Raises cyclical work when it is due rather than when it fails, books it across the portfolio while there is slack in the diary, and keeps the schedule from quietly slipping a year.
An AI agent for planned maintenance runs the cyclical side of a portfolio: boiler services, gutter clearing, communal cleaning, alarm testing, pump and lift servicing, external decoration cycles. It watches when each item is next due, raises the job with enough lead time to be scheduled comfortably, arranges access where a visit enters somebody's home, and reports what is overdue. Planned work loses every contest for a coordinator's attention, because reactive work has a tenant attached and planned work does not. So it slips, quietly and without anybody deciding to let it, until the boiler that was not serviced fails in February and becomes an emergency call-out at four times the cost with a cold family waiting.
Gives cyclical work the deadline it does not naturally have.
Tracks the next due date for every cyclical item on every property
Raises the job with lead time rather than in the week it is due
Groups nearby properties so a contractor can run them in one trip
Arranges tenant access for anything entering an occupied home
Records completion and rolls the next due date forward
Reports what is overdue, so the slippage is visible while small
A serviced boiler and an unserviced boiler cost the same to own until one of them fails at the worst possible moment, at emergency rates, with a habitability problem attached and a landlord asking why. The economics of prevention are not seriously disputed by anyone in property management; the reason it does not happen is purely that cyclical work never shouts. It also has a second-order effect that shows up in the annual numbers rather than the daily ones: an office running heavily reactive is an office whose contractors are always being asked to attend urgently, which is the most expensive way to buy anybody's time and steadily erodes the goodwill you need when something genuinely is an emergency.
A three-step flow around a rolling maintenance calendar.
The agent tracks the next due date for each item on each property and opens the job at the lead time you set for that type of work.
It groups due items by area and trade so one visit covers several properties, then places and schedules the work like any other job.
On completion it files the certificate or report, sets the next due date, and adds anything still outstanding to the overdue report.
Annual boiler servicing across a street of managed properties.
Scenario: an office was booking boiler services individually when somebody noticed them, and was consistently running a backlog into the winter. In August the agent flags eleven boiler services falling due between October and December, and notices that seven of them sit within two adjacent postcodes. It groups those seven and offers the batch to a gas engineer on the panel, who takes the lot for a fortnight in September. Access is then arranged tenant by tenant, and two of the seven turn out to be difficult: one tenant works away and one refuses the first three windows offered. Both are flagged early enough that they can be worked around rather than becoming a December problem. The remaining four properties are scheduled individually. By the end of September ten of the eleven are complete, and the one outstanding sits on the overdue report with the reason attached rather than being discovered in January.
Anybody whose winter is decided by what happened in September.
Planned work loses to reactive work every single day unless something protects it.
Communal plant carries servicing cycles that residents never see until they fail.
Preventive schedules are the whole job and the easiest thing to let slip.
Emergency call-out rates are what turn a repair budget into a bad year.
Cyclical programs across thousands of homes cannot be tracked by memory.
Identical units mean identical cycles arriving in large simultaneous batches.
Holds the cycle, books the batch, files the outcome.
Holds each property's cyclical items, intervals and next due dates.
Books the planned visits alongside reactive work so clashes are visible.
Offers batched work to contractors and receives the completed reports.
Arranges access with tenants for anything entering an occupied home.
Stores service records and certificates filed against the property.
Builds the overdue report the office reviews on a regular cadence.
The cycles that only become visible when they are missed.
Questions about protecting work that has no tenant chasing it.
An AI agent for planned maintenance runs the cyclical side of a portfolio — servicing, clearing, testing, decoration cycles. It watches when each item is next due, raises the job with lead time, groups nearby properties, arranges access, records completion and reports what is overdue.
Compliance work has a legal deadline and a certificate; planned maintenance is discretionary work you do because it is cheaper than the alternative. The mechanics of scheduling overlap heavily, but the consequence of missing one is a penalty and the other is a failure at an inconvenient moment.
Enough to absorb a contractor being busy and a tenant refusing two windows. Six to eight weeks suits most annual work. Raising a job in the month it is due converts planned maintenance back into urgent maintenance, which defeats the purpose entirely.
It is one of the few genuine wins available here, because travel is a real cost a contractor prices in whether or not you see it itemized. It only works where the properties are close and the timing is flexible, which is exactly what lead time buys you.
That is their decision to make and it should be recorded plainly. The useful contribution is that the refusal is on file with a date, so when the unserviced item fails there is no argument about whether it was raised. Do not let a refusal silently delete the item from the cycle.
For anything entering an occupied home, yes, and this is where planned schedules actually break. The engineer is rarely the constraint; the tenant who works and has already declined two windows is. Starting early is what makes that solvable rather than terminal.
It converts drift into a number. Most offices know planned work slips but cannot say by how much, which makes it impossible to argue for resource. A count of overdue items by type and age is the thing that turns a vague concern into a case somebody can act on.
Raises cyclical work when it is due rather than when it fails, books it across the portfolio while there is slack in the diary, and keeps the schedule from quietly slipping a year.