The Critical role of quality real estate data for AI

Nick Wadge, Chief Technology Officer at Knight Frank, explains why data should be the solid bedrock upon which AI infrastructure can be built.

AI’S EXPANDING FOOTPRINT IN PROPERTY MARKETS
For the past few years, the use of AI has been rapidly expanding across all sectors, including real estate. Whether it’s market forecasting, optimising portfolios or ensuring compliance with ESG targets, just about every tech company’s salesperson will proudly (and perhaps somewhat disingenuously) tell you how their AI software improves their offering. However, as enthusiasm for these AI applications grows, so do cautionary tales and scepticism. 

The effectiveness of any AI is only as good as the data feeding it. In real estate, we know that millions of pounds often hinge on decisions, and the quality and reliability of data inputs are fundamental to building trust in the insights that guide those decisions. In other words, if the data underlying an AI model is flawed, fragmented or biased, the model’s output will be unreliable – a dangerous proposition when investment decisions or development plans are on the line.

“GARBAGE IN, GARBAGE OUT” STILL RULES
The real estate industry has long grappled with inconsistent data. Property records might be spread across hundreds of local registries, each updated at different intervals and with different data standards. Leasing and sales data might exist on thousands of spreadsheets or in siloed systems. As a result, property data can be filled with errors – incorrect square meterage, omitted maintenance records, or outdated occupancy rates. Feeding these fragmented and frequently inaccurate datasets into an AI model creates a classic “garbage in, garbage out” scenario. If an AI algorithm trains on flawed inputs, the resulting insights will, at best, be incomplete and, at worst, deeply misleading.

THE DATA QUALITY IMPERATIVE
AI systems don’t possess any magic immunity to bad data – they amplify it. In practical terms, this means an AI model can’t properly learn market dynamics if the inputs are flawed. Any AI’s accuracy heavily depends on data quality, as biased or noisy datasets will inevitably lead to flawed predictions. Data are the ontological bedrock of AI. AI’s outputs (recommendations, classifications, decisions, etc.) are not standalone entities but emergent properties that arise from the structure and quality of the data they are trained on. Without well-formed, relevant and representative data, AI lacks a coherent substrate from which to form anything meaningful. Treating AI as a magic layer to be applied after the fact without attending to the underlying data is like expecting your house to stay standing without solid foundations. 

A DATA‑DRIVEN FUTURE
The real estate industry is on the cusp of an AI revolution, and standing still is going backwards. As we have seen, however, the true enabler of this revolution is high-quality data. When data are accurate, granular, and trustworthy, AI can unlock tremendous value, uncovering patterns we humans might miss and bringing new efficiency to an old industry. Conversely, AI will struggle or even backfire if data remains fragmented or inaccurate. It is critical to recognise this and act on it by investing in data quality; through standards, verification, and unified technology platforms. This paves the way for AI that consistently delivers reliable, unbiased insights to better service our clients.

The foundation of successful AI in real estate isn’t the algorithm; it’s the data.