Built for equipment, not for tickets
Most service AI is adapted from support software, where work ends when a reply is sent. Field work ends when a machine runs, and the part, the skill and the window in between are the entire problem.
Agent to Agent Workflows
AI agents for field service
Field service is not customer support with a van. The constraint is physical — a technician, a part, a location and a window, and every one of them has to line up before anything actually gets fixed. This is the shape of the platform for teams whose work ends in front of a machine.
- Built for keeping physical equipment running
- Diagnosis before a truck rolls, not after arrival
- Dispatch that accounts for skills, parts and location
- Knowledge delivered to someone standing at the asset
The technician becomes a parts runner
A generic ticket says the unit is not cooling. The technician arrives, diagnoses a leaking valve, checks the van, and the part is not there. Half a day gone and nothing fixed.
The technician spent it driving rather than repairing, and nobody learned anything that would prevent it next week. None of that was a mistake — it is what happens when work moves as a chain of handoffs, none of them able to see far enough ahead to prepare.
What has to happen before the van leaves
Every one of these is a decision currently made on site, too late to change the outcome.
- Step 1
Predict rather than react
Telemetry from connected equipment is read for the correlated signals that precede a failure, so work is scheduled against a predicted fault rather than a phone call.
- Step 2
Diagnose before dispatch
The likely cause is established from the symptom, the service history and the documentation before anyone is sent.
- Step 3
Confirm the part
Availability is checked against the van and the nearest depot as part of the scheduling decision, so a job is not planned into a shortage.
- Step 4
Match the technician
Assignment weighs skills, the parts position and location together, instead of routing on availability and hoping.
Why field service needs its own agents
The differences from support software are structural, not cosmetic.
The agents work as a system
Diagnosis does not help if the part is missing, and parts do not help if the wrong technician is sent. The value sits in the handoffs between them.
What the expert knows gets captured
The most expensive fact about a field organisation is that its judgement lives in a handful of people and leaves when they do.
Why the usual approaches fall short
Each of these is good at one part of a problem that has four parts.
FSM scheduling software on its own
Excellent at moving people around and blind to what is wrong with the machine.
Customer support AI applied to field work
Optimises for closing an interaction. Field work closes when equipment runs.
A knowledge base the technician can search
Puts documents in a back pocket and leaves the interpretation to somebody standing in front of a fault.
Preventive maintenance on a calendar
Services healthy assets on schedule and misses the one that failed early.
What things are called
Field service vocabulary, and the two terms Ascendo uses differently.
- First-time fix
- The job resolved on the first visit with no return trip — the metric most field organisations are really managing.
- Truck roll
- Sending a technician to site. The most expensive step in the process and the one worth preparing or avoiding.
- Install base
- The population of equipment you are responsible for keeping running.
- Dataflow
- Agents acting on the same job in parallel, as opposed to a workflow where each step waits for a human handoff.
- Dispatch
- Matching a job to a technician on skills, parts position and location together.
- Intelligent asset fingerprint
- The learned picture of normal and abnormal behaviour for one specific unit.
Frequently Asked Questions
An FSM system moves people and jobs around efficiently. It does not know what is wrong with the machine. These agents work on the diagnosis, the parts position and the knowledge, and they are built to sit alongside a scheduling system rather than to replace one.
Where this fits
Autonomous field service
The same agents running as one continuous loop, and the decisions that stay with people.
Telemetry Analysis tour
The predictive layer: correlated signals that precede a failure, scored per unit.
Cognitive Spares tour
The parts side: predicted demand per part and depot, with SLA coverage scored per customer.
More Agent to Agent Workflows tours
Field Service
Start where your first-time fix rate hurts
Tell us where jobs actually fail today — the diagnosis, the part, or the person who was sent — and we will show you which agents change that, and in what order.
Talk to us