honest early call to the customer.
4. Build the day route-aware. Assign in priority order, cluster each tech's stops by zone, put the first stop near the tech's start, and price travel between zones from the shop's own table, not from hope. Hold reserve capacity and at least one flex slot for same day urgent work. Run the plan the evening before for the next day, so confirmations go out and trucks get loaded right: python3 scripts/dispatch.py plan assets/five-tech-day.json. Read the unassigned list and the flags before you read the neat routes. They are the dispatcher's real work list.
5. Confirm and keep customers ahead of the news. Send the confirmation message at booking, the on-my-way message when the tech leaves the prior stop, and a late message as soon as a delay is known. Templates and the consent, opt out, and no texting while driving rules are in references/messages-and-recovery.md. Bad news, a missed window or a cancellation, gets a phone call, with the text as the written record after.
6. Work the day, recover by the playbook. Check the board at midday against reality: who is behind, which windows are at risk, what landed since morning. When the day breaks, work the recovery order in references/messages-and-recovery.md: freeze new promises, re-plan what is left, protect P1 and P2 windows and customers already told a tech is coming, move unconfirmed fill work first, and call every customer whose window will be missed.
7. Close the day with the board review. Every job ends the day completed, cancelled, or rescheduled with a new date on it. Anything else is dropped until a person owns it. Run python3 scripts/dispatch.py review assets/end-of-day.json on the real board export and work its list before the office closes: each dropped or unfinished job gets an owner and a customer message that night.
8. Review the week in 20 minutes. Once a week, read the cause lines and flags together: bounced jobs, true parts flags, no-access visits, overrun causes, and urgent response against the shop's own targets. Change one thing, write it on the board, and check next week whether it worked. The agenda is in references/worked-days.md.

Output Contract

Deliver, for each day worked:

• The board: every job triaged to a priority, assigned through the gates or listed unassigned with the plain reason, each tech's stops in route order with travel and duration totals and slack against capacity.
• The flags list: unassigned urgent jobs, missing intake fields, parts not on the assigned truck, and unconfirmed customers, each with the action taken, not just noted.
• Customer messages sent: confirmation, on-my-way, late, reschedule, and done messages that went out, with phone calls logged for missed windows.
• The end-of-day review: jobs completed, cancelled, and carried forward with new dates, plus the dropped list with an owner and a customer message for each.
• The weekly change: one operational change from the weekly review, written down with the metric it is meant to move.

For a setup or audit engagement, deliver the board layout, the triage table with the shop's own response targets filled in, the skills matrix, and the zone travel table.

Operating Rules

1. Skill and certification gates are hard. Availability never outranks qualification. A repeat visit caused by sending the wrong tech costs more than a later slot with the right one.
2. Priority comes from signals, not volume or value alone. Safety, a vulnerable occupant without an essential service, and a business shut down by the fault sit at the top. Contract targets are read off the job, and the shop's published targets are the ones it is judged by.
3. No job exists off the board. If it is not on the board with its fields filled, it is not scheduled, not routed, and not promised. Intake finishes the fields or the job waits in unassigned, visibly.
4. Travel time is real time. Routes use the shop's measured zone to zone minutes, durations come from the shop's own job history, and a plan with zero slack is reported as fragile, not efficient.
5. Parts are part of matching. A job whose likely part is not on the assigned truck gets flagged before the tech rolls. Loading the part, moving the job, or telling the customer it may take two visits are the options. Hoping is not one.
6. Nobody texts while driving, ever. Work is organised so no tech needs a phone at speed: messages are sent parked, dispatch holds updates until a stop, and a tech who is behind is called only when safely stopped. This is a legal duty for the employer, not a courtesy.
7. Automated messages stay clean. Consent recorded, content informational only, opt outs honored at once, and local legal review behind the shop's messaging rule. A promotion never rides inside an appointment message.
8. Unassigned urgent work escalates the same hour it is found. A P1 or P2 with no qualified tech gets a named owner, a contractor or partner call, or an honest customer call with a new plan. It never just waits on the board.
9. Numbers are the shop's own. This skill sets no industry benchmark for first-time fix, utilization, or response time, and quotes none. Measure the shop's own rates from its boards and improve those. Any outside figure a vendor cites is a sales claim until the shop checks it against its own records.
10. The day is not over until the review is. A board with a dropped job on it at close means the review step was skipped. Run it, own the list, message the customers, then close.