The Engine That Runs Your Maintenance Operation
Every maintenance request that arrives at your operation — a tenant complaint, a sensor alert, a PM due date, a compliance deadline — requires someone to read it, classify it, find the right vendor, dispatch them, follow up, and close the loop. In an operation with 10 or 50 properties, that coordination layer runs constantly, in parallel, across every location. It consumes more FM time than most operations realize — and it's the first thing that breaks when volume spikes or a key person is unavailable.
The coordination layer that runs itself.
The AI Dispatch Engine is the coordination layer that runs itself. It handles every step of the work order lifecycle — from detection to close — automatically, within the rules your operation sets. When a decision requires judgment, it stops and surfaces it. Everything else moves without anyone having to push it.
What the Engine Does — And What It Doesn't
Understanding the engine starts with understanding what "semi-autonomous" means in practice. It doesn't mean the system operates without oversight. It means the coordination work — the steps that currently consume FM time without requiring FM judgment — executes automatically. The decisions that require judgment stay with your team.
The engine handles:
- Work orders created from every source — scheduled PMs, sensor alerts, tenant reports, compliance requirements, vendor updates, and recurring issue patterns. Each one enters the same workflow, gets classified by priority and trade, and moves toward dispatch without waiting for someone to process it manually.
- Vendor dispatch within configured parameters — matching the work order to the right vendor by trade, certification status, location coverage, and performance history. When the vendor, cost, and scope fall within pre-approved thresholds, dispatch happens automatically. When they don't, the engine flags it for your review.
- SLA monitoring from the moment the work order is created. The engine tracks every open work order against its response window and escalates automatically when a breach is predicted — before it happens, not after.
- Documentation requests at work order close. The engine doesn't close a work order until the vendor has submitted the required service report, technician sign-off, and any compliance documentation required for that asset type. The record exists because the workflow required it.
Your team handles:
- Work orders where the cost estimate exceeds the approved NTE.
- Vendor substitutions on active accounts.
- Compliance findings that require regulatory response.
- Scope changes identified during the work.
- Any pattern of recurring issues that the engine surfaces as requiring strategic review.
These are the decisions that require judgment — and they're the ones your team should be making, not chasing status updates.
Every Source. One Workflow.
The engine receives maintenance inputs from multiple sources simultaneously and converts all of them into the same structured work order format — with the same SLA, the same documentation requirement, and the same tracking logic regardless of where the request originated.
Sensor alerts from IoT devices monitoring asset condition
temperature anomalies, vibration patterns, runtime thresholds, water detection, energy spikes. When a sensor crosses a configured threshold, the engine evaluates whether to create a work order automatically or flag for FM review based on the severity and the asset's compliance criticality.
Scheduled preventive maintenance
the engine generates PM work orders based on the frequency configured for each asset at each location. PM work orders don't require anyone to check a calendar. They appear when they're due, assigned to the right vendor, with the right scope and documentation requirements pre-loaded.
Tenant and occupant reports submitted through any channel
web portal, phone, email integration. The engine classifies the request, assigns priority, and moves it into the dispatch queue without requiring an FM to manually create the work order.
Compliance requirements with defined inspection frequencies
NFPA 25 sprinkler inspections, NFPA 10 extinguisher checks, NFPA 110 generator load tests, elevator certifications, and the full range of regulatory obligations for your property types and markets. The engine creates the work order before the deadline, not after the inspection finds it missing.
Invoice anomalies detected during payment processing
when an invoice doesn't match the contracted rate or approved scope, the engine flags it as a work order requiring resolution rather than letting it pass through unchecked.
Recurring issue patterns identified by AI
when the engine detects multiple corrective work orders on the same asset within a defined period, it surfaces the pattern as a diagnostic recommendation before the next failure event occurs.
The Dispatch Logic — How the Right Vendor Gets to the Right Job
Dispatch is not a simple routing decision. It involves trade matching, certification verification, location coverage, cost threshold evaluation, and performance history — all of which have to be current and accurate at the moment of dispatch, not at the moment of vendor onboarding.
Trade and certification match.
The vendor dispatched for HVAC refrigerant work holds a current EPA 608 certification. The vendor dispatched for fire suppression ITM holds the required NICET certification. The elevator contractor holds a current state mechanic license. These are verified at the time of dispatch — not assumed from onboarding records.
Location coverage.
The vendor's confirmed service area covers the property location. Multi-site portfolios don't send vendors outside their verified coverage zones.
Cost threshold.
The estimated cost range for the work order falls within the pre-configured auto-approval threshold. Work orders above the threshold are flagged for client approval before dispatch — with the vendor recommendation, cost estimate, and scope pre-populated so the decision takes seconds, not an email thread.
Performance score.
The engine routes to vendors with the strongest performance history for that trade and location — response time compliance, first-time completion rate, invoice accuracy, and documentation quality, updated automatically from every completed work order.
Availability.
For time-sensitive work orders, the engine selects from vendors who have confirmed capacity for the required timeframe.
When all five criteria are met within auto-dispatch parameters, the vendor receives the work order notification immediately. When any criterion requires client input, the engine surfaces the decision with full context — not an alert that something needs attention, but a specific decision with a recommended action.
What Your Operation Looks Like When This Runs
The operational difference is not visible in any single work order. It's visible in the aggregate — in what the FM team is doing with its time and what the operation is catching before it becomes a problem.
the FM's morning starts with reviewing what came in overnight, manually creating work orders for the requests that arrived, chasing vendors for status on open items, following up on a PM that was due last week, and discovering that the compliance inspection scheduled for Thursday hasn't been confirmed with the vendor yet.
the FM's morning starts with the engine's exception queue — the three items that require human judgment today. Two are cost approvals above threshold. One is a recurring issue pattern on an HVAC unit that the engine has flagged for review. Everything else is already moving.
The shift is not from work to no work. It's from coordination work — pushing items through a manual process — to management work — making the decisions that require judgment and reviewing the outcomes the process produces automatically.
According to IFMA FM Pulse Q4 2025, only 10% of FM organizations report all scheduled maintenance completing on time. The bottleneck is not capacity for the work itself — it's the coordination model that requires human action at every step to keep work moving. When that model changes, the completion rate changes with it.
How the Engine Learns
The engine doesn't operate on fixed rules alone. It improves over time based on the outcomes of every work order it processes.
When a client overrides a vendor recommendation, the engine captures the reason — wrong trade, cost concern, site preference, certification issue — and adjusts future routing for that location and work order type. When a vendor consistently misses SLA windows for a specific trade, their routing priority decreases automatically. When a sensor threshold generates a false positive that results in a work order with no finding, the threshold is calibrated for that specific asset.
The engine review cycle — 30 days after activation, 60-day stabilization, quarterly operational review — is built into the onboarding process. The system doesn't run on the same configuration in month 12 that it ran on in month one. It runs on a configuration shaped by the actual performance of vendors, assets, and workflows in your specific operation.
Part of a Complete Maintenance Operation
The AI Dispatch Engine is the coordination layer of Sweven FM's Maintenance Infrastructure as a Service model. It doesn't operate independently — it runs on inputs from IoT asset monitoring, dispatches through the verified vendor network, closes the payment loop through payment automation, and produces the compliance documentation that the reporting layer surfaces.
Each component of the service feeds the engine. The engine coordinates everything else.
→ IoT Asset Monitoring — the signals the engine acts on → Vendor Marketplace — the network the engine dispatches through → Payment Automation — the financial layer the engine closes → Compliance & Reporting — the record the engine builds
See How the Engine Runs in Your Operation
Most commercial maintenance operations have the same coordination problem at different scales. The engine is configured to your properties, your vendors, your approval thresholds, and your compliance requirements — not to a generic template.