Government department deploys Google Cloud’s generative AI across local government to automate council planning tasks.
Public sector administrations process huge amounts of unstructured data that slows down infrastructure development. The UK central government has set a target of building 1.5 million new homes by 2029. Local planning authorities are facing backlogs caused by dense paperwork, delaying development schedules.
To address these constraints, the Ministry of Housing, Communities and Local Government (MHCLG) and the Department of Science, Innovation and Technology (DSIT) have expanded two machine learning tools designed to accelerate municipal processes. Officials speaking at Google Cloud Summit London confirmed progress on the national rollout of the Extract application and the Augmented Planning Decisions (APD) prototype.
Lila Ibrahim, Chief AI Readiness Officer at Google DeepMind, said: “The UK has an opportunity to build the homes communities need, but local councils face a mountain of red tape. That’s why we’re co-creating sophisticated planning tools directly with councils to solve real-world bottlenecks.”
“This will significantly reduce decision-making time and enable planners to focus on the future of building Britain faster.”
Head of household applications (including routine domestic changes such as loft conversions and site extensions) account for almost 70% of all planning applications submitted each year. Manually evaluating these standard deliverables requires planners to spend hours cross-referencing local policy documents, historical archives, and unstructured PDF files.
This iterative evaluation process consumes administrative time to support key infrastructure and commercial developments. Automation deployments target this administrative dispersion and aim to reduce application decision-making timelines by 50%.
Core features of Google Cloud generative AI tools
Engineers at MHCLG and the government’s applied AI team, Incubator for AI (i.AI), built the Extract tool in-house using the Gemini foundation model. After trials in more than 20 local planning authorities, administrators have extended its application to all councils in England.
Extract parses unstructured data locked within legacy PDF records and transforms hundreds of pages of historical planning documents into structured digital datasets within minutes. Operational data from the trial phase shows that the tool will save approximately 255 hours of manual data entry per year per city council. This reduction will allow local authorities to reallocate personnel to complex assessment tasks.
Integrating large language models into public sector workflows requires an enterprise-grade security environment. Local governments handle sensitive citizen records and require strict risk management protocols to prevent data breaches.
The government hosted the Gemini model on Google Cloud to establish a protected operating environment where data sovereignty is maintained. Cloud environments have active security controls that block malicious input such as prompt injection attacks. This technology framework ensures that sensitive local government data remains secure during both test and production computing cycles.
APD systems, on the other hand, act as analytical assistants for urban planners by automating four key administrative tasks:
The system integrates incoming documents by preprocessing data backlogs, flagging missing information gaps, and extracting core geographic site data into a unified user interface for personnel review. The software identifies relevant national and local zoning laws, assesses compliance margins, and adds accurate policy citations for manual verification. The application analyzes public consultation letters and summarizes stakeholder objections and past case law. The model generates an initial draft of the final evaluation report, including technical rationale and recommendations. Approval conditions.
The protocol stipulates that human planners retain final decision-making authority on all applications. This software does not independently automate final approval or rejection. A staff member reviews every line of text produced by the machine learning model and corrects the analytical inferences before validating the report.
To maintain regulatory accountability, the APD prototype records internal processing steps sequentially. This mechanism establishes an auditable chain of thought and creates a validation trail of all applications processed to support the final decision of the personnel.
Planning trials and expansion schedules by local governments
Development of the APD prototype relies on a collaborative framework that brings together public sector administrators and engineering teams from Google Cloud, Google DeepMind, and faculty.
The alpha version is undergoing live testing within three local authorities: London Borough of Barnet, Dorset City Council and London Borough of Camden. Testing across these different local jurisdictions allows developers to utilize different local government datasets to test their software against different local policies.
The central planner has completed the alpha phase and plans to roll out the APD tool across all 300-plus UK local authorities by 2027. Google Cloud provides the elastic computing infrastructure needed to manage the thousands of concurrent inference queries generated during daily operations.
Paul Maltby, director of public services at the school, said: “The UK’s planning system is at a standstill. Planning officers are forced to spend half their time assessing attic conversion applications, while applications for housing estates and warehouses are put on hold.”
“Our AI system, built with planners, removes the tedium of reviewing simple planning applications and enables faster decision-making. This allows planners to focus on key developments that matter, and importantly, allows families to improve their homes without months of delay or uncertainty.”
Barnet Council’s Executive Director of Growth, Nysha Polaine, added: “The tool’s ability to gather relevant information, carry out preliminary assessments and create the basis for a report has the potential to save significant staff time managing planning applications and speed up the decision-making process for residents. In turn, this will make a significant contribution to achieving the borough’s housing growth targets.”
The collaboration between MHCLG, i.AI, Google DeepMind, and faculty creates a structured division of labor for enterprise software engineering. Ministries define policy guidelines and legal boundaries, and external technology partners design and deploy the underlying model architecture.
The successful integration of these systems demonstrates the feasibility of hosting advanced language models, handling core management workloads, and modernizing public service delivery within a secure public cloud infrastructure.
SEE ALSO: EU publishes AI content labeling handbook ahead of August deadline for AI law
Want to learn more about AI and big data from industry leaders? Check out the AI
AI News is brought to you by TechForge Media. Learn about other upcoming enterprise technology events and webinars.
