Certification-rich, interoperability-poor: to our knowledge, the first building-data interoperability assessment written specifically for Asia-Pacific.
The diagnosis, in 100 words
Your BMS speaks BACnet, but BACnet standardizes transport, not meaning. The points themselves are named in integrator shorthand — AHU-3-SAT, CHWP-2-RUN-STS — and no two buildings share a convention. Fewer than 5% of operational buildings carry any ontology tagging (practitioner estimate). The maintenance system, the property-management system, access control and the meters each live in their own silo. Every analytics platform you onboard re-pays the same mapping tax, because the tags are not stored in the BMS and do not survive upgrades. And the silos persist because someone, somewhere, bills by the hour to bridge them.
That is the diagnosis. Now the part nobody is writing about.
Every interoperability conversation is happening somewhere else
Read any building-data interoperability discussion from the past two years — the ontology debates, the "Plaid for buildings" pitches, the legal commentary — and notice the geography. The standards arguments are about North American office stock. The regulatory analysis is about Brussels. The funded platforms sell to New York and London.
We verified what that means in practice. Between April and June 2026 we checked the major data and agent platforms against the frameworks APAC portfolios actually report into:
| Platform | What we verified | APAC reality |
|---|---|---|
| Cambio ($18M Series A, Jan 2026) | Compliance suite covers GRESB, CRREM, TCFD, SFDR, EPCs | No NABERS, no Green Mark, no EEWH, no BEAM Plus, no CASBEE. Offices: NY, SF, London — zero APAC offices (verified May 2026) |
| Altus ARGUS Assist (launched Apr 2026) | Agentic layer over ARGUS valuation data; no API for the agent layer; zero BMS/IoT integration | Initial scope: US Industrial + Office. APAC clients hold the license, not localized capability (verified May 2026) |
| CoStar | No API — confirmed; Excel export only | No Taiwan, Singapore, Hong Kong or Japan coverage; Australia build-out only began in 2026 (verified Apr 2026) |
| VTS Asset Intelligence (launched Apr 2026) | AI lease abstraction, asset chat | Zero confirmed APAC deployment (verified Apr 2026) |
The pattern extends upstream. Yardi Virtuoso (September 2025) and Cherre Agent.STUDIO (October 2025) shipped agent marketplaces — inside their own walls. The major BMS clouds remain vendor-cloud onboarding paths with varying degrees of API openness. Mapped, the most credible independent data layer, raised its Series B in June 2025 and still deploys engineers building by building.
None of these are bad products. The point is narrower and harder: the buildings of Asia-Pacific are not in the room when interoperability is being defined. If your portfolio reports to NABERS in Sydney or Green Mark in Singapore, the global interop wave was not built with you in mind.
To our knowledge, this is the first interoperability assessment written specifically for APAC building data. That it took until 2026 is the finding.
What APAC actually has
Here is what the US/EU conversation misses: Asia-Pacific is not behind on building assessment. In one important respect, it was first.
Taiwan's Architecture and Building Research Institute (ABRI), under the Ministry of the Interior, introduced the Taiwan Intelligent Building Label in 2004 — making Taiwan the first country in the world to operate a national smart-building certification. Five tiers, Certified through Diamond, scored on ICT infrastructure, system integration, security automation, safety, and occupant comfort. That is two decades of institutional practice in asking "how intelligent is this building?" — and the label was refreshed again in the current 2026 cycle.
The rest of the region runs an equally serious assessment stack:
| Market | Scheme | Character |
|---|---|---|
| Taiwan | EEWH + Taiwan Intelligent Building Label | Dual track: green building + smart building (world's first, 2004) |
| Singapore | BCA Green Mark | Green-building scheme woven into national building policy |
| Japan | CASBEE | Built-environment efficiency assessment |
| Australia | NABERS | Measured, in-use performance rating — not design intent |
| Hong Kong | BEAM Plus | Green-building assessment scheme |
NABERS deserves particular respect from anyone who has run a measurement and verification program: it rates what the meters actually recorded, which is exactly the discipline IPMVP tries to enforce on retrofit claims.
But here is the regional condition in four words: certification-rich, interoperability-poor.
A certification assesses; it does not interconnect. A Green Mark plaque does not make the chiller plant's point names legible to your analytics stack. A five-tier Intelligent Building Label scores system integration within the building — it says nothing about whether the building's data can travel to the portfolio level. An M&V engineer running IPMVP Option C across a Singapore–Taipei–Sydney portfolio still spends the first weeks reconciling meter naming conventions before a single baseline regression gets fit. The certificates prove the building performs. They do not make the building legible.
And as far as we could verify, none of the region's schemes comes with a data-portability right. For that, you currently have to look at Europe.
The legal lever exists — it just isn't here
The EU Data Act has been fully applicable since 12 September 2025. Its user data-access rights (Articles 4–6) explicitly cover connected products including HVAC and building-automation equipment: the user of the equipment — not just the manufacturer — has a legal right to the data it generates. A second wave follows: "data by design" obligations for new products from September 2026, cloud-switching provisions from 2027. In parallel, the revised EPBD requires building automation and control systems in larger non-residential buildings (the >290kW tier was due end-2024; >70kW by 2029), plus digital building logbooks and Smart Readiness Indicator assessments arriving from 2027.
There is no APAC equivalent. No parliament in the region has written a building-data access right into law.
But two paths bring this to APAC anyway — today, not in some legislative future:
1. APAC owners with EU assets are already in scope. The Data Act's rights have been live for nine months. If your fund holds a Frankfurt or Amsterdam asset, you can exercise data-access rights on its connected building systems now. Most legal commentary addresses manufacturers' obligations; almost nobody is translating the owner-side opportunity — least of all for owners headquartered in Singapore, Tokyo, Taipei, or Sydney.
2. The EPBD wave reaches APAC through capital, not parliament. European investors are recalibrating what "transparent building data" means in disclosures, and APAC funds report into the same GRESB- and CRREM-shaped frameworks. The disclosure bar set in Brussels travels through investor reporting straight into APAC annual cycles — no local statute required.
The LLM turn — and what an APAC portfolio can do Monday
For twenty years the semantic gap survived every standards effort because closing it meant human work. That changed in the past eighteen months, and the evidence is now peer-reviewed: a 2025/2026 study of retrieval-augmented LLM point classification against the Brick ontology reported F1 scores in the 85–100% range across six real-world building datasets, beating static few-shot baselines by double digits.
The practitioner translation: roughly 75–85% of points become auto-taggable, with the remainder needing human review (our estimate; ambiguous shorthand and multi-role points concentrate the errors). Human-in-the-loop stays non-optional wherever fault detection rides on the tag — a wrong tag propagates a wrong diagnosis.
The money side, stated honestly. Industry estimates put manual point mapping at $10K–$75K per building, mid-range around $30K (roughly 1,200–3,000 interop-relevant points at $8–25 per point, all-in). We label these as estimates because there is no authoritative public per-point figure — a data gap that says a lot by itself. Two more honesty notes: at the low end, small simple buildings were never expensive to map, so the headline savings live in the mid-range and above. And the commonly cited 15–20-year average BMS age is exactly that — commonly cited, primary sourcing thin. The real economic argument is not one building's mapping bill. It is that the tax is paid again for every analytics platform you onboard, because the tags live in nobody's system of record. A verified, portable mapping corpus replaces a recurring tax with a one-time cost.
Meanwhile the agent-tooling substrate went industrial: the Model Context Protocol was donated to the Linux Foundation in December 2025 and now sees 97M+ monthly SDK downloads across 10,000+ servers. The built world's share of that: exactly one open-source BACnet MCP server (a five-star GitHub project), and — as of June 2026 — no Haystack or Brick server that we could find. The plumbing exists; buildings simply haven't shown up yet.
Here is the part that matters for an APAC portfolio today, and it requires no new hardware, no API contract, and no change to your OT security posture. Standard practice keeps the BMS on an isolated network, and your facility managers are right to defend that. But nearly every supervisory head-end already produces scheduled exports — CSV trend dumps, periodic reports, emailed summaries. For the large majority of commercial buildings (practitioner estimate: ~95%), those exports are the realistic 2026 access path. An advisory-only agent fleet that reads what the building already emits has zero write path to the OT network. Nothing to misconfigure, nothing to break.
Concretely: schedule the exports your BMS, meters, and CMMS already know how to produce. Land them somewhere an agent can read. Let the agents do the semantic normalization and the cross-system reasoning — energy against work orders against occupancy — and answer portfolio questions humans never had time to ask. Advisory only. We run an agent fleet on building data every day; this pattern is not a thought experiment.
"Agent-ready" — and what comes next
APAC's certification stack spent twenty years asking whether a building is green, and — first in the world — whether it is smart. The next question is different in kind: is the building legible to software? Can an agent read it? We think "agent-ready" will become a property owners screen for, the way energy performance became one. That framing deserves its own piece, and it will get one.
This report is the first in a series. Coming next:
- The State of Building Data Interoperability 2026 — the global companion report
- The Interop Report Card — vendor-by-vendor API openness, graded strictly from public documentation
- An open semantic-mapping benchmark — anchored in public real-world datasets, with the messy multi-contractor naming chaos real buildings actually contain, corpus published CC-BY (coming later this month)
If you own or operate buildings in APAC and want each of these as they land, subscribe to The Intelligent Building Brief, or write to hello@ai-smart-buildings.com — especially if your portfolio has a naming-convention horror story to contribute to the benchmark. Those stories are the dataset.
Research compiled by the AISB agent fleet from primary sources; every claim verified against the public record. Full source list available on request.
Research compiled by the AISB agent fleet from primary sources; every claim verified against the public record. Cost figures are labeled industry estimates. Full source list available on request — hello@ai-smart-buildings.com.