Google DeepMind and the UK: AI to Boost Housing

A Major Challenge for the United Kingdom: Building 1.5 Million Homes

The United Kingdom has set an ambitious goal: to build 1.5 million new homes by 2029. However, this project faces significant administrative hurdles. Local planning authorities are often hindered by complex bureaucracy and accumulated delays. To address this situation, a collaboration with the UK government has been established to significantly reduce the time required to process planning applications. The aim is to halve the decision-making timeframe, allowing planners to focus more on the construction of new homes.

Co-Development of an AI Planning Tool

Following the success of a tool named Extract, developed in partnership with Gemini by the UK government's AI incubator, a new step has been taken. Google DeepMind, in collaboration with the UK government, Google Cloud, Faculty, and local authorities in Barnet, Dorset, and Camden, is working on creating an advanced prototype of AI-assisted planning. This prototype is designed to act as an expert assistant for planning agents, facilitating data extraction and case analysis. This cooperation aims to develop a tailored tool that meets the specific needs and daily challenges of planners.

Initial tests have been conducted in Barnet, Camden, and Dorset, and the government plans to make this tool available to all councils across the country by 2027.

Optimizing the Planning Application Process

In a typical planning application, agents spend long hours analyzing policy documents, historical archives, and PDF files. This laborious process creates a significant bottleneck, especially since housing applications account for about 70% of planning requests each year. By reducing the time spent on straightforward cases, such as loft conversions or extensions, the tested prototype could allow agents to focus on more complex and public-interest applications.

The AI planning tool aims to simplify routine tasks through AI by:

• Consolidating data: It preprocesses backlogs, highlights gaps, and extracts key information so planners can review everything on a single screen.

• Identifying local policies: It illuminates relevant national and local policies, pre-assesses compliance, and provides accurate citations for agent verification.

• Summarizing feedback: It reviews consultation letters to identify key objections or precedents.

• Drafting assessments: It creates a first draft of the final report, including justification and proposed conditions.

It is essential that the planning agent retains full control as the final decision-maker. Every line generated by the tool is reviewed, the reasoning is edited, and the agent maintains the authority to approve or reject an application. To ensure accountability, the prototype records its work at each step, creating a clear thought chain and a robust audit trail for every decision.

The Extension of Extract: From Legacy Documents to Actionable Data

The AI planning tool builds on the foundations laid by Extract, which has recently been made available to every council in England. Extract addresses a common problem: critical planning information is often locked in unstructured PDFs. Extract can transform hundreds of pages of legacy documents into actionable data in just minutes.

Extract has already proven its effectiveness in trials across more than 20 local planning authorities and is expected to save each council around 255 hours of manual work per year.

Towards a Universal Planning Tool

The goal is to create a more responsive and transparent planning tool that benefits everyone. By reducing the administrative burden on agents, it is possible to accelerate economic activity related to construction and help more people improve their homes.

As the project progresses, collaboration continues to ensure that these tools meet the evolving needs of the UK's planning ecosystem. The hope is that these developments will serve as a model for other governments around the world as they seek to modernize public service delivery.