An Intelligence Layer for Geospatial & UK Property Data

‍

"We would highly recommend Passion Labs to any organisation looking for serious AI capability combined with pragmatic execution."

Thomas McAlpine, Founder, TYTL (prev. Addland)

‍

Business Context

‍

Our client operates one of the UK's most extensive property intelligence platforms, bringing together Land Registry information with tens of millions of records spanning ownership, planning, geospatial and market data.

The opportunity was not simply to make this information searchable. It was to create an intelligence layer capable of understanding the relationships between parcels of land, locations, ownership, planning constraints and commercial signals, and reasoning across them simultaneously.

‍

‍

‍

The Challenge

‍

Traditional property analysis requires experts to move between databases, maps, planning records and market information, manually joining together evidence before reaching a conclusion. This becomes particularly difficult with geospatial data. A question such as “Which sites within this area meet these development criteria?” is not a conventional search problem. It can require spatial reasoning across boundaries and polygons, combining multiple datasets, applying commercial constraints and validating the resulting opportunities. The challenge was therefore to give AI the tools and context required to reason reliably across an extremely large, heterogeneous property data environment.

‍

Our Process

‍

Passion Labs designed an intelligence architecture connecting frontier AI models with the client's proprietary property infrastructure. We developed a specialised Model Context Protocol (MCP) layer that exposed complex property and geospatial operations as tools that AI agents could reliably call. On top of this, we built a multi-agent reasoning system capable of decomposing property questions into a sequence of searches, spatial operations, analysis, validation and evaluation. Rather than attempting to place millions of records into an LLM's context window, the architecture allows models to intelligently navigate the underlying data infrastructure and retrieve, interrogate and combine the information required for each task.

‍

The Solution

‍

The resulting system acts as an intelligent reasoning layer across the client's property universe. Users can ask complex questions in natural language and the system determines which datasets and tools it needs, performs the relevant geospatial and database operations, combines evidence across multiple sources, tests its assumptions and returns a structured answer. This transforms previously fragmented land and property information into a system that can actively reason across geography, ownership, planning and commercial data. Instead of simply retrieving property records, the system can discover relationships and opportunities hidden across tens of millions of data points.

‍

‍

Impact & ROI

‍

• 90% reduction in time to surface viable opportunities

• Analyst workflows accelerated from days to seconds

• 3× increase in opportunities reviewed per week

• Significantly higher confidence in investment and development decisions

‍

This solution unlocked true needle-in-a-haystack discovery, delivering faster insights, better decisions, and measurable commercial advantage.

‍