Land disputes are one of the most common legal conflicts in Kenya. Titles are lost, boundaries are disputed, and sales are contested. Resolution moves through courts for years.
The land registration problem is frequently described as a technology problem. It is not. It is a data structure problem.
The core issue: land records exist in multiple systems that do not share identifiers. A parcel appears in the Lands Registry under one reference number, in county GIS systems under a different boundary polygon, in property tax records under a different owner ID, and in court records under a case number that references none of the above.
No single query returns the complete picture of a parcel's status.
What AI can and cannot do
An AI agent connected to the Lands Registry API can retrieve a parcel record. It cannot determine whether that record conflicts with an adjacent parcel's surveyed boundary. It cannot determine whether a case filed at the Environment and Land Court two years ago affects the title. It cannot determine whether the property tax payments reflect the registered owner or a subsequent purchaser.
These limitations are not AI limitations. They are data integration limitations.
What the coordination layer addresses
ardhi-mcp (github.com/gabrielmahia/ardhi-mcp) provides structured access to land registry coordination tools for Kenya. It does not solve the data integration problem — that requires years of government data work. What it provides is a structured interface layer that can grow as data integration improves.
The east-africa-legal-historical-pd dataset at huggingface.co/datasets/gmahia/east-africa-legal-historical-pd provides historical customary land law records (pre-1928, public domain) that give AI agents context for the parallel customary land tenure systems that operate alongside formal registration.
The land registration problem is a data problem. The tools that address it must be built at the data layer, not the AI layer.
Top comments (0)