Singapore Facility Teams Can Learn from the Softgrid–EM Engineering AI Facilities Partnership

Artificial intelligence is becoming more relevant to facility management, but the value does not come from buying a prediction tool alone. It comes from connecting asset data to people, procedures and engineering action.

A recent Singapore example is the memorandum of understanding announced on 22 September 2026 by Softgrid and EM Engineering. The partnership is described as an exploratory collaboration to examine AI-enabled estate and facilities-management applications, including intelligent lift monitoring, smart-building visibility and integrated facilities workflows.

For Singapore facility managers, warehouse operators, building owners and SMEs, the announcement provides a useful adoption lesson. It shows why AI projects should be assessed as operational improvement programmes rather than standalone software purchases.

What the Softgrid–EM Engineering announcement actually establishes

According to the announcement, Softgrid and EM Engineering will explore the use of AI predictive intelligence, edge IoT, video analytics and integrated monitoring across mutually agreed estate and facilities-management projects. The initial areas mentioned include intelligent lift monitoring, vendor-agnostic smart-building visibility and integrated facilities management.

Softgrid also refers to existing experience in lift monitoring and Town Council estate operations. This makes the partnership relevant to organisations managing distributed assets, repeated maintenance activities and service conditions that can change throughout the day.

However, the announcement should be understood accurately. It establishes an exploratory strategic partnership, not a confirmed full-scale deployment, guaranteed project outcome or published record of savings. That distinction matters. A memorandum of understanding can indicate a direction for collaboration while the technical scope, data requirements, pilot design and performance results are still being developed.

Lesson one: begin with a clearly defined operational problem

Facility teams should avoid starting with a broad question such as, “How can we use AI?” A better starting point is a specific operational issue that can be observed and measured.

Examples might include repeated lift faults, incomplete equipment condition records, slow escalation of abnormal readings, difficulty coordinating multiple vendors or limited visibility across a multi-building portfolio. For a warehouse, the issue could involve monitoring critical material-handling equipment, refrigeration systems, loading areas or building services across operating shifts.

A well-defined problem helps the team identify the right data, users and success measures. It also prevents the project from becoming a dashboard exercise with no change to daily work.

Lesson two: connect predictive intelligence to engineering workflows

Predictive intelligence becomes useful only when someone can act on it. A system may detect an unusual vibration, temperature trend, operating pattern or visual condition, but the facility team still needs to decide what happens next.

A practical workflow could connect an alert to an asset record, work order, inspection task, escalation rule or vendor instruction. The responsible engineer or supervisor should be able to review the alert, consider operating context and determine whether the correct response is immediate intervention, planned inspection or continued observation.

This is the important engineering integration angle in the Softgrid–EM Engineering example. AI capability and engineering delivery are being considered together, rather than treating software as separate from maintenance practice. For SMEs, this can mean starting with existing communication and maintenance processes before introducing more complex automation.

Lesson three: assess data readiness before selecting technology

AI performance depends on the quality and relevance of the information available. Before beginning a pilot, teams should review:

• Which assets are included and whether each has a consistent asset ID.
• What sensor, building-management, inspection, video or maintenance data already exists.
• Whether timestamps, locations and operating conditions are recorded consistently.
• How often data is missing, duplicated or manually entered.
• Who owns the data and who is authorised to access it.
• How historical faults, repairs and maintenance outcomes are documented.

Edge IoT can help process information closer to the equipment or site, which may support faster monitoring and reduce dependence on constant cloud connectivity. It does not remove the need for sound asset records, sensible data governance and appropriate cybersecurity controls.

Lesson four: define asset coverage and the limits of the pilot

A pilot should be narrow enough to manage but meaningful enough to test the operating model. Teams could begin with one building, a defined group of lifts, a critical plant room or a selected warehouse zone. The right scope depends on the operational problem, asset criticality and available data.

Before deployment, document the assets included, the signals being monitored, the alert recipients, the response process and the period of evaluation. It is also useful to identify assets that are deliberately excluded. This prevents stakeholders from assuming that a pilot covers every system or that results automatically apply across an entire estate.

For warehouse operators, pilot boundaries should also consider shift patterns, vehicle movement, access restrictions and the effect of normal operational changes on sensor readings. A useful system must distinguish relevant abnormalities from ordinary changes in workload or environment.

Lesson five: keep human oversight in the operating model

AI should support professional judgement, not quietly replace it. The Ministry of Manpower has described SAGE as an AI-powered workplace-safety prototype that analyses workplace photographs, identifies common safety risks and recommends relevant guidance while complementing professional judgement. This offers a useful principle for facilities work as well: automated analysis can assist review, but accountable people remain important.

Facility teams should define who reviews alerts, who can close or override a recommendation, how uncertain results are escalated and how decisions are recorded. Training should cover both system operation and the limitations of the predictions.

Human review is especially important where equipment access, safety, service continuity or customer operations could be affected. An alert should be treated as decision support unless the system has been specifically validated for a more automated use case.

Lesson six: measure the pilot without assuming savings

Do not promise energy savings, reduced downtime or maintenance cost reductions before the relevant baseline and measurement method are agreed. Instead, define practical pilot indicators such as:

• Alert relevance and the number of false or duplicate alerts.
• Time from alert to review and from confirmed issue to work order.
• Completion of inspections or corrective actions.
• Availability and quality of asset data.
• Acceptance of the workflow by engineers, supervisors and vendors.
• Changes in unplanned incidents, where the measurement period and comparison method are appropriate.

These measures help teams decide whether to improve the data, adjust the workflow, expand the asset scope or stop the pilot. They also create a more credible basis for future budgeting.

How ISS can support the next step

For Singapore businesses, the main question is not whether AI is fashionable. It is whether a specific facilities challenge can be improved through better monitoring, clearer workflows and responsible automation.

ISS can help organisations discuss the engineering, facility-management and AI automation requirements involved in that assessment. A practical starting point may include mapping the current workflow, reviewing available data, identifying suitable assets, defining human approval points and preparing a pilot scope that matches the organisation’s operational capacity.

The Softgrid–EM Engineering partnership is a timely reminder that successful AI facilities management depends on collaboration between technology and engineering teams. SMEs do not need to begin with a large transformation. They can begin with a controlled problem, reliable information, clear accountability and measurable learning.

Contact ISS to discuss your engineering, facility management or AI automation requirements.

Sources and further reading

• Softgrid and EM Engineering Sign MoU – Predictive AI for Estate Management
• Opening Address at The Singapore WSH Conference 2026
• National AI Strategy Update