Home · Solutions · Operations & quality
Solution · Operations & qualityWork orders leave the plan on time, the part is on the shelf, the report closes the same day
Preventive maintenance that runs to plan
The maintenance plan releases its own work orders against the production schedule, the parts are reserved before the technician walks to the machine, and the completion closes the order the same day.
Executive summary
Stop releasing work orders by printer and typing the completion reports back a week later.
The maintenance plan stays exactly where it is; what we build sits around it.
Administration per work order falls from about 14 minutes to a short review, and the time returns to planners, storemen and technicians.
the CMMS or SAP S/4HANA Asset Management for orders, confirmations and equipment records; purchasing for requisitions; the SharePoint certificate library
Business problem
Maintenance
Preventive maintenance exists because the alternative is more expensive, and in a food plant it also exists because the law says so. Steam, refrigeration and lifting equipment carry inspection dates that do not move; filters, seals and calibration checks carry intervals an engineer set once and the system generates forever. The plan is usually sound. What fails is everything between the plan and the machine.
The plan knows neither the production schedule nor the store. An order for line three in week twelve is a proposal; whether line three stops that week is settled at the planning meeting, on a different screen. Somebody reconciles the two calendars by hand, and maintenance usually loses, because a running line is a fact and a due service is an intention. Free stock is checked when the technician reaches the machine and finds the box empty.
Closure is the quiet half. The completion report is written on paper and typed in days later by someone who was not there, so readings get rounded and observations worth following up do not survive. The history that results cannot answer what matters most: is this asset failing more often than it used to, and should we keep repairing it.
How it works today
The shape below repeats across food, chemicals and packaging, whatever the maintenance system is called.
- SystemThe maintenance plan generates the coming period's work orders on a fixed schedule
- PersonThe planner compares them with the production plan in a second window and decides by hand which can go ahead
- PersonThe store is asked about the parts; a missing one becomes a phone call to purchasing and, sometimes, a requisition
- WaitingWork that collides with a production run is pushed to next week, and often to the week after
- PersonOrders are printed for the technicians, who write the completion on the sheet: tasks done, readings, parts used
- WaitingSheets wait in the tray until a planning assistant has time, typically three to seven days
- Risk of errorOrders closed without a reading or without the parts booked leave the equipment record wrong, and the next due date is calculated from it
Why the current process costs more than it appears
The cost grows where nobody is looking.
- Rescheduling is the expensive part of planning. Every collision is handled twice, once by the planner who moves the order and once by whoever explains the compliance figure at the monthly review.
- A part missing on the day costs far more than the part: the walk to the machine, a production window that cannot be sold again, and an express delivery at whatever the supplier asks.
- Deferred orders do not spread evenly. They pile up against the assets hardest to stop, so an acceptable plant average hides the machines that matter most going unserviced.
- Equipment histories written a week late are written from memory, and rounded readings support neither a warranty claim nor a decision to replace.
- Statutory inspections queue behind filter changes. When the queue slips the inspection slips with it, and nobody notices until somebody asks for the certificate.
Cost of inaction
Deferred maintenance behaves like a loan. The plant borrows an hour of production by pushing an order to next week and repays it later with interest, in an unplanned stop that takes a shift. The rows above price only the administration, because that is what a spreadsheet can count. Beside it, in no budget line, sits the compliance rate: orders released late are done late or not at all, and the asset due in week twelve runs until something on it decides the date instead.
The second cost accumulates in the equipment record. A history written from memory a week after the job cannot support a warranty claim, answer an inspector, or settle whether a fifteen-year-old filler should be repaired again or replaced. That gets decided on opinion, about assets worth far more than everything above.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
A food producer with two plants in Central Europe, about 640 registered pieces of equipment across process lines, packaging, refrigeration, steam and utilities; 1,300 employees; an ERP with a maintenance module; Microsoft 365 E3 in the offices, F3 on the shop floor.
1,100 planned work orders a month, from weekly lubrication rounds to annual overhauls, about 14 minutes of administration each; one order in six needs a part that is not in free stock.
The plan generates orders, a planner reconciles them with the production schedule in a second window, the store is asked by phone, orders are printed, and an assistant types the sheets back in days later.
Release and closure, not the work itself. Orders leave the plan late, arrive without a confirmed part, and return to the record a week after the job.
Robots release the plan's orders into windows the production calendar leaves free, reserve the parts and requisition what is short, put the job card in Microsoft Teams with the checklist and the asset's history, then close the order and the equipment record the same day.
In the modelled case administration falls from about 14 minutes an order to a short review, orders reach the technician the day they are released, and compliance becomes a current figure rather than a monthly reconstruction; every number here is modelled on the assumptions above.
Proposed solution
The maintenance plan stays exactly where it is; what we build sits around it. On the release run a robot takes the orders the plan has generated, reads the production calendar and the shift plan, and books each into a window the line is not running. Orders that collide, and any order whose asset carries a statutory date that period, become a task for the planner in Microsoft Teams: three proposed windows, the consequence of each, one click. The same run checks free stock for every part, reserves what is there and raises a purchase requisition where the reserve level would be breached. A robot raises requisitions; it never approves one.
The technician's day then starts in a Power Apps job card inside Teams: today's orders, the checklist for each, the asset's last five interventions, and the reserved parts with their location in the store. Completion is captured on the same screen, with readings, parts used, photographs and anything noticed outside the order's scope. A robot posts the confirmation, books the parts, updates the counter and lets the plan calculate the next due date from a real number rather than an assumed one. Nothing in the flow interprets free text: every decision is a rule the maintenance team can read, change and defend. This is the planned side only; the alarm at night and the diagnosis that follows have their own page in this library.
UiPath Orchestrator time triggers, queues, retries and audit; UiPath Integration Service connectors for SAP (BAPI and OData) and Microsoft OneDrive & SharePoint; UiPath Action Center tasks completed inside Microsoft Teams; a Power Apps canvas app with camera control as a Teams tab; Microsoft Forms read by Power Automate; Adaptive Cards through the Workflows app; Power BI semantic model
The release rules against the production calendar, the collision and statutory-date exceptions, the parts and reserve-level logic with its requisition trigger, the job card and its checklists, the confirmation and counter-update logic, the escalation ladder and the report
Your ERP's maintenance module through UiPath SAP activities or your CMMS vendor's API; the production calendar from the planning system or its scheduled export
How the automated process works
- AutomationA robot pulls the orders the plan has generated and matches each against the production calendar and the shift plan
- AutomationFree stock is checked for every part; what is there is reserved, and a requisition is raised where the reserve level would be breached
- PersonOrders that cannot be fitted, and anything with a statutory date at risk, reach the planner in Microsoft Teams with three proposed windows and their consequences
- AutomationReleased orders are scheduled, assigned to a technician group and posted to the job card with the checklist, the asset history and the location of the parts
- PersonThe technician works the checklist on the phone, recording readings, parts used, photographs and anything noticed outside the order's scope
- AutomationA robot posts the confirmation, books the parts, updates the counter and lets the plan recalculate the next due date; certificates are filed against the asset
- AutomationOverdue orders escalate to the technician group and then the supervisor; Power BI refreshes compliance, overdue ageing, part-caused delays and repeat failures
Human-in-the-loop model
Automation handles
- Releasing and scheduling the plan's orders against the production calendar and the shift plan
- Reserving parts, raising requisitions at the reserve level and tracking them against the order's date
- Confirmation, parts consumption, counter readings, the next due date, escalations and reporting
People decide
- Which window a colliding order takes, and what production gives up for it
- Whether a checklist result is acceptable, and whether an observation becomes a new order or waits
- The rules themselves: reserve levels, escalation ladders, statutory dates and checklists stay with maintenance
Before and after
Systems and integrations
We do not add technology to make an architecture look serious. Every element below has a specific job in this process.
Inputs
- the maintenance plan and its orders
- the production calendar and shift plan
- free stock and reserve levels
- Microsoft Forms observations from operators
- results from external inspection bodies
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- UiPath Action Center
- Power Automate cloud flows
Target systems
- the CMMS or SAP S/4HANA Asset Management for orders, confirmations and equipment records
- purchasing for requisitions
- the SharePoint certificate library
- Power BI
Human touchpoints: the Power Apps job card in a Microsoft Teams tab; Action Center scheduling tasks; Adaptive Cards for assignment and escalation; the Power BI report in the plant channel
Technologies used
run the release, parts and closure jobs on schedule; queues, retries, audit trail
Areads the plan and order list; posts reservations, requisitions, confirmations
Athe planner's scheduling decision on colliding orders
Athe technician's job card: checklist, asset history, readings, parts, photographs
Ainspection results from external bodies and observations from line operators
AAdaptive Cards for assignment and escalation; Forms responses into the flow
Acompliance, overdue ageing, part-caused delays and repeat failures
Asystem of record for plans, orders, equipment and parts
AIllustrative economic model
What it is worth, with the arithmetic shown.
Nothing here prices the maintenance itself. The fourteen minutes belong to four people who never pick up a tool on the order: the planner who reconciles it with production, the storeman who looks for the part, the technician who fills in the sheet and the assistant who types it back. €28 is a blended fully loaded hourly cost across those roles in Central Europe. Both figures are illustrative rather than measured, and the model shows capacity released, not posts removed.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- Administration per work order falls from about 14 minutes to a short review, and the time returns to planners, storemen and technicians
- Orders leave the plan into a window the line is genuinely free, so the reschedule conversation happens once rather than weekly
- The part is reserved before the technician walks to the machine, and a shortfall becomes a requisition days earlier, at a normal price
- The equipment record is current on the day of the job, so the next due date follows the machine and not the calendar
- Statutory inspections are tracked as dates that cannot slip, and repeat failures surface while they are cheap to investigate
The management view
- Schedule compliance stops being a monthly reconstruction: the figure per plant and line is current and built from the records an auditor would read
- Deferred maintenance becomes a decision with a name on it rather than a silent slip, because every reschedule is a recorded choice with a reason
- The budget can be argued asset by asset, since parts, express orders and repeat failures attach to machines rather than to a cost centre total
Board-level KPIs
Security and governance
Control is not an add-on.
- Robots sign in to the maintenance system with their own named accounts, limited to the transactions the flow needs, so every posting carries the robot's identity in the system's log
- Secrets live in the vault connected to Orchestrator rather than in the packages, which makes rotation a store operation and not a code change
- Certificates, orders, job cards and photographs sit on SharePoint in your Microsoft 365 tenant inside the EU Data Boundary; the control plane runs in UiPath Automation Cloud, EU region
- Technicians see only their own plant's assets through group membership in Microsoft Entra ID, and statutory dates, reserve levels and escalation ladders are versioned with the maintenance manager's sign-off
Why now
European law on the use of work equipment requires periodic inspections by competent persons and that the results be recorded and kept available to the authorities; a history a week behind and written from memory is a weak answer to an inspector and a weaker one to an insurer
Energy and downtime cost more than they did, and food shift patterns leave narrower maintenance windows; the modelled 257 hours a month, about €7,200, is what the paperwork consumes before a single window has been negotiated
The building blocks are ordinary now: scheduled robots against an ERP interface, a canvas app in a Teams tab, cards through the Workflows app since the legacy Teams connectors were retired in May 2026, and reporting on licences you hold
Relevant executive roles
Planned work stops arriving as an ambush; the reschedule is a decision taken once, with the cost of each window visible to both sides
Compliance and overdue orders per line become current figures, and the argument about deferred maintenance has evidence behind it
Maintenance spend attaches to assets rather than to a cost centre total, which is what a repair-or-replace case needs
One scheduled automation against the existing maintenance module and one app on the tenant, not a second maintenance system nobody sponsors
Common questions and objections
It does, and it carries on doing so. What it does not do is check whether the line is running, whether the part is on the shelf and whether the completion ever arrived. Those three gaps are where orders slip; the plan stays the system of record.
Then the argument moves to where it belongs. When every collision is a recorded decision with the alternative windows attached, deferred maintenance stops being a maintenance failure and becomes a production choice the plant can price.
The job card replaces a sheet, not a conversation, and it is shorter: the checklist is prefilled, the last five interventions are on screen, and the parts listed are the reserved ones. Without coverage it submits on reconnection.
When this is not the right solution
- Fewer than a few hundred orders a month on a single site, where a disciplined planner and one shared calendar cost less than an automation
- The maintenance plan itself is wrong: intervals copied from a manual nobody has revisited, assets missing from the register, or parts lists that do not match the machines; the plan comes first
- The pressing problem is the unplanned stop rather than the planned one, in which case the alarm-to-resolution flow covered elsewhere in this library is the better start
A question for the next management meeting
Ask this company this afternoon which machines are overdue for a planned service and why: does the answer come back today, or after a week and three spreadsheets?
Implementation approach
What we deliver, and what we need from you to start.
We deliver
- Discovery on one plant: plan structure, order types, checklists, statutory dates, parts and reserve levels
- The release rules against the production calendar, with the collision and statutory-date exceptions
- The parts check, reservation and requisition logic, tied to reserve levels and supplier lead times
- The Power Apps job card in Teams, and robots for confirmation, parts consumption, counter updates and certificate filing
- The Power BI report, a runbook, then a pilot on one plant and rollout with hypercare
We need from you
- Three months of work-order history with release, due and completion dates and the parts consumed
- The current production calendar, and how far ahead it is firm
- A maintenance process owner, a production planner and a storeman to settle the rules
- Technical accounts for the maintenance system, plus your checklist templates
Stages
Discovery
Plan structure, order types, exceptions, parts and statutory dates with the maintenance team
Design
Release rules, reserve levels, escalation ladder, job card and security model
Build
Robots, job card, flows and report in your own environment
Validation
Replay of three months of real orders, exception handling, acceptance by planners
Go-live
One plant and one asset class under supervision, then the rest with hypercare
Optimisation
Rule tuning, checklist revisions, reliability review on the repeat-failure data
Departmental. Effort follows the number of order types and checklists, how firm the production calendar is, and whether the maintenance module has an interface or must be driven through its screens.
Next week's work orders are printed. Last week's are still waiting to be typed in.
Send us last month's work-order export with release, due and completion dates and the parts consumed. You get back a release-and-closure flow for one plant, plus the overdue and repeat-failure picture your own export already contains.
Time the admin behind one week of ordersThe neighbouring process usually has the same problem
Broken air conditioning, a badge, a desk move: all of it lands on one office manager's memory.
View solution Finance & accountingCapital projects that reach the asset register on timeYour newest production line is earning money and your books still call it a project.
View solution Operations & qualityThe daily production report without the 6 a.m. spreadsheetStop paying production controllers to retype MES numbers into a spreadsheet the meeting does not trust.
View solutionIndustries we deliver this in most oftenManufacturing & industryTransport & logisticsEnergy & utilities