The 2026 AI in Real Estate survey published by Remit Consulting and the UK PropTech Association put a precise number on something the industry has been sensing for two years. Ninety-three percent of real estate organisations now provide their teams with access to AI services. Only seven percent say AI is fully integrated into their business operations.
That gap — between access and integration — is the defining technology challenge for institutional real estate in 2026. And the primary reason for it is not the AI tools themselves. It's the data they're supposed to work with.
What "fully integrated" actually requires
When AI tools in real estate work well, they do things like identify anomalous vacancy trends across a portfolio before they appear in quarterly reports, benchmark rent levels against comparable assets automatically, flag properties where capex forecasts are diverging from historical patterns, or generate first-draft commentary on portfolio performance from structured data.
All of these applications require the same thing: a data layer that is clean, validated, consistently structured and updated frequently enough to be useful. An AI model that predicts vacancy trends from data that arrives quarterly in inconsistent formats and is manually reconciled before use is not going to produce reliable results. It will produce outputs that look plausible but require expert review before anyone can act on them — which eliminates most of the efficiency gain.
This is the data readiness problem. It is not a problem that AI vendors are well placed to solve. It is a problem that must be solved at the data layer before AI tools are deployed.
Shadow AI and the data quality risk
The Remit survey also identified a phenomenon that deserves attention from compliance and risk officers: the rise of shadow AI. A significant proportion of real estate professionals are supplementing their firms' corporate AI tools with personal use of platforms such as ChatGPT, Gemini and similar services — often to process internal data.
For Swiss institutional fund managers, this is a governance issue with FINMA implications. Data from client portfolios — including property-level financial data, valuation information and tenant data — is being pasted into consumer AI services with unknown data retention and processing policies. Many organisations are unaware of the extent to which this is happening.
The structural response is to provide AI capabilities within controlled, compliant data environments rather than leaving teams to improvise with public tools. But this requires a controlled data environment to exist in the first place — which brings the discussion back to data infrastructure.
What the clean data layer looks like
The data layer that makes AI tools reliable in institutional real estate has four characteristics. It is consolidated — drawing from all relevant sources (property managers, accounting systems, valuers) into a single schema. It is validated — with systematic checks applied at ingestion rather than manual review at reporting time. It is structured — using consistent classifications and definitions that don't vary between data sources or reporting periods. And it is current — updated frequently enough that the data an AI tool processes reflects the actual state of the portfolio.
This description is not a description of what AI needs — it is a description of what good portfolio management infrastructure looks like independent of AI. The AI use case is an additional argument for building it, not a substitute for the underlying work.
Where Swiss fund managers are on this journey
Among Swiss institutional real estate fund managers, the data infrastructure picture is mixed. Most FINMA-regulated funds have some form of systematic data collection from property managers, but the quality and consistency of that collection varies significantly. Quarterly batch delivery is standard; real-time or near-real-time data flows are rare. Validation is typically manual. Classification mapping between régie data formats and fund-level schemas is usually undocumented, maintained by specific individuals.
This means that the majority of Swiss institutional real estate funds are not yet in a position to deploy AI tools reliably on their portfolio data — not because the AI tools aren't available, but because the data those tools would operate on is not sufficiently structured and validated.
The sequence that works
The consultancy and technology firms closest to this problem are converging on the same recommendation: address the data layer first, deploy AI second. This is not a conservative position — it is the position that produces reliable AI outputs rather than impressive-looking outputs that require extensive expert review to trust.
For Swiss fund managers, the practical question is not "which AI tool should we buy?" It is "does our current portfolio data infrastructure give us the validated, consistent, structured data that AI tools require?" For most teams, the honest answer is not yet.
The good news is that addressing the data infrastructure problem — consolidating sources, applying systematic validation, standardising classifications — delivers value immediately, before any AI tool is deployed. Quarterly reporting becomes faster and more reliable. Audit trails improve. The correction cycle that consumes analyst time each quarter shortens significantly. The AI capability is a second-order benefit that the infrastructure also enables.
Build the data foundation that makes AI trustworthy
STREETS provides the validated, structured, consolidated data layer that institutional real estate AI tools require. Built for Swiss fund managers, in production since 2017, hosted in Switzerland.
Book a demo