Singapore GMIS-EB 2.0: How AI Energy Optimisation Can Support Existing-Building Retrofit Applications

For Singapore building owners and facility managers, an existing-building retrofit is not simply a matter of purchasing new equipment. The project must be planned around a clear baseline, measurable energy outcomes, suitable documentation and practical implementation controls.

This is where AI energy optimisation and connected building data can help. Used properly, AI can identify patterns in electricity consumption, highlight abnormal operation, prioritise retrofit opportunities and support ongoing verification. It does not replace a qualified professional, an energy audit, engineering judgement or the formal requirements of a funding application.

With the latest update to the Green Mark Incentive Scheme for Existing Buildings 2.0 page on 25 August 2026, building owners should review the current eligibility conditions, terms and application requirements before committing to a project. Applications submitted from 5 May 2026 are subject to the latest terms and conditions. Funding is available until funds are fully committed or 31 March 2027, whichever comes first.

What GMIS-EB 2.0 means for retrofit planning

GMIS-EB 2.0 supports outcome-based energy improvement works for eligible existing buildings. The scheme can include qualifying light-industrial properties of at least 5,000 sqm, subject to the applicable requirements. Support is linked to energy and carbon-reduction outcomes and Green Mark achievement, so a strong application needs more than a list of proposed equipment.

Depending on the project, application documentation may include a retrofit proposal, endorsement by a qualified professional, a project schedule and ACMV schematics. Owners should confirm the latest requirements directly with BCA, as eligibility, documentation and funding conditions may change.

AI is most useful when it helps the project team create a reliable evidence trail from the existing building condition through to post-retrofit performance.

Five practical ways AI and digital data can support an application

1. Establish a usable energy baseline

Before recommending a retrofit, the team needs to understand how the building currently consumes energy. Monthly utility bills may provide a starting point, but they are often not detailed enough to explain the performance of ACMV systems, tenant areas, warehouses, production spaces or common services.

Smart meters, submeters and building management system data can provide a more granular view. AI-assisted analytics can organise readings by time, equipment group or operating zone, then identify recurring patterns such as overnight consumption, weekend loads, demand spikes or abnormal operating hours.

The baseline should also record relevant operating conditions. These may include occupancy, operating schedules, weather exposure, production activity, loading patterns and major changes in building use. Without this context, an algorithm may identify correlation without proving the actual cause of energy use.

2. Detect faults and prioritise retrofit opportunities

In many buildings, the first opportunity is not a major equipment replacement. It may be incorrect schedules, simultaneous heating and cooling, excessive fresh-air operation, poorly tuned setpoints, blocked filters, sensor issues or equipment running when spaces are unoccupied.

AI can screen BMS and ACMV data for unusual behaviour and rank possible issues for investigation. For example, it may flag an air-handling unit that continues operating beyond the expected schedule or a cooling plant whose energy use has increased under similar conditions.

These findings should be treated as investigation leads, not final diagnoses. A facility engineer or qualified professional must verify the physical condition, control logic, safety implications and suitability of each proposed intervention.

3. Compare retrofit scenarios

Once the baseline and operational issues are understood, digital models and analytics can help compare different improvement options. These may include controls optimisation, variable-speed drives, chiller or air-side improvements, sensor upgrades, lighting controls or other energy improvement works relevant to the building.

A useful scenario comparison should show the assumptions behind each option. The project team should consider expected operating hours, equipment capacity, maintenance requirements, integration constraints, disruption to tenants or operations, and the interaction between systems.

AI can accelerate the comparison of scenarios, but it should not generate unsupported savings promises. A technically credible proposal distinguishes between measured historical performance, engineering calculations, modelled outcomes and targets that will require verification after implementation.

4. Improve retrofit documentation and coordination

Retrofit applications can require coordinated information from building owners, facility managers, consultants, contractors and system vendors. Missing schematics, inconsistent asset names or unclear responsibility for data can delay decision-making.

A structured digital data environment can help organise meter registers, equipment schedules, BMS points, ACMV schematics, inspection records, quotations, project milestones and commissioning documents. AI-assisted document search or workflow automation may help locate information and identify missing fields, provided that confidential data is managed appropriately.

BCA’s Smart Facilities Management guidance promotes a structured approach to planning and implementing Smart FM. In practice, this means starting with business and operational needs, assessing current systems, defining the desired outcomes, implementing suitable technology and reviewing performance. The technology should support the FM operating model rather than become a disconnected pilot.

5. Support measurement and verification after the works

A retrofit is not complete when equipment is installed. The project team needs to check whether the building is operating as intended and whether energy performance has improved under comparable conditions.

Connected meters and BMS trends can support ongoing monitoring. AI can help identify whether post-retrofit consumption follows the expected pattern, detect performance drift and create exception reports for the FM team. It may also help compare actual operating data with the baseline, subject to an agreed measurement and verification approach.

Measurement should account for changes in operating conditions. A warehouse with different throughput, a building with changed occupancy or a facility with altered operating hours may not be directly comparable with the original baseline. This is why the method, assumptions and data quality should be agreed before implementation.

Data readiness checklist for building owners

Before starting an AI-enabled retrofit study, review whether the building has:

• Reliable electricity bills and available interval data.
• Smart meters or submeters covering major energy-consuming systems.
• Accessible and time-synchronised BMS data.
• Current ACMV schematics, equipment schedules and asset information.
• Clear records of operating hours, occupancy and major process loads.
• Defined data ownership, access controls and cybersecurity responsibilities.
• A named person responsible for validating alerts and recommended actions.

For SMEs and smaller FM teams, the first step may be data consolidation rather than a complex AI platform. Evidence from Singapore’s SME AI adoption research also points to the importance of foundational digital systems and complementary technologies. AI is more useful when the underlying meters, workflows and records are dependable.

Know the limits of AI-generated recommendations

AI can process more data than a person can review manually, but it does not automatically understand the full operational context of a Singapore building. It may misinterpret a temporary shutdown, tenant activity, weather event, maintenance issue or process requirement.

Recommendations should therefore pass through a human review process. Facility teams should ask:

• What data produced this recommendation?
• Is the data complete, accurate and correctly labelled?
• What physical inspection supports the finding?
• Could the change affect comfort, indoor environmental conditions, safety or production?
• How will the expected result be measured?

AI tools should also be configured with appropriate access controls. BMS data, floor plans, equipment information and tenant-related records may be commercially sensitive.

A practical next step

Owners considering GMIS-EB 2.0 can begin with a digital and engineering readiness review. Map the current meters, BMS points, ACMV documentation and operating schedules. Establish a defensible baseline, identify the largest uncertainties, and prioritise a small number of retrofit opportunities that can be technically verified and documented.

AI energy optimisation is not a substitute for engineering design or compliance. It is a way to make building data more actionable, improve retrofit decision-making and support the continuous measurement needed for outcome-based improvement works.

For the latest GMIS-EB 2.0 eligibility and application conditions, refer to BCA’s official scheme page. To discuss engineering, facility management or AI automation requirements for your building, contact Intelligence Solution & Service Pte. Ltd. (ISS).