The FM Accountability Stack is a three-layer operating model that defines how FM organisations execute work, make decisions, and assign accountability for outcomes. The three layers are Activity (what was done), Intelligence (what should be done), and Governance (who owns whether it was right). The governance gap it addresses has always existed in FM — AI makes it impossible to ignore.
CMMS platforms, IoT sensors, BMS integrations — the data exists. Work orders, inspections, asset records. What these systems cannot answer is whether any of it is working.
A completed work order is proof of compliance. Not proof of value.
Who decided that asset's maintenance interval? Who set that SLA threshold? Who owns whether that prioritisation call was right?
In most FM operations those decisions are made by habit or experience — not by any defined, reviewable logic. AI makes it urgent. But the gap was always there.
The operational surface of FM — work orders, inspections, technician dispatches. Most FM technology is built here. Accountability is person-level: a technician completes a task, a supervisor signs off.
This creates a record of execution. But only of execution. A completed work order is proof of compliance — not proof of value. Activity metrics measure process adherence, not decision quality.
What was done · Who did it · When
Whether the work was right · Whether it prevented failure · Whether it improved outcomes
The person who performed the task
Where data becomes direction — patterns identified, predictions formed, recommendations generated. This layer answers: what should happen next?
Accountability belongs to the logic behind the system — not the system itself. Every recommendation depends on thresholds, rules, and goals set by humans. If those are misaligned, the system produces poor decisions consistently. The system does not own the decision. The configuration does.
What is likely · What action is recommended · What outcome to expect
That the logic is still valid · That conditions haven't changed
Whoever configured the decision logic
The most critical — and most absent — layer. Governance is not reporting. It is the operating layer that defines who owns decision logic, who reviews whether that logic produces the right outcomes, and who has authority to change it.
Without it, activity produces unvalidated records and intelligence produces unowned recommendations. Governance is what makes both layers accountable.
Decisions are not ad hoc · Logic is validated · Outcomes are traceable
Named outcome owners · Review cadences · Explicit change authority
Leadership responsible for the portfolio
When procedure is followed, work orders close on time, and an asset still fails — the failure sits not in execution, but in the decision logic no one owned.
Asset life, cost efficiency, SLA performance, energy, tenant experience — these must be named, owned by a function, and reviewed on a cadence. Without this, there is no basis against which any decision can be evaluated.
Outcome ownership is explicit · Targets are assigned · Review cadence is set
A System of Decisions holds these outcomes inside the operating layer — not in a policy document separate from where decisions are made.
An ongoing comparison between the logic in use, the actions being taken, and the outcomes produced. Any drift between these three is a governance signal — whether the decision was made by a person or a system.
Decision quality is tracked · Drift is caught early · Logic stays current
A System of Decisions makes this comparison continuous — not a quarterly review that happens after drift has already compounded.
When conditions change — new assets, new regulations, new cost pressures — decision logic must be updated. The governance layer names who makes that call, on what basis, and how frequently.
Change authority is named · Updates are traceable · No rule changes without ownership
A System of Decisions gives change authority a place to live — and creates a record every time it is exercised.
Reactive operations. Work is done but not optimised. The organisation responds to what has happened — not what is about to. Effort is high; impact is uncertain.
Unowned decisions. Recommendations are generated but no one formally owns the logic behind them. When something goes wrong, accountability has nowhere to land.
Oversight without substance. The review structure exists but the underlying data and logic are too weak to govern meaningfully. Governance becomes procedural, not consequential.
© Xempla. All rights reserved.
Terms & Conditions