The Future of Facilities Management: From Building Data to Accountable Decisions

Research perspective | Logic Foundry | October 8, 2026. This article synthesises published evidence and proposes an operational design approach. It is not a peer-reviewed paper or a report of experiments conducted by Logic Foundry.

Abstract

Facilities management is moving toward richer sensor information, building analytics and AI-assisted decisions. The operational challenge is to turn those capabilities into work that people can understand, own and verify. Our central hypothesis is that the value of a building signal depends on the quality of the decision process surrounding it: asset context, evidence, responsibility, action and follow-up. This article separates published findings from our design proposals and outlines a small evaluation that a multi-building team could undertake.

1. Why the operating layer matters

The International Energy Agency's Energy Efficiency 2025 report states: “Buildings account for around 30% of global energy demand.” Its buildings chapter discusses global consumption in 2024 and distinguishes residential demand from commercial and public buildings. This is a global sector statistic, not the energy share of a particular property portfolio and not an estimate of avoidable waste. [1]

For a facilities manager, the practical implication is to treat daily operation as a subject worth measuring. A capital upgrade may change the equipment available, but schedules, occupancy, maintenance condition and operating decisions continue to shape how that equipment is used. A useful future FM platform should help teams examine those relationships without assuming that every unusual reading represents a fault.

2. Evidence that analytics can support improvement

Berkeley Lab's 2020 report, Proving the Business Case for Building Analytics, describes the 2016–2020 U.S. Smart Energy Analytics Campaign. It covered 104 organisations and approximately 6,500 buildings. The report documented median annual energy savings of 3% for energy information systems and 9% for fault detection and diagnostics. These are separate technology categories and participant outcomes, not percentages to add together. [2]

The campaign combined commercially available analytics with energy-management and monitoring-based commissioning practices. It was not a randomised evaluation of CORE-FM or a promise for every building. Its geography, participating organisations, implementation practices and historical period limit direct transfer to another portfolio. The finding supports a narrower conclusion: analytics used within an operating programme can be associated with measurable improvements. It does not establish that installing a dashboard alone produces them.

The U.S. Department of Energy describes re-tuning as identifying and correcting operational problems using building automation data. Its process includes collecting information, analysing trends, walking the building, implementing appropriate actions, reporting and continuing the practice within operations and maintenance. This is federal-building guidance, not a universal savings guarantee or a substitute for qualified site-specific engineering. [3]

3. Design hypothesis: every signal needs a decision record

Our proposal is to organise a potential issue around six connected elements: the building and asset; the observation and its source; the operating context; the proposed action and reasoning; the person responsible; and the evidence needed to close the issue. This is a design hypothesis for evaluation, not an established industry standard or a tested Logic Foundry result.

Consider a hypothetical air-handling unit running outside its expected schedule. The first task is to confirm that the point maps to the right equipment and that its timestamp and units are understood. The next is to check whether the schedule reflects a legitimate occupancy change. Only then should a responsible operator decide whether an investigation or authorised adjustment is appropriate. Closure should record what changed and whether the observed behaviour changed afterward.

This approach also helps with maintenance. A completed inspection is an activity record. A finding linked to an asset, assigned action and verified repair is a fuller decision record. Across a portfolio, that distinction can make repeated issues easier to examine without confusing a high task-completion rate with demonstrated reliability.

4. AI should make its evidence inspectable

NIST's 2023 AI Risk Management Framework identifies characteristics including “valid and reliable” and “explainable and interpretable.” The framework is voluntary guidance for AI risk management; it does not certify any FM product. Its relevance here is a design principle: evaluate an AI system in the context of its intended use, rather than treating a convincing answer as proof of correctness. [4]

For an AI-assisted work-order proposal, the facilities team should be able to inspect the observations that prompted it, the asset mapping, relevant history and information that is missing. The output should distinguish an observed condition from a possible explanation. Where evidence is weak, the appropriate response may be a request for inspection rather than a diagnosis. Operator corrections should remain traceable so that future evaluations can examine the same failure cases.

5. Digital twins need a defined purpose

NIST's July 2026 Digital Twins Workshops Summary Report concerns manufacturing. It identifies interoperability, verification, validation, uncertainty, cybersecurity and workforce readiness as persistent challenges. Applying those concerns to buildings is our inference, not a building-specific finding from that report. [5]

A useful FM twin should begin with the decision it is meant to support. A model for evaluating operating schedules has different requirements from a model used to predict component failure. Define the scope, required data, update frequency and acceptable uncertainty before investing in a complex visual representation. A model that appears current while relying on stale or incomplete data can create misplaced confidence.

6. BMS integration and future energy controls

Integration must preserve meaning as well as transport values. Our proposed design checklist includes point identity, units, time zones, quality indicators, equipment relationships and an explicit distinction between observation and control authority. A read-only observation integration and a system permitted to write control commands require different operating and security arrangements.

Future energy-efficient controls should be evaluated against comfort, ventilation, equipment constraints and the authorised operator's ability to intervene. The test should include lost connectivity, stale sensors, incorrect mappings and changed occupancy. A future reporting tool should expose those limitations, rather than silently replacing missing evidence with assumptions. These are research design priorities, not claims that Logic Foundry has implemented autonomous building controls.

7. A practical evaluation a team could try

Select one building, one asset family and one recurring operational issue. Keep existing safety procedures and control responsibilities in place. Establish the observation period and baseline before reviewing results. Record occupancy, weather where relevant, equipment changes and missing data so that apparent improvements are not attributed to the wrong cause.

For a short observational pilot, record each flagged condition, whether a qualified operator judged it actionable, the next step assigned and the evidence collected after action. Track the proportion of useful alerts, duplicate or misleading alerts, time to an owned action and unresolved findings. Energy can be an additional outcome where measurement and comparison are suitable; a short pilot should not be presented as proof of annual savings.

Compare the proposed process with the team's existing workflow. Retain examples where the process failed: the wrong asset, an unexplained recommendation, an action with no owner or a closure without verification. Those cases are useful design evidence. This is a suggested evaluation for readers, not an experiment we claim to have completed.

8. Discussion and Logic Foundry's research direction

Two useful discussion prompts are: Which evidence must accompany a recommendation before your operators will act on it? Where does your current process lose continuity between a building observation and a verified outcome? A productive review can also begin with one unresolved issue and reconstruct its path through the organisation.

CORE-FM currently connects properties, assets, work orders, preventive maintenance and sensor monitoring. Logic Foundry's confirmed research areas include predictive maintenance, AI-assisted work orders, sensor anomaly detection, energy optimisation, digital twins, BMS API integrations and future smart controls, reporting and decision tools. We are working to better understand how those ideas can support accountable FM decisions. These research areas do not establish shipped integrations, autonomous controls, measured savings or release dates.

Our intended role is to connect useful building information with practical operational context, while making uncertainty and responsibility visible. The future of FM should be judged by the decisions teams can make and verify, not simply by the amount of data they can collect.

References are linked below. Evidence was reviewed on October 8, 2026. Statistics retain the geography and scope of their original sources.