Field dispatch optimization automation.
AI classifies every job ticket into structured assignment requirements; VRP solver optimizes routes across the active fleet against multi-objective constraints (drive time + jobs/day + load balance + emergency priority); three-lane routing handles emergency disruption, same-day adjacency-based fills, and multi-day scheduled clustering. Tech mobile app with live GPS + cascade visibility on disruptions. Capacity utilization climbs from 65-72% to 82-88%.
A real dispatch pipeline has four jobs.
Most field service dispatch is a coordinator with a whiteboard, an Excel route sheet that gets re-sorted manually every morning, and a mobile app the techs ignore because it lies about ETAs. Customers wait '8am-5pm' for the tech to arrive somewhere in there. Capacity utilization sits at 65-72% because manual coordination can't see the optimal route. Emergency calls disrupt entire days because the cascade isn't visible. The job of a real dispatch pipeline is to industrialize the route optimization while keeping the human judgment where it belongs (customer relationship, complex scope decisions, exception escalation).
Four jobs. One: classify every incoming job ticket into actionable structure — job type, required skills + certifications, required equipment, estimated duration based on similar historical jobs, urgency tier. Without classification accuracy, route optimization runs on bad inputs. Two: optimize routes via VRP (Vehicle Routing Problem) solver across the active fleet. Multi-objective: minimize drive time, maximize jobs-per-day, balance load, prioritize emergency. Continuous re-optimization as conditions change. Three: route by urgency tier. Emergency disrupts nearest tech's route with cascade visibility. Same-day slots into today's routes via adjacency-based filling. Scheduled plans 2-4 weeks ahead with geographic clustering for efficient routes. Four: dispatch via tech mobile app with full context, live GPS tracking, automatic next-job loading. Customer-facing comms: confirmation, reminders, live tracking link, post-job CSAT. Exception patterns drive structural process improvements.
Done right, your tech utilization climbs 15-20 percentage points, your customer no-show rate drops 60-70%, your first-time-fix rate improves because techs arrive with proper context + parts, and your fleet runs fewer trucks doing more work. Done wrong, you ship route optimization that ignores tech-customer relationships, your emergency cascade visibility breaks down, and you create operational fragility worse than the manual whiteboard.
Whiteboard dispatch + 8am-5pm windows
Dispatch coordinator at 6am whiteboarding today's 47 jobs across 12 trucks. Manual sequencing by ZIP code; 'good enough' routes. 9:15am: emergency call from customer with no heat in winter. Coordinator scans whiteboard for nearest tech; calls tech1 — voicemail. Calls tech2 — already started a job. Calls tech3 — 25 minutes away. Re-shuffles tech3's day; emails 4 customers about delays. 11am: tech1's job runs 90 minutes long; route cascade unclear. Day ends with 38 of 47 jobs completed; capacity utilization 67%. Customer 'window' was 8am-5pm; arrival actually 3:40pm; customer was at appointment 1pm-3pm — missed visit; reschedule. CSAT: 3.2/5.
AI classify + VRP + cascade-aware dispatch
Same 47 jobs. AI classifies each into structured requirements. VRP solver builds optimized routes — 12 trucks, 87% utilization target. 9:15am emergency: solver disrupts nearest qualified tech's day; reshuffles next 3 stops; cascade visibility shows which 4 customers got bumped, by how much. Auto-comms to all affected customers within 90 seconds with new ETAs + apology gesture. 11am tech1's job runs long: solver immediately reroutes downstream stops; tech2 picks up tech1's 4pm job because it's adjacent to tech2's route. Day ends 44 of 47 completed; utilization 84%; on-time arrival rate 89%. CSAT: 4.5/5.
Who this is for, who it isn't.
Field dispatch automation pays back fastest for service businesses with 8+ field techs, 30+ daily jobs, and customer-facing arrival commitments. Below 5 techs, manual dispatch is fine. Below 15 daily jobs per coordinator, the build complexity isn't justified.
Build this if any of these are true.
- You have 8+ field techs and 30+ daily jobs and your dispatch coordinator is bottleneck or burning out. That's the throughput being recovered.
- Your tech capacity utilization is below 75% and you can identify drive-time waste. VRP optimization typically lifts utilization 12-20 percentage points.
- Your customer 'arrival window' is wider than 4 hours and customer complaints reference unpredictability. Tighter windows directly improve CSAT.
- You have GPS-equipped fleet vehicles + tech mobile devices. Without live location data, the optimization runs on stale assumptions.
- You have an ops or dispatch leader willing to own ongoing rule tuning. Without ownership, optimization rules drift from operational reality.
Skip or wait if any of these are true.
- You have under 5 techs and under 15 daily jobs. Manual dispatch with a good coordinator is often more efficient than build-and-tune.
- Your jobs are highly variable in scope and duration is unpredictable. VRP optimization assumes duration estimates; high variability undermines the math.
- Your existing FSM platform (ServiceTitan, Workiz, Housecall Pro) handles your needs adequately. Built-in scheduling has caught up; orchestration on top is for businesses with specific gaps.
- Your techs resist mobile-app discipline. Without status updates flowing through the app, the optimization runs blind. Build operational discipline first.
- You're a single-trade specialty (one type of work, no skill matching needed). The matching component of optimization isn't doing real work; simpler tools may be sufficient.
What this saves, by the numbers.
The savings come from three sources, in order. Capacity utilization gain (the largest line — 12-20 percentage points across the fleet directly translates to revenue or fleet reduction). Reduced no-show + cancellation rate from better customer comms. Coordinator time recovered from manual route building. Most teams see 1.5–2× the conservative numbers below by year two.
The architecture, end to end.
Dispatch architecture has a single trunk (job ticket intake, AI classification, VRP optimization across fleet) feeding 3 urgency lanes. Emergency disrupts nearest qualified tech's route with cascade visibility for downstream customer comms. Same-day slots into today's routes via adjacency-based filling for capacity utilization. Scheduled plans 2-4 weeks ahead with geographic clustering for efficient routes. All three lanes converge at dispatch with tech mobile app + live GPS. Completed jobs feed CSAT + utilization analytics; exceptions surface field disruptions with cascade reroute and pattern analysis for structural improvements. Click any node for the architectural detail; click a path label to highlight one route.
Click any node to expand. Click a path label below to highlight one route through the graph.
Call, portal, IoT alert, recurring schedule. Customer + location + urgency + skills + equipment.
Customer history loaded. Without classification accuracy, route optimization works on bad inputs.
Multi-objective: drive time + jobs/day + load balance + emergency priority. Continuous re-optimize.
Water main, gas leak, no heat in winter. ETA within 5 minutes. Wins customer-for-life when handled well.
Disrupted customers proactively notified with apology + small gesture. Cascade visibility prevents downstream chaos.
Specific window with confidence, not "between 8am-5pm." Adjacency to existing routes.
Capacity utilization climbs from 65-72% (manual) to 82-88% (optimized). Fuel + fatigue savings.
2-4 week multi-day planning. Geographic clustering by ZIP. Premium 1-hour windows for upcharge.
Tech name + photo, ETA window, prep instructions, live tracking. No-shows drop from 12-15% to 4-6%.
Full context: prior notes, parts, safety. Live re-routing. Customer privacy-controlled GPS visibility.
Photos, signature, payment at point of service. Tech's next job loads automatically.
Tech utilization, revenue per truck-day, first-time-fix rate, on-time arrival. Quarterly territory review.
Parts missing, scope expansion, customer not home, weather. Cascade reroute + proactive comms.
Recurring "parts not on truck" = stocking process. Recurring "ran long" = duration model retraining.
Stack combinations that actually work.
Three stack combinations cover most builds. The decision usually comes down to your FSM platform commitment — ServiceTitan dominates HVAC + plumbing + electrical mid-market; Workiz and Housecall Pro cover SMB; custom builds offer the most flexibility for unusual operations.
Tradeoff: The HVAC + plumbing mid-market stack. ServiceTitan as FSM source of truth; OptimoRoute or Routific for VRP optimization; Claude layers AI job classification + cascade communications. About $1,800/mo all-in for 30+ tech operations. Best for established trades with multi-skill technician pools.
Tradeoff: The SMB stack. Workiz or Housecall Pro for FSM at smaller scale; custom OR-Tools VRP solver for optimization (Google's open-source tool); GPT-4o for AI orchestration. Best for $2M-$15M revenue trades. Lower per-tech cost than ServiceTitan; higher build complexity but more control.
Tradeoff: The enterprise stack. Salesforce Field Service Lightning with Einstein for AI-augmented dispatch; integrated with Salesforce CRM for unified customer data. Best for $50M+ revenue with established Salesforce investment + national coverage. Higher per-tech cost; lower build complexity; less flexibility than custom.
Cheapest viable. ServiceTitan native scheduling + OptimoRoute for routing + manual job classification by coordinator. Skip the custom AI layer for v1. About $400/mo above existing ServiceTitan. Validates whether VRP-only optimization (without AI classification) covers your needs. Builds in 1 week.
Production stack for $20M+ revenue with 30+ techs. ServiceTitan ($600+/mo at scale), OptimoRoute ($300+/mo), Claude Sonnet ($150-$400/mo), GPS fleet tracking ($25/truck/mo), Slack with cascade alerts. About $1,500-$2,200/mo all-in. Adds the optimization quality, customer cascade comms, exception pattern analysis, and quarterly process tuning rhythm.
How to actually build this.
Six steps from zero to a production dispatch pipeline. The biggest mistake teams make is shipping VRP optimization before duration estimates are calibrated — optimal routes built on wrong durations are worse than manual routes built on tech intuition.
Calibrate duration estimates
Pull 6-12 months of historical jobs with actual durations. Build duration estimation model per job type (HVAC tune-up: avg 75 min, p90 110 min; water heater install: avg 240 min, p90 320 min). Validate model accuracy against held-out historical data. Without good duration estimates, VRP optimization builds routes that consistently run late. The duration model is the foundation; build it before automation.
Wire AI job classification
AI classifies job tickets into structured assignment requirements: type, skills, equipment, urgency, duration estimate, complexity tier. Validate against 200 historical tickets with known correct classifications; AI must match expert classification 90%+ before scaling. Customer history loaded as classification context — repeat customers with documented preferences route differently than new customers.
Build VRP solver
Vehicle Routing Problem solver across active fleet. Multi-objective: drive time + jobs/day + load balance + emergency priority. Hard constraints (skill match, equipment, time windows, certifications). Soft constraints (preferred tech, customer relationship, tech home location). Solver must find solutions in under 30 seconds for typical day's dispatch or coordinator UX breaks. Performance tuning matters as fleet grows.
Build the three urgency lanes
Emergency: nearest-tech disruption + cascade reroute + customer cascade comms. Same-day: adjacency-based slot finding into today's routes. Scheduled: multi-day planning with geographic clustering + tight customer windows. Build them in volume order — scheduled first (highest volume), same-day second, emergency third with most care because emergency response defines brand.
Wire mobile app + GPS + customer comms
Tech mobile app with full job context, turn-by-turn navigation, status transitions (en-route/arrived/working/completed), photo capture, signature capture, payment processing. GPS tracking visible to coordinator + customer (privacy-controlled). Customer comms cadence: confirmation, 24-hour reminder, 30-min reminder, 'on the way' with live tracking link, post-job CSAT survey.
Add observability + quarterly tuning
Dashboards: tech utilization, revenue per truck-day, on-time arrival rate, first-time-fix rate, no-show rate, customer CSAT. Exception pattern analysis: which job types run long? Which trucks under-stocked? Which territories show capacity gaps? Quarterly ops review uses data to drive: tech training, parts inventory, territory rebalancing, duration model retraining.
Where this fails in real deployments.
Five failure modes that wreck dispatch operations in production. Every team that's built this hits at least three of them.
VRP optimization breaks tech-customer relationships
Customer always sees Mike for HVAC service over 3 years. Mike knows their system, their preferences, their dog. Optimization sends Sarah because she's 8 minutes closer. Customer is annoyed, asks for Mike, has to reschedule, marks dispatcher down on CSAT. Multiplied across 200 long-term customers, the trust erosion costs more than the drive-time savings.
Emergency cascade buries scheduled customers
9am emergency: tech1 reassigned. Tech1 had 4 stops scheduled today; downstream customers get bumped to tomorrow. Tech1's 2pm customer is at home waiting; gets text 'we'll be there tomorrow' at 12:30pm. Customer is angry — they took half-day off work. CSAT crashes; review on Google: 'they cancelled my appointment with 90 min notice for an emergency that wasn't even mine.'
Duration estimates drift over time
Duration model trained on 2024 data. By mid-2025, common service patterns have shifted (new equipment models take longer to service, new techniques speed certain repairs). Model still uses 2024 averages. Routes pack to old durations; cascading delays compound; tech utilization on paper shows 88% but real-world experience is overstuffed days running late.
Tech mobile app status updates lag reality
Tech finishes job; doesn't update status to 'completed' for 30 minutes (driving to lunch, parking, multitasking). Solver thinks tech is still working; doesn't reassign next slot. Coordinator manually overrides. Multiplied across 30 techs over a week, the manual oversight overhead is significant.
Solver finds 'optimal' routes that techs hate
Solver builds routes that minimize drive time mathematically — bouncing techs between distant jobs because aggregate time saved is small. Tech feels like routes are 'all over the place'; cognitive load high; mistakes climb; CSAT on installs declines. Tech feedback: 'the old way had us in one neighborhood per day; this has me crisscrossing the city.'
Build it yourself, or get help.
This is a Tier-3 build because duration calibration + VRP design + cascade visibility are all hard problems. Done well, it pays back in months and dramatically improves fleet operations. Done sloppily, it ships optimization-by-numbers that erodes both tech and customer relationships.
Build it yourself
If you have engineering, ops leadership, fleet GPS, and tech buy-in.
Hire a partner
If fleet capacity is bottlenecking growth and you can't wait 10 weeks.
Want to get in touch with a partner to build this for you? Run the free audit first. It gives any partner the context they need on your business — your stack, your volume, your highest-leverage automation — so the first conversation is about scope, not discovery.
Run the free auditAutomations that pair with this one.
The matchups that come up while building this.
Want to know if this is the highest-leverage automation for your business?
Run a free audit. We'll tell you what would save you the most money — even if it isn't this one.
No credit card. No follow-up call unless you ask.