Maintenance Risk Agent
Says which asset fails next, and which service is causing failures
- 8hard rules
- Things it must never do, enforced in every run
- 5handover rules
- Conditions where it stops and gives the work to a person
- 9output fields
- A structured result, not a paragraph you have to re-read
- ≤6steps per run
- Bounded work: at most 4 tool calls, enforced by the runtime
Read-only by design. This agent has no path to sending, spending, publishing or committing anything. Where that limit is the product, removing it would remove the reason to trust the output.
Overview
Preventive maintenance intervals come from a manufacturer's recommendation written for an average machine in average conditions. The asset in front of you is not average — it runs harder or barely at all, in dust or damp, operated by people with different habits. The same interval is simultaneously too frequent for half the fleet, servicing things that did not need it, and too slow for the other half, which fails between services.
The evidence to correct this is already in the work order history and nobody reads it. Every organisation running assets has years of records showing what broke, when, after how many hours, and what was done just before it broke. That last one is the finding nobody looks for: failures clustering shortly after a service is not coincidence, it is induced failure, and it means the maintenance itself is causing downtime.
Two more patterns hide in the same data. The same fault fixed four times on one asset means the fix is wrong, not that the asset is unreliable. And an asset whose maintenance cost has passed what it is worth to keep is a replacement decision nobody makes, because no single line item ever looks large.
This agent reports all four, with the evidence and the confidence behind each. It schedules nothing and touches no maintenance system, and it never makes a safety determination.
Benefits
Failures prevented rather than attended
Which is the whole point of preventive maintenance.
Servicing that stops being wasted
Half a fleet is usually serviced more often than it needs.
Replacement decisions with evidence
Rather than when something finally dies.
What it does
Finds induced failures
Breakdowns clustering shortly after a service. Not coincidence — the maintenance itself is causing downtime, and nobody looks for it.
Corrects intervals to actual use
The manual describes an average machine. Yours runs harder, or barely at all, in conditions it never mentions.
Catches the fault fixed four times
A repeat failure means the fix was wrong, not that the asset is unreliable. Different problem, different action.
Names the asset past its economic life
Maintenance cost against replacement value. No single work order ever looks large enough to trigger the question.
States confidence per finding
Some assets have hundreds of records and some have nine. The report says which is which.
Schedules nothing
No work order is raised, changed or closed, and no safety determination is ever made.
What this changes
What this changes about your week
- It takes over
- Reading years of work orders to find what fails next and why.
- Done today by
- Nobody. Intervals come from the manufacturer's manual instead.
- How often
- Never, which is why intervals never change.
- Typical manual effort
- Correlating every failure against usage, service history and the fleet baseline.
Why Field Services in particular. Failures clustering shortly after a service means the maintenance is causing downtime — the one pattern nobody looks for, because it inverts the assumption the whole schedule rests on.
Runs unattended
Started by you or by an event, and it finishes on its own. Nothing waits for someone to be at a desk.
The same standard every time
The two-hundredth item is held to the bar the first one was. Consistency is the part people cannot sustain.
It cannot act on its own
Maintenance Risk Agent has no path to sending, spending or committing. That limit is why its output is safe to act on.
How it runs
This agent runs server-side through the PROMIVO runtime. Each run is logged step by step and every tool call is permission-checked before it executes.
- 1InputValidated against the agent input schema
- 2Modelclaude-opus-5 with automatic fallback
- 3Toolsextract_structured_data
- 4PermissionsRuns READ actions automatically; anything higher waits for approval
- 5OutputStructured result matching the output schema
- 6LoggingTokens, duration and every step recorded
Demo
Demo dataIllustrative sample output, abridged.
Input
{
"period": "2024–2026",
"currency": "USD",
"workOrders": []
}Output
{
"escalate": true,
"disclaimer": "An analysis of your own maintenance records. No work order has been raised or changed, no maintenance system was accessed, no asset has been assessed as safe or unsafe to operate, and no interval change is recommended for any safety-critical or statutorily inspected system.",
"repeatFaults": [
"GEN-02 has had the same coupling fault repaired 5 times in 18 months. That is a fix that does not hold, not an unreliable asset, and replacing the part again will not change it."
],
"performingWell": [
"The packing line assets have had no unplanned failures in two years at a 400-hour interval. There is room to lengthen it, and nobody has asked."
],
"inducedFailures": [
{
"pattern": "Across the conveyor fleet, 22 of 61 unplanned failures occurred within 40 running hours of a scheduled service. The fleet baseline for any 40-hour window is 4. The servicing is causing failures.",
"confidence": "high",
"windowHours": 40,
"failuresWithinWindow": 22
}
],
"escalationReason": "Failures are clustering after servicing across the conveyor fleet, and one asset has had the same fault recur five times.",
"intervalFindings": [
{
"asset": "PUMP-14",
"evidence": "Runs at 3.1x the fleet average usage. Four failures in two years, each between 610 and 690 hours after service — consistently before the interval is reached.",
"direction": "too-long",
"confidence": "high",
"safetyCritical": false,
"currentInterval": 750
},
{
"asset": "COMP-03",
"evidence": "Serviced 11 times in two years with no fault found on 9 of them. Runs at a third of fleet average usage.",
"direction": "too-short",
"confidence": "moderate",
"safetyCritical": false,
"currentInterval": 500
}
],
"pastEconomicLife": [
{
"asset": "FORK-07",
"downtimeHours": 310,
"cumulativeCost": 41200,
"replacementValue": 28000
}
],
"insufficientHistory": [
"Six assets have fewer than 10 work orders. No prediction is offered for them."
]
}Requirements
Integrations
No integrations required.
Before you start
- Work order history: asset, date, fault, action taken, cost, downtime
- Asset register with running hours or usage, and replacement value
- Current maintenance intervals
- Requires the All-Access plan or higher
Use cases
Annual maintenance planning
Intervals set from your own history rather than the manual.
A rising downtime trend
Whether it is age, intervals, or the servicing itself.
Capital planning
Which assets have passed the point of being worth repairing.
After acquiring a fleet
What the records say about what you just bought.
Pricing
All-Access
Billed monthly through your PROMIVO subscription. Cancel at any time.
Runs consume your plan allowance for agent executions and tokens. See plan limits
FAQ
Does it schedule maintenance?
No. It raises, changes and closes nothing, and has no access to a maintenance system. It reports what the history shows.
Does it make safety decisions?
Never. Where a pattern touches a safety-critical system it escalates immediately to a person. It gives no safety determination of any kind, in any circumstance.
How much history does it need?
More is better and it tells you when there is not enough. A prediction from nine work orders is reported as low confidence rather than presented alongside one drawn from four hundred.
Reviews
Reviews open once customers have run this agent.
Need this agent to work differently?
Tell us what to change and our team will scope a customised version for your business.
Customize this agent