A facilities maintenance request starts with "the pump sounds wrong" and a photo sent to a site manager. Another arrives by email from a tenant. A contractor phones about access. The asset register holds the service history, while the permit-to-work record sits in a shared folder and the purchase order lives in finance.
Everyone is busy, but nobody can see the whole job.
A connected facilities maintenance system can turn those messy reports into clear jobs. It brings each work request together with the asset record, risk assessment, permit, contractor evidence, purchase order, and return-to-service decision. AI can read the first message and propose the details. Named people still decide the risk, authorise the work, and confirm that equipment is safe to return to service.
Make every request complete enough to route
People reporting faults rarely know the information a maintenance planner needs. They describe the symptom in their own words and use the location name familiar to them.
A useful intake process captures the original message, then helps establish:
- The site, area, and likely asset.
- The observed symptom and when it started.
- Whether the asset is running, isolated, or unavailable.
- Any immediate risk to people, property, production, or service.
- Photos, documents, and contact details.
- The person who reported it and who is present on site.
AI can classify an email, extract a location from a message, suggest an asset match, or summarise a long history. The result should be presented as a proposal. A site reference, asset tag, or manual confirmation can validate the match before work is raised.
This is a good fit for AI because language varies and the output is bounded. The system should refer unclear requests rather than invent a confident asset or risk category.
Avoid asking the reporter to complete a long universal form. Conditional questions are better. A water leak, access-control fault, and unusual plant noise need different evidence.
Show what is urgent and who can decide
A maintenance job can be urgent for different reasons. People may be at risk. A customer service may be about to fail. A room may need to be ready before staff arrive. Those reasons affect who should act and what they are allowed to do.
The system should show the reason for the priority in plain language. A red label on its own does not tell a manager whether to isolate equipment, call a contractor, or wait for an authorised person.
AI can spot words that may signal danger or disruption and bring the request to a person's attention. It cannot approve the safe way to carry out the work, declare equipment safe, or reduce the priority because earlier cases were harmless.
The HSE guidance on maintenance says maintenance should be planned, carried out by competent people, and supported by a safe system of work. Some high-risk jobs require a formal permit to work.
Send each decision to the named person who has the authority to make it. Give them the request, photos, asset history, and risk information, then record what they decided and when.
Keep permits and competence visible
Treat the permit as a live operational control. Attaching a PDF after the job misses its purpose.
HSE describes a permit to work as a formal system that states what work will be done, when it will happen, which parts are safe, and which precautions apply. It also coordinates site management, plant supervisors, operators, and the people carrying out the work.
The maintenance queue should show whether a permit is required, who can authorise it, its current state, and the equipment or area it covers. It should prevent the asset from being returned to normal service until the required declaration is complete.
It also needs to handle shift changes and contractor handover. HSE's permit-to-work guidance stresses clear responsibility, communication, and instructions when work continues beyond one shift.
Competence records belong near the assignment decision. The system can confirm that a named person or supplier has a current recorded qualification for the work. A competent manager still decides whether that evidence is sufficient for the actual task.
AI may summarise method statements or compare submitted evidence with a checklist. Approval remains an accountable human action.
Connect each job to asset and supplier records
Another isolated maintenance application will create more reconciliation unless it connects to the records already in use.
The system may need links to:
- The asset register for identity, location, criticality, and service history.
- Building or production controls for alarms and operating state.
- Procurement for approved suppliers, orders, and spend.
- Access control for visits and restricted areas.
- Document storage for manuals, risk assessments, and evidence.
- Finance for cost allocation and invoice matching.
The integration does not have to copy everything. It should fetch the facts needed for the decision and write back the event each system is meant to own.
An asset register may remain authoritative for planned service dates. The connected system owns the live fault, responsibility, status, and evidence. Procurement owns the purchase order. Clear ownership avoids several versions of "complete".
For older systems, a daily file or monitored database view may be enough. Use real-time integration only where a delay would change safety or service.
Measure delay and repeat work
A good system should reduce waiting and missing information. Digitising the existing inbox leaves the underlying problem in place.
Track the moments that expose friction:
- Time from report to complete triage.
- Time waiting for access, evidence, approval, parts, or a contractor.
- Jobs reopened after an incomplete repair.
- Repeat faults against the same asset and symptom.
- Planned work displaced by avoidable reactive work.
- Cases where the first visit lacked the right skill, part, or permission.
These measures help an operations director decide whether to improve intake, supplier response, stock, asset data, or the maintenance plan. They also show whether AI-assisted triage saves work or creates more checking.
Start with one site or asset class. Map ten recent jobs from first report to return to service, including every wait and handover. Build the first version around the missing facts and decisions revealed by those real cases.
ORBN can help design and build this kind of connected facilities maintenance system. We can integrate the facilities or maintenance software, asset register, documents, procurement, and finance records you already use, then build the missing workflow around the decisions your operation needs to control.
If maintenance requests still disappear between inboxes, folders, and systems, bring us ten difficult jobs. We can trace where context or responsibility is lost and identify the smallest useful first release.