A practical guide to connecting building systems, governing operational data and piloting AI safely in Singapore facilities.

Infographic showing connected Singapore facility systems including ACMV, electrical, pumps, lifts, fire protection and IoT feeding into a governed AI facilities-management workflow.

Artificial intelligence can help facility teams detect abnormal equipment behaviour, prioritise maintenance and automate selected work-order processes. However, AI is only as reliable as the operational data, system integration and governance behind it.

For Singapore building owners, warehouse operators and facilities-management SMEs, the practical question is not simply whether to adopt AI. It is how to connect existing systems safely and consistently before introducing AI into daily operations.

Why Singapore’s Open Digital Platform matters

On 9 July 2026, JTC announced a collaboration with Panasonic to develop and validate AI-enabled smart infrastructure and facilities-management solutions using the Open Digital Platform at Punggol Digital District. According to JTC, the platform aggregates real-time data from more than 20,000 sensors and supports capabilities including a digital twin, predictive-model training and virtual test-bedding.

This provides a useful reference architecture for the wider facilities-management sector. The lesson is not that every building needs to replicate Punggol Digital District. Rather, it shows the value of treating building data as a connected operational layer instead of keeping each system in a separate information silo.

In a typical facility, useful signals may sit across the building-management system, ACMV equipment, electrical distribution, fire-protection systems, lifts, pumps, access control, energy meters and standalone IoT devices. If these systems use inconsistent asset names, timestamps, locations and permissions, an AI model may struggle to identify the operational context behind an alarm.

Interoperability comes before automation

Before selecting an AI platform, facility managers should establish how data will move between systems and how people will interpret it. A practical interoperability blueprint should address five areas.

1. Create a common asset register

Start with a clear list of assets, locations, equipment types, system owners and criticality. A pump should not appear as “P-01” in one database, “Pump Lobby 1” in another and “Asset 0034” in a third without a reliable cross-reference.

Define consistent identifiers for equipment such as air-handling units, chillers, distribution boards, fire pumps, lifts and water pumps. Include relationships where useful—for example, which sensor belongs to which asset, which asset serves which zone and which equipment depends on another system.

2. Standardise the data needed for decisions

Not every data point needs to be collected or integrated immediately. Identify the information required for a specific operational decision: temperature trends for ACMV optimisation, vibration or run-time information for rotating equipment, current and energy readings for electrical monitoring, or alarm and inspection records for fire-protection workflows.

Define basic rules for units, timestamps, sampling frequency, quality status and missing values. Historical maintenance records should also be linked to the correct asset wherever possible. These details can make the difference between a useful anomaly alert and a noisy list of disconnected warnings.

3. Separate OT and IT responsibilities

Operational technology controls physical equipment, while information technology manages applications, networks and business data. Connecting the two requires careful boundaries.

Facility teams should document which systems are read-only, which systems can issue commands and which actions require a person to approve them. A sensible early-stage design may allow an AI service to read trends from a BMS or equipment database, generate an explanation and recommend a work order, while leaving control changes and equipment shutdowns to authorised personnel.

Network segmentation, account management, logging, backup arrangements and vendor access should be considered before connecting legacy controllers or remote monitoring services. These technical decisions should involve the relevant building, engineering, IT and cybersecurity stakeholders.

4. Build permissions and anonymisation into the design

Operational data may include information about occupants, contractors, access events, work activities and business operations. Use role-based access so that users see the data required for their responsibilities, rather than a complete copy of every system.

Where analytics do not require personal information, remove or anonymise it. For example, an equipment-health model may need temperature, pressure, run-time and alarm history, but not the identity of a person who entered a plant room. Define retention periods and maintain an audit trail for data access and workflow decisions.

5. Keep human approval in the loop

AI anomaly detection can support engineering judgement, but it should not automatically be treated as proof of equipment failure. A model may identify an unusual pattern caused by a sensor fault, temporary operating condition, planned maintenance or a change in occupancy.

For early pilots, use a human-in-the-loop workflow: the system detects an issue, provides the relevant trend and asset context, recommends an action and routes it to an authorised engineer or supervisor. The person confirms, rejects or reclassifies the recommendation. This feedback can improve the workflow while preserving accountability.

A phased pilot plan for Singapore facilities

Building owners and FM SMEs can reduce risk by starting with one operational problem rather than attempting to connect every system at once.

  1. Define the business outcome. Choose a measurable objective such as reducing nuisance alarms, improving response to ACMV faults, identifying abnormal pump behaviour or shortening the time needed to prepare a work order.
  2. Select a bounded asset group. Begin with one building zone, plant room, warehouse area or equipment family. Confirm data ownership, system interfaces and the people responsible for responding to alerts.
  3. Map the data and interfaces. Document sources, asset identifiers, signal quality, update frequency, access permissions and OT/IT boundaries. Identify gaps before committing to model development.
  4. Run in observation mode. Allow the AI service to generate alerts and explanations without automatically changing equipment settings. Compare its recommendations with actual site conditions and maintenance findings.
  5. Introduce controlled workflow automation. Once the alert quality is understood, allow approved recommendations to create draft work orders, assign priority or request inspection. Keep high-risk actions subject to human approval.
  6. Review and scale. Measure useful indicators such as alert precision, response time, rejected recommendations, repeat faults and data completeness. Expand only when the operating model is stable.

What smaller operators should prepare now

A smaller facility does not need a large digital twin programme to become AI-ready. It should first establish reliable asset records, clean naming conventions, accessible maintenance history and a documented list of system interfaces.

It is also useful to identify an integration owner who can coordinate the building owner, FM team, equipment vendors, system integrator and IT stakeholders. Without clear ownership, a pilot can produce a technical demonstration but fail to become a usable operating process.

Singapore’s wider smart-built-environment direction reinforces this approach. BCA’s Smart Facilities Management and AI for the Built Environment resources highlight integration, data-driven operations and structured adoption. MOM’s WSH technology guidance also identifies practical categories such as electronic permit-to-work, video analytics and IoT environmental monitoring. These technologies still depend on appropriate governance, site procedures and competent human oversight.

From disconnected systems to dependable decisions

The main value of an open digital approach is not simply the number of sensors or the sophistication of an AI model. It is the ability to connect trustworthy information to a clear operational decision.

For Singapore facility managers, the right starting point is therefore practical: map the assets, standardise the data, define permissions, protect OT systems and test one workflow with people involved at every stage. Once that foundation is in place, AI can be introduced as a controlled support layer for anomaly detection, maintenance planning and selected automation.

ISS can help businesses assess engineering data, facility-management workflows and AI automation opportunities before moving into implementation. Contact ISS to discuss your engineering, facility management or AI automation requirements.

Sources and further reading