Skip to content

Politics & Governance

Governing in the Age of AI: Reimagining Local Government


Paper7th May 2025

Contributors: Isabel Atkinson, Nathan Boroda, Ruthie Gottesman


Chapter 1

Executive Summary

At the turn of the 20th century, local government was the engine of national renewal. As cities swelled under the weight of rapid industrialisation, councils built vast housing developments that moved hundreds of thousands of citizens out of slums, overhauled public-health infrastructure and laid the foundations of the modern welfare state.

A century later, while the challenges are just as urgent – an ageing population, a housing crisis and mounting pressures on public health – this engine is stuttering to a halt. By 2029, local government in England will need an additional £16 billion to deliver services compared with 2020. Many councils are risk averse when it comes to innovation – even as they are asked to do more for less. So, amid these pressures, they have reduced access to services through lengthy waiting times, dipped into dwindling reserves and redirected resources to meet demand for statutory services. And still the backlogs grow.

The limits of the existing operating model have been reached. Starved of resources by cuts inflicted by previous governments over the past 15 years, many councils are on the verge of bankruptcy even though local taxes are at their highest level. Residents wait too long for care, too long for planning applications and too long for benefits; many people never receive what they are entitled to. Public satisfaction with local services is sliding.

Today, however, there are new tools – enabled by artificial intelligence – that would allow councils to tackle these challenges. The day-to-day tasks of local government, whether related to the delivery of public services or planning for the local area, can all be performed faster, better and cheaper with the use of AI – a true transformation not unlike the one seen a century ago.

These tools would allow councils to overturn an operating model that is bureaucratic, labour-intensive and unresponsive to need. AI could release staff from repetitive tasks and relieve an overburdened and demotivated workforce. It could help citizens navigate the labyrinth of institutions, webpages and forms with greater ease and convenience. It could support councils to make better long-term decisions to drive economic growth, without which the resource pressure will only continue to build.

In our Governing in the Age of AI series of publications, the Tony Blair Institute for Global Change argues that AI offers a transformative pathway for reimagining how the state works, making governments more efficient, transparent and agile. Our first paper, Governing in the Age of AI: A New Model to Transform the State, set out the benefits of harnessing AI in the centre of government. Our subsequent paper, Governing in the Age of AI: Reimagining the UK Department for Work and Pensions, used bespoke analysis to estimate the time savings that embedding AI could deliver for the department. In this paper, we examine how AI could help reimagine local government.

To understand the potential, TBI partnered with one local government to map the tasks that were performed by its staff to our unique database of 19,000 tasks ranked according to the potential impact of AI. The analysis showed that using AI could automate or improve at least 26 per cent of tasks – or one million hours of work per year – which is equivalent to a productivity gain of £30 million per year.

If similar gains were achieved at national scale, we estimate an annual time saving worth £8 billion across local authorities in England and Wales, equivalent to 380 million hours and £325 per household. Councils could use these savings in several ways, including improving the speed and quality of services, reducing pressures on the workforce, bringing services back in house and moving towards a more preventative service model.

Councils – operating at a smaller scale than central government – are well placed to test innovative tools and services, becoming the nation’s innovation lab for public services, but this transformation will not happen without support. Increased centralisation of policy and operational decision-making over the past 40 years has coincided with the neglect of the operational and delivery role of local government. In addition, organisational siloes within and between local governments, as well as limited market power to challenge incumbent technology providers, increase the cost of change.

As a result, councils lack the confidence, capabilities and infrastructure required to unleash this innovation. While some councils – such as Newham, Hertfordshire, and Hammersmith and Fulham – are innovating with AI, mechanisms are not in place to scale these initiatives across the country.

The government has proposed creating unitary authorities – single entities responsible for all local-government services in their area, rationalising the complex system of overlapping local authorities that currently exists – and devolving more power to local areas. These are important steps in transforming local government and a critical opportunity to ensure it delivers for citizens.

To make the most of this moment of change, TBI is calling for the creation of a new institution: the Devolved AI Service (DAIS). DAIS would act as a co-operative platform for local government to support fast, frugal innovation. It would provide the vision and governance for local innovation, build and incubate local AI tools and services, transform the infrastructure for scaling innovation and improve local capabilities.

With seed funding from the Department for Science, Innovation and Technology and the Ministry of Housing, Communities and Local Government, DAIS would in the medium term transition to self-sufficiency, sustained through a subscription model. In the longer term, it would develop a funding stream by monetising international access to tools and services. The first local authorities and related organisations to join DAIS should become test beds, adopting three bold flagship programmes:

  1. Introduce AI co-workers in high-volume, high-cost services. AI could, for instance, support social workers with the assessment backlog, helping to reduce waiting times, allowing social workers to focus on higher-value caring instead of lower-value coordination, and saving money. If adopted, TBI estimates that the care-assessment backlog (currently 227,000 cases) could be cleared in one month.

  2. Pilot a Local Navigation Assistant (LNA) to support citizens’ interactions efficiently and effectively. The LNA would improve the provision of information about services, including about citizens’ eligibility and the progress of their applications, and provide pre-approval for some services to reduce avoidable contact, starting with benefits that rely on locally held data, such as housing benefits. Data from this tool could be used to enable service redesign and target interventions. TBI estimates that AI could save one-and-a-quarter days per week of additional time for staff in these citizen-facing roles.

  3. Introduce a data and decision-making platform for local-plan development and an AI Planning Assistant to streamline the planning process. These tools would create good-quality and dynamic local plans, streamline the application process for citizens and improve the quality of decision-making for applications. TBI estimates that AI could free up almost two days per week for housing and economic planners.

These new services must be both effective and safe. DAIS should make sure these tools are responsibly adopted, including through the use of TBI’s PEARS framework, as set out in A New Model to Transform the State, to ensure decisions made with support from AI systems are predictable, explainable, accountable, reversible and sensitive.

In addition to establishing DAIS, there are reforms that councillors can pursue immediately: championing AI innovation, adopting tools that have been proven to work, coordinating with leading local authorities and other stakeholders to share best practice, adopting data standards, and adapting the workforce and attracting the right talent.

Through these interventions, local government can regain its ability to transform communities and drive economic growth. Instead of firefighting immediate crises, councils can refocus on long-term priorities – building vibrant, resilient local economies and delivering meaningful improvements to citizens’ lives.


Chapter 2

The Need to Reimagine Local Government

Local government should be the cornerstone of modern public administration, given its prominence in people’s lives. There are 317 local authorities in England[_] and ten combined authorities (formal partnerships between local governments). These bodies are involved in more than 800 public services,[_] many of which are statutory responsibilities[_] such as caring for the elderly and supporting the homeless. They should also be an engine of economic growth through the planning process, and support business development.

Spotlight

What do local authorities do?

Local authorities have three key roles.

  • Commissioning services: identifying community needs, developing service specifications and procuring providers.

  • Service delivery: managing and operating citizen- and business-facing services such as waste collection, public transport and financial transfers, often as a delivery partner for national-level organisations such as the National Health Service.

  • Planning: managing land use for housing, infrastructure and commercial purposes, and approving planning applications.

Over the past 15 years, local authorities have been expected to do more with less. For instance, in 2013 councils gained responsibilities related to public health, yet between 2015 and 2024 the grant that supports this work was cut in real terms.[_] More challengingly, demand for statutory services is outpacing councils’ ability to pay for them. By 2029 local services in England will require almost £16 billion more in spending compared with 2020,[_] stretching resource-constrained local governments to breaking point.

Most local governments have adopted three tactics to deal with this increased demand.

First, many have increased waiting lists and indirectly rationed support. As of March 2024, more than 78,000 people in England had been waiting for at least six months for their council to carry out a care assessment.[_] In the 2022–23 financial year, 28,655 people aged 65 and over died waiting for care provision.[_] In the planning system, only 20 per cent of major and 38 per cent of minor planning applications were settled within statutory timeframes in 2023–24 – exacerbating housing shortages.[_]

Second, around 85 per cent of top-tier councils (which include county councils, London boroughs, metropolitan boroughs and unitary authorities) have increased council tax by the maximum amount allowed in the 2025–26 financial year (4.99 per cent, which is above inflation).[_] Even this is not enough: an increasing number of councils have used financial reserves to pay for in-year spending.[_] This approach has driven several councils to bankruptcy – seven since 2018.[_]

Third, many councils have diverted money from other programmes – especially those that invest in the local economy such as planning and development – to meet the demand for statutory services (Figure 1). Today, children’s and adult social care consume 64.8 per cent of the total budget for local government in England.[_]

Figure 1

Local governments in England are spending more money on just two services

Source: Institute for Fiscal Studies [_]

Note: Percentages are rounded

These efforts to raise and divert money have not been enough. Councils are increasingly unable to deliver for citizens or local communities. Backlogs are still growing. More than 1.6 million older people in England have an unmet care need today[_] and England has the highest homelessness rate in Europe.[_] Public satisfaction across all local services has been trending downwards over the past decade (see Figure 2).

Figure 2

Satisfaction with services is declining across all areas

Source: Local Government Association[_] [_]

Local government is running out of options and time. It cannot sustain rationing services, dipping into reserves or diverting resources indefinitely. In the face of persistent financial constraints, the operating model of local government itself must change.

The good news is that advances in AI capabilities can help reimagine how local government delivers without making these trade-offs. The current model is bureaucratic, labour-intensive and reliant on reactive decision-making. A new approach that uses AI to make interactions better, faster, cheaper – and that creates space to accomplish transformational change – is needed.

In the late 19th century, amid rapid urbanisation, local government built vast housing developments, moved hundreds of thousands out of slums and overhauled public-health infrastructure. Councils had the confidence, resources and freedom to drive major change. [_] Today, local government faces equally severe challenges – an ageing population, a housing shortage for young people and growing public-health pressures – but funding cannot match demand. Equipping local government with the capabilities and infrastructure to deliver AI-enabled transformation is essential.

The Potential Impact of AI in One Council

Understanding the operating model is key to improving it. In Governing in the Age of AI: A New Model to Transform the State, the Tony Blair Institute for Global Change and UK-based applied-AI company Faculty used a framework allocating daily tasks across government departments to three categories – citizen-engagement flows, operational flows and decision-making flows – to set out the benefits of harnessing AI in the public sector. Our report estimated that, based on the likely costs of implementation and the best available assessment of impact at the time, doing so could potentially unlock £40 billion in net annual productivity gains.

In our subsequent paper Governing in the Age of AI: Reimagining the UK Department for Work and Pensions, TBI used a bespoke database of 19,000 tasks ranked according to the impact of using AI for each of them to estimate time savings for the department. Our analysis found that time savings of around 40 per cent, or a productivity gain of close to £1 billion a year, could be achieved by harnessing this technology.

In this paper, we have used the same bespoke database to assess the potential productivity gains of implementing AI in a local-government setting (see Methodology for further details). TBI partnered with one council to analyse its operational data, subject to strict governance procedures.

Our analysis of the data showed that at least 26 per cent of tasks could be automated or improved using AI, delivering a productivity gain worth just under £30 million per year for this council. Altogether, one million hours of employed staff hours could be saved per year with the use of AI (Figure 3). The source data covered council employees only and did not include currently outsourced services, where further gains would be possible.

Figure 3

Embracing AI could significantly improve productivity in one council

Source: TBI analysis of local-government data[_]

TBI was also able to calculate the potential time savings across salary bands. As shown in Figure 4, AI is likely to have a similar impact across the first four pay bands (under £69,999), with between 25 and 31 per cent of tasks being improved. Given the concentration of roles in the middle pay brackets, the productivity gains in monetary terms are likely to be greater for these roles.

Figure 4

The impact of AI on time savings is similar across pay bands

Source: TBI analysis of local-government data[_]

TBI also calculated the total cost saving across England and Wales by extrapolating these figures nationally and adjusting based on mean pay for full-time local-government public servants in England and Wales. We found that implementing AI would provide councils with a total time saving of 380 million hours each year (full-time equivalent), or a productivity gain of £8 billion. This is the equivalent to £325 for every household in England and Wales annually.

These are estimated productivity gains that can be deployed by local governments in several ways, including improving the speed and quality of services, reducing pressures on the workforce, bringing services back in house and moving towards a more preventative service model. But these outcomes will not be achieved without significantly strengthening the foundations for innovation. We turn to this in the next section.


Chapter 3

The Case for a Devolved AI Service to Scale AI-Enabled Transformation

Councils – operating at a smaller scale than central government – are well placed to test innovative AI tools and services, but this transformation will not happen without support. Forty years of increased centralisation of policy and operational decision-making have been accompanied by the neglect of the operational and delivery role of local government, especially in England. Many councils do not have the resources, confidence, infrastructure or capabilities to invest in AI.

Some local governments buck this trend.

  • Newham Council has partnered with the University of East London to form the UK Centre for AI in the Public Sector, with the aim of developing low-cost tools that can support high-volume public services, starting with areas under extreme pressure such as temporary accommodation, housing and social care.[_]

  • Hertfordshire County Council has partnered with the University of Liverpool and robotics company Robotiz3d to trial an AI robot to identify and repair road defects before they become potholes.[_]

  • Hammersmith and Fulham Council is introducing AI-based analytics supporting its CCTV cameras and backend CCTV platform, which will allow monitoring teams to analyse objects in real-time and enable additional capabilities, such as the ability to identify fly tipping and anti-social behaviour. Automated video search will enable officers to find the right timeframe in a video – which can turn hours of manual searching into minutes.[_]

  • Hammersmith and Fulham Council has also piloted damp and mould sensors to help keep residents’ homes dry and safe.[_] A dynamic remote-monitoring system enables it to remotely assess the performance of assets including lifts, boilers and electrical appliances.

These are early-stage projects, the full impacts of which have not been evaluated. Most importantly, they represent only a small number of local authorities: 4 per cent of council leaders report using AI innovatively.[_]

Several factors play a role in successfully adopting AI tools quickly: strong local leadership to set direction and a culture supportive of innovation, sustainable financing, good-quality data to test, train and run AI tools, and capabilities to build or procure the right tools and evaluate outcomes. Many of these mirror the digital-transformation cycle. [_]

The digital-transformation cycle

  • Set the direction. Align on a focus across the organisation.

  • Make the case for people-centred change.

  • Allocate resources. Get the right people together to allocate resources.

  • Procure and build. Procure the technology needed.

  • Deliver. Integrate the transformation with business as usual.

  • Measure and learn. Share data and learn.

However, many councils lack these components. Local leaders might not yet have a strong understanding of, or confidence in, AI and are concerned about the perceived risks of AI to citizens (including marginalised groups) and the impact on public trust if things go wrong.[_] This risk is transferred to financial departments, whose strict return-on-investment (ROI) requirements over short timeframes for business-case proposals for AI projects[_] mean many never get off the ground or have the sustained support needed to test, iterate and improve over time. This situation is compounded by, and contributes to, a shortage of reliable evidence of the efficacy and safety of AI tools.

Councils also struggle to recruit and retain people with the skills to build or procure AI tools. Digital and data roles are 2 per cent of headcount in local government, compared with a 4 per cent benchmark across the public sector, and there is a higher proportion of non-technical roles within the digital and data workforce than in the private sector.[_] This increases reliance on outsourcing, which risks hollowing out internal capacity. In addition, only a quarter of council spending on digital, data and technology suppliers is channelled through government frameworks intended to simplify procurement of trusted services – such as G-Cloud, which offers easy access to pre-approved cloud providers – compared with 72 per cent in central government.[_]

Finally, local data are fragmented, opaque and of variable quality. Around seven in ten councils in one survey highlighted the availability of high-quality data as a barrier to innovation.[_] A large proportion of local-government workloads and data remain on site rather than in the cloud – which can lead to additional costs, limit data sharing and prevent services scaling.[_] Cloud infrastructure is an important foundation for scalable AI-based solutions.

In addition to these immediate problems, there are two persistent systemic barriers that make innovation almost impossible within the current institutional model.

Organisational siloes within and between councils increase the cost of innovation and waste potential for shared efficiencies. Councils mostly negotiate their own technology agreements.[_] Teams pitch, procure and/or build tools to solve similar problems, whether task-focused (for instance, automating data entry) or domain-focused (for instance, producing a care plan). This similarity could in theory enable a shared vision, tools, data and capabilities to be used to drive service transformation across administrative boundaries. Instead, each local area competes for AI capabilities, builds or procures duplicative products and collects data in different and mutually incompatible ways.

Councils also find it difficult to manoeuvre against concentrated market forces, especially for critical business-management tools which are dominated by a few providers in each service domain.[_] These incumbents, such as providers of case-management systems, guard access to their technology and data, and have little incentive to update their services. Requests for better services or data sets for a specific purpose are charged at a premium, with little guarantee of quality – but there are few alternative options for councils in the market. The high cost of changing providers – due to breaking long-term contracts, data migration and workflow changes – exacerbates the financial and political risk of moving to a system that would enable open innovation. Most councils use each system for a median of ten years.[_]

These factors lead to limited innovation, driving high costs, slow adoption and vendor lock-in. Councils spend £1.8 billion per year on technology systems,[_] yet between £300 million and £1 billion goes on maintaining legacy IT systems.[_] And where innovation is occurring, it is often in applications that improve the efficiency of a service, with a modest cost saving. Few councils have transformed entire services.

Progress towards improving the innovation environment is slow, patchy and mostly limited to sharing knowledge through organisations such as the Local Government Association (LGA). Other improvements include the introduction by the Ministry of Housing, Communities and Local Government (MHCLG) of the Local Digital Declaration, data standards such as Open Referral UK[_] and the Local Digital Fund[_] to enable pockets of innovation.[_] The London Office of Technology and Innovation (LOTI) has enabled a network of professionals and sandboxes for innovation for London councils. In parallel, the Greater London Authority is creating a Data for London Library to connect data across London.[_]

The last few months have seen a welcome burst of energy and activity in digital government in the UK. The Digital Centre of Government in the Department for Science, Innovation and Technology (DSIT) has brought together previously disjointed digital teams in central government. It has a set of ambitious goals, going beyond “keeping up” with the pace of technology to catalyse genuine transformation and work in partnership with the wider public sector, including local government. It has already begun to release useful tools to help councils digitise paper records to speed up planning decisions.[_] Experts within the Government Digital Service (GDS) are investigating how it can better support councils, with a focus on better procurement, data sharing, digital leadership and access to GDS products.[_] [_] But there is more that would need to be done in parallel to build sustainable local capacity and capabilities to create new tools, and create the right environment to scale existing ones, including practical support for infrastructure changes and the implementation of standards. Critically, these reforms must be owned by local government.

In the short term there may be a risk that innovation is disrupted while restructuring linked to the local-government reorganisation proposals occurs.[_] Yet the broader Devolution Priority Programme will expand combined authorities to all areas of the country, provide local areas with more freedom to act through better funding agreements and deepen the powers of mayors to help deliver economic growth.[_] It is also a rare opportunity not only to simplify structures but to genuinely reimagine the operating model of the newly combined authorities. To harness this potential and avoid paralysis, local government needs to be furnished with confidence, tools and infrastructure.

More can be done now. The speed and scale of the challenge requires the establishment of a new institution with the funding and mandate to act quickly to overcome the obstacles outlined. The government should introduce a Devolved AI Service (DAIS), initially funded by central government but owned by councils. DAIS would act as an ecosystem orchestrator, taking on five key roles:

  • Create a shared and locally-owned vision and governance framework for AI adoption. Building on the Local Digital Declaration, DAIS should align councils with a vision for the use of AI technologies, as well as with guidance for adequate data and product governance. This will help reduce risk aversion within local government.

  • Pilot products and services. Where the market for innovation is nascent or the costs of entry into the market too high, DAIS may identify use cases and build reusable tools and services that solve local problems – providing a base level of functionality that could then be customised locally or built on using private-sector engagement. It could also introduce competition funds, modelled on the “test-and-learn” approach used by the UK government’s Innovation Fund.[_] Ongoing costs such as compute resource should be absorbed as far as possible by DAIS.

  • Curate the right environment to scale innovation – including getting the basic technology right. Where there is a flourishing market for innovation but conditions limit the spread of this innovation (for instance, in monopolistic markets), DAIS should perform a curator role. It should work with councils to signal to the market what they will procure, help to evaluate the safety and efficacy of locally developed tools, introduce framework agreements to facilitate procurement, and support councils with data cleaning or cloud migration. DAIS should work with MHCLG – and especially its Local Digital team – to introduce common standards[_] (including on data collection and operational processes, as well as cyber, privacy and accessibility protocols) which can help prise open poorly functioning monopolistic markets. DAIS would provide the practical support for local governments to implement these standards, and build and manage open application programming interfaces (APIs).

  • Enhance AI and procurement capabilities. The core operational team should comprise staff from multi-disciplinary backgrounds (both subject-matter experts and technical-support personnel) sourced from the labour market as well as through secondments from local governments. Councils which second staff should be the first to trial the new components or services. In this way, labour is traded for products, while capabilities are developed. DAIS should work with GDS on its existing programme to support procurement capabilities in local government.[_] DAIS should also help to upskill the local-government workforce by helping to scale talent-development programmes, such as in Newham.[_]

  • Monetise international access to tools and expertise to help de-risk innovation and de-link innovation from individual business cases. While productivity gains through the adoption of AI may encourage further reform, there is an opportunity for DAIS to enhance the financial sustainability of innovation (and reduce cost pressures faced by councils) by exporting expertise and products to localities around the world. The recent growth of international groups focused on sub-national digital transformation, such as the Open Government Partnership Local,[_] highlight the potential. DAIS should create a commercial arm, modelled on the success of e-Estonia, and all funds should be reinvested into DAIS, helping to financially de-risk innovation for the most ambitious projects.

DAIS should be owned by councils (which could mirror the set-up of PD in Germany,[_] the in-house public-sector consultancy which is owned exclusively by public-sector shareholders and delivers 80 per cent of services to them), with a rotating steering group from principal councils and combined authorities[_] to set priorities and direction, helping to achieve local buy-in.

While seed funding for DAIS should come from DSIT and MHCLG, over time revenues from international access to tools would displace this funding. Council subscriptions – set at a reasonable price given funding constraints – should also be introduced to ensure buy-in. If councils choose to leave DAIS, they would lose access to its continuous support, providing strong incentives to remain.

The body could initially be incubated within GDS to align goals and build capabilities within the organisation. While not directly controlled by GDS, it should be seen as closely linked, sharing resources, expertise and tools.

Over time, DAIS should position itself as a single collaboration point with central government and support the practical adoption of regulation (for instance, by creating tools to help databases conform to standards). Conversely, it could use learnings from its work to amplify local voices for central reform and negotiate data-sharing agreements with central departments that would allow councils to plan and deliver services and reduce duplication. As a voice for local innovation, it may also support DSIT to vet new AI vendors or tools from startups and small and medium enterprises (SMEs) using sandboxes or accelerator challenges.

While DAIS would unlock several barriers to local innovation, its success would rely on local authorities continuing best practice related to service design, including diagnosing problems, co-designing changes with end users and identifying data needs. Later in this paper we outline several steps that councillors can take now to champion local innovation.

Finally, DAIS will not succeed without the right private-sector partners; as described in our paper A New Model to Transform the State, these often have significant capabilities including in cyber-security, large language models (LLMs) and compute. The current relationship between councils and the private sector is limited by internal constraints and data-access issues, or extractive because of vendor lock-in. DAIS would form a central entry point for the private sector to work with local government to share capabilities and capacity, inform standards and build common tools.

Spotlight

What would DAIS do?

DAIS would overhaul the incentives that block fast-paced innovation and reduce the number of failure points in the digital-transformation lifecycle by:

  • Being owned by councils, for councils.

  • Providing shared leadership on AI innovation.

  • De-linking funding from individual business cases and strict ROI.

  • Having “protected” capacity and capabilities to develop AI tools – and helping to upgrade local government’s capabilities.

  • Providing practical support for councils to implement data standards, make infrastructure changes and evaluate new tools.

  • Reducing duplication of effort and money by removing innovation siloes between and within local governments.

  • Using its collective power to bring incumbent technology providers to the table and actively work with them to improve their products.

  • Providing a single point of collaboration between local and central government to co-innovate and provide feedback on regulation.

The impact of DAIS would be felt across different time horizons. In the short term, DAIS would help councils save money and improve services by rapidly testing innovations. As DAIS works to implement standards across the tech stack and builds council capabilities, this innovation could be scaled rapidly across locations. The potential to export products globally would provide a sustainable funding stream for continuous innovation.

To highlight what could be achieved, we turn next to three specific ways that DAIS could help local governments harness AI to transform services.

1. Introduce an AI Co-Worker in High-Volume, High-Cost Services

Local government is stuck in a service doom loop. Access to and the quality of services are being rationed. While money and time are focused on a small number of high-volume, high-cost services such as adult social care, even these struggle to provide a good service and waiting lists have ballooned.

Tackling these waiting lists has involved transferring workforce capacity to admissions, data entry and administration, and away from service delivery (Figure 5). A strategic approach would be to eliminate these administrative burdens, creating more time for service delivery.

Figure 5

Councils are increasingly spending time on assessment and administration instead of service delivery

Source: Future of local government, PwC[_]

One example is care assessments in adult social care, which are used to determine whether a person needs care and what this care should look like. This process involves a social worker or occupational therapist, who conducts a meeting by phone, in person or online to discuss the care applicant’s ability to perform day-to-day tasks.[_] Backlogs for assessments had reached 227,000 by March 2024; more than 78,000 of these applicants had waited more than six months.[_] There are further backlogs when it comes to the creation of care plans. As a result of these delays, many citizens end up in hospital as their conditions deteriorate or after an accident such as a fall – and many experience delays being discharged from hospital while they wait for care to be arranged.[_]

Under this operating model, to tackle growing demand, local government must either increase the size of its workforce or the caseload per worker. The first approach takes significant time – the social-worker vacancy rate stood at 9.3 per cent in Scotland in June 2024[_] and 10.5 per cent in England in September 2023.[_] While both these figures have come down slightly compared to the previous year, the use of agency social workers in England increased by 25 per cent between September 2022 and 2023, which is costly.[_] The previous central government provided more funding to employ and retain social workers to help clear backlogs,[_] but these were short-term measures that have only helped at the margins while neglecting the need for broader transformation.

The second approach – increasing the caseload per worker – is unsustainable. Social workers are increasingly overworked, including with administrative tasks. The turnover rate for social workers stood at 14.5 per cent in adult social care in 2023,[_] with common causes being burnout, overly bureaucratic systems and a culture of blame.[_]

The growing number of errors in the care-assessment process and poor quality plans are linked to these pressures. The Local Government and Social Care Ombudsman received 40 per cent more complaints about assessments and care plans in 2023–24 compared to the previous year,[_] and upheld 80 per cent of those it investigated, up from 67 per cent in 2022–23.

This does not need to be the case. The tasks involved in these types of high-volume, high-cost services can all be done better and faster using AI, as highlighted by leading local governments and partners in the private sector.

Spotlight

How are organisations innovating?

Private companies are using AI to support the key tasks in social work:

  • Several local governments, such as Kingston Council, are trialling Beam’s AI tool, Magic Notes. This is a generative-AI tool that streamlines the documentation process for social workers online, via phone or in person, and stores information in a secure, centralised database. It has 98 per cent accuracy on transcription and uses a human-in-the-loop approach to ensure accuracy. This tool has been shown to reduce the time spent on administrative tasks by 63 per cent.[_] [_]

  • Cera has created a generative-AI tool to transcribe social workers’ conversations and convert them into outcome-based care plans with tasks for carers. The company estimates the tool could save staff two to three hours per day.[_]

  • NextGen’s Ambient Assist creates AI-generated notes and care plans for patients in the health system, integrated with electronic health records. According to the company, it can save up to two hours of time spent on documentation per day.[_]

The integration of these tools in high-volume, high-cost services could be transformative. Rather than replacing workers, these tools would allow social workers to focus on the high-value aspects of their role.

The impact of AI on high-cost, high-volume services

According to TBI’s analysis, harnessing AI in the administration of high-cost, high-volume services related to the delivery of children’s and adult social care could create a time saving of around 12 per cent, representing half a day per week for staff working in this space.[_]

Focusing more narrowly on care assessments conducted by social workers, using AI could result in a 50 per cent reduction in the amount of time spent on administrative tasks – a conservative estimate in relation to real-world evidence from companies such as Beam. This reduction could, in turn, create 268,250 hours of time savings per week for social workers in England. These savings could be directed towards improving the quality of care interactions, which in turn could improve both staff retention and outcomes for people receiving care. Alternatively, they could be used to eradicate the assessment backlog within the first month.[_] [_] [_]

However, more can be done. If councils procured different tools across the assessment, care-plan and review process, this would lead to unhelpful friction for social workers. Instead, DAIS should seek to integrate different aspects of the social-work process with the same tool – indeed, some of the companies referenced earlier have started to do this.

The introduction of this kind of AI co-worker tool will face systemic barriers. First, conversations with stakeholders for this paper highlighted the fact that information-governance teams within local government can stall or block innovation over fears of litigation if something goes wrong, especially in services supporting vulnerable people.

Second, a small number of companies dominate the space for the case-management systems (CMSs) that hold case notes, assessments, care plans and other relevant personal information, making the effort and cost of changing providers greater than potential benefits of going elsewhere. Many of the commercial contracts with these companies do not offer flexibility or provide access to data (which are often held in different formats, making data migration costly). As a result, these companies block integration with their systems, often on the pretext of safety, and charge a premium to local governments to access their data, limiting the spread of innovation and the use of data in decision-making.

New market entrants must also go to each council to sell their products, with each local-government body requiring slightly different interfaces or data-collection processes. For instance, care assessments are conducted in different ways across local authorities, requiring time-consuming adaptations to tools to meet these requirements.

Rapidly adjusting these market barriers, and reforming local government’s procurement processes and its role in curating the innovation environment, is key. DAIS would be best placed to overcome these barriers and scale these co-workers. This involves several steps:

  • DAIS should set a goal for every council to adopt an end-to-end AI co-worker to support public servants in high-volume, high-cost services, such as adult social care. It should identify local authorities that are willing and capable of rapidly piloting these tools.

  • DAIS should prime the market by running innovation competitions to identify companies to build these tools. In each competition, DAIS should outline modular components for the AI co-worker based on the tasks involved. In social work, this would include note taking and summarisation, assessing needs, recommending action, generating care plans and monitoring changes in needs. DAIS could also work with the government’s Incubator for Artificial Intelligence (i.AI) to incorporate existing products, such as the minute-taking tool,[_] into a foundational version of the co-worker tool, which can then be customised and adapted through private-sector involvement.

  • Depending on the maturity of the products, and to encourage SMEs to participate, DAIS could invite companies to bid as a consortium, using existing tools as components to build out end-to-end products.

  • DAIS should create critical data sets to support training of tools and policy decision-making. DAIS should review and publish findings on the variations in the way data are collected and stored across CMSs and recommend quick improvements in data infrastructure. Following the example of Open Banking, it should require “read-only” data-sharing from existing CMSs[_] and set a timetable for the staged release of pseudonymised aggregate data sets (replacing identifiable information with codes, allowing data across systems to be linked and decisions on care made without disclosing individual identities) to be provided to decision-makers. DAIS should rapidly prototype AI tools to restructure or reformat historical data held in these systems – or adapt similar tools trialled by i.AI – which would save councils time and money.

  • DAIS should also work towards establishing standards on key processes and data collection for the AI co-worker’s different tasks to underpin the scaling of innovations and provide consistent data sets – with continuous feedback loops to ensure that standards remain up to date. For instance, in social work there should be a standardised approach to conducting care assessments, writing care plans and data-collection requirements. These standards would allow businesses to rapidly scale AI tools and expand the number and type of people who can perform key functions.[_] [_]

  • Over time, DAIS should build and manage open APIs to integrate the functionality of AI co-workers with CMSs, allowing these tools to edit data within existing CMSs. DAIS can introduce these open APIs by cooperating with CMS providers and compensating them for their work to integrate the APIs, working with MHCLG to update national legislation to force CMS providers to comply with integrating open APIs and/or by introducing an innovation fund to create competing CMS products once contracts have ended.

  • Standardising how data are structured and stored within CMSs - and enabling trusted AI tools to safely access and update that data - would make it easier for councils to switch between CMS providers, reducing their dependence on a small number of dominant vendors by lowering the technical and financial barriers to migration.

  • As data access from CMSs improves, DAIS should link these data with other valuable data sets which can then be analysed by AI tools and used to support operational and policy decision-making. For instance, in adult social care, this would require CMS data being linked to data from the digital social-care records, which include day-to-day data on a person’s care, and national databases, including operational data on the workforce. Better data standards will also facilitate integration with the health-record systems in the NHS.

  • DAIS should create AI tools to support councils to make better decisions on this linked data, including tools that can predict care needs, understand required staffing levels and notify workers of actions across the care workforce, such as when a person has just left hospital and requires an updated care plan. All of this supports councils to plan in their area, respond quickly and intervene early.

  • As AI reduces the administrative burden in high-demand services such as social care, workflows must be carefully redesigned to ensure adequate breaks from potentially higher-value but higher-stress work.

  • Public servants involved in purchasing and outsourcing the provision of AI-based technologies should be supported with resources to ensure AI-enabled services are compliant with data and equality laws, like those produced by the LGA.[_]

High-volume, high-cost services can be transformed using AI. In adult social care, for instance, there are too few social workers to meet the growing care needs of an ageing population. Many of these workers are overworked and leave soon after joining. The person receiving care gets a low-quality service: assessments are delayed and inaccurate, and care plans can be ineffective.[_]

Introducing an AI co-worker for social workers would enable three things. First, it would lead to better and quicker assessments and care plans, which would improve the care recipient’s satisfaction, help prevent the escalation of their needs, and would allow social workers to focus on caring work rather than administration and coordination. Second, it would facilitate a dynamic approach to changes in people’s care needs and provide a mechanism to rapidly update care plans when needed. Third, the use of AI to analyse data from care assessments and plans could help to predict people at risk of needing care before they reach a crisis point or to understand staffing requirements, helping to shift to a preventative model of care that reduces costs for local governments.

2. Pilot a Local Navigation Assistant

Tackling high-volume, high-cost services would help create the capacity and space to address wider access issues across local services. For most citizens, councils are the first point of contact to register a business, book a venue, tell the council about a pothole, check the status of an application, apply for a variety of benefits or pay parking fines. These interactions fall into six broad categories: register, book, tell, check, apply and pay.[_]

However, the fact that issues are not resolved in a timely manner leaves people feeling abandoned. According to KPMG’s analysis in 2022, local government lags other sectors in the net promoter score (NPS) – a measure of citizens’ satisfaction – standing at -25 points, compared to +5 points for the public sector and +29 points for private-sector leaders.[_] When councils fail to address citizens’ concerns and issues quickly and effectively, the NPS drops from +1 to -59. [_]

Applying for services is one of the biggest issues for residents. In 2024, just half of residents felt their council kept them informed about the services and benefits it provides, down from 66 per cent in 2012.[_] When councils do keep people informed, information flows favour quantity over quality of information, such as mass distribution in newsletters rather than information tailored to the individual or area. As a result, engagement is low and many people do not get to the right services the first time.

In addition, many locally administered benefits have duplicate information-collection processes and extensive requirements for citizens to navigate. For instance, the Housing Benefit and Council Tax Reduction form is 32 pages[_] and requires people to fill in information that is likely already known to the local government from previous interactions or elsewhere in government (for instance, via interactions with the Department for Work and Pensions (DWP)). These complicated forms and the complexity of the information required can lead to prolonged interactions and errors, delaying the process further and leaving vulnerable citizens without critical support.

Delays then occur in the processing of applications. More than three-quarters of new housing-benefit claims (76 per cent) took longer than the two-week statutory period to process in FY 2023–24.[_] Applications undergo manual assessment and evaluation against set criteria, which is resource- and time-intensive. People have no visibility on the progress of their application, which results in them contacting call centres – which in turn reduces capacity and diverts resources away from complex queries.

Siloes between councils mean that when citizens move to another local-authority area, they will likely have to repeat application processes – for instance, to apply for housing benefit or register for council tax. (There are some exceptions; processes are in place to transfer information held within Education, Health and Care Plans for children with special educational needs and disabilities across local government, for example).[_] More than 2.84 million people who moved local authorities in the year before the 2021 census[_] would have been left to navigate new points of contact, information sources and application processes by themselves.

The cumulative effects of poor access to information and laborious processes have led to a situation in which a significant proportion of locally administered benefits are unclaimed.[_] Around £3.4 billion in council tax support is untapped annually, affecting more than 2.25 million households, while unclaimed housing benefits reached a total of £1.3 billion in 2024.[_]

Figure 6

Significant support went unclaimed across locally administered benefits in 2024/2025

Source: Policy in Practice[_]

These service-access challenges could be addressed by developing better ways of navigating citizens successfully to the services they need. Examples in the health space show what is possible: as described in TBI’s paper Preparing the NHS for the AI Era: Why Smarter Triage and Navigation Mean Better Health Care, Rapid Health, which is trialling AI triage and patient navigation in NHS GP practices, has shown in a real-world study that the use of these tools can reduce repeat appointments by 70 per cent and demand for urgent same-day requests from 62 per cent to 19 per cent.

Some local governments have begun exploring AI solutions to improve navigation of citizens. For instance, Salford City Council has developed a Digital Inclusion Triage Tool[_] to assess users’ needs and signpost them to recommended services. Hillingdon Council has deployed automated voice and web-chat solutions, which handle around 40 per cent of all incoming calls and can be accessed by citizens at any time of day. To date this has saved the council £5 for every pound spent.[_] But more can be done. DAIS should introduce a Local Navigation Assistant (LNA) to streamline access to services across the six key functions of local government. Like the digital public assistant outlined in TBI’s paper Governing in the Age of AI: A New Model to Transform the State, the LNA should provide a single point of contact between citizens, businesses and their council.

Underpinning this LNA would be good-quality data. There are four main types of data that are required to perform each of the six key functions of local government. These are:

  • Request data: information provided by citizens about why they are getting in touch.

  • Personal data: includes prior interactions, background data and identity data.

  • Public data sets: geospatial and demographic data, for example.

  • Rules data: includes lists of available services and eligibility criteria.

However, these types of data are inaccessible or unusable in their current form. Personal data are split across departments within or outside the local authority. For instance, data on adaptations for housing are not linked to social-care records despite similar citizens using both services[_] and data on disability assessments are held in proprietary data systems in the DWP.

There has been some progress towards making better use of local data in some areas of the country. The London Borough of Barking and Dagenham has created One View,[_] which brings together personal and publicly available household data from across local-government services (including adult services, children’s services and homelessness) into a single view. Advanced analytics flag high-risk cases early and enable proactive intervention months before crises. The tool has enabled proactive support of 1,000 households with debt and predicted 93.6 per cent of people who needed to shield during the Covid-19 pandemic before the national government released its list.[_]

Finally, data standards such as Open Referral UK have codified the way that community services are categorised, creating common rules for any directory,[_] including complex eligibility criteria for services.[_] These standards are a critical foundation for the functioning of AI as a navigation assistant. Yet they have been adopted by only eight local authorities.[_]

The impact of AI on citizen-facing services

According to TBI’s analysis, using AI in citizen-facing services could lead to a 27 per cent time saving, representing one-and-a-quarter days per week of additional time for staff who are involved in local services with the most interaction with citizens, including in call centres and benefits administration.[_] [_]

More can be achieved. Resolving issues related to siloed, poor-quality and inaccessible data, and harnessing AI present an opportunity to break free from the confusing and bureaucratic processes currently required to navigate services. Linked data and the application of AI solutions can personalise information, identify services and support citizens in their interactions, including by pre-filling information on forms and pre-approving eligibility checks. For public servants, AI can help the workforce prioritise urgent and complex cases, optimising workflows. Several steps should be taken to deliver a Local Navigation Assistant:

  • DAIS should create a framework, co-designed with councils, outlining the key modular components of the LNA. This should map onto the six key functions of local government: register, book, tell, check, apply and pay. The LNA should be open source (modelled on LocalGov Drupal, which provides common code for websites) and interoperable with the digital public assistant outlined in TBI’s earlier paper.

  • DAIS should work with the private sector to procure key parts of the lower tech stack (for instance, compute power and commercial LLMs that can be fine-tuned).

  • In the short term, DAIS should rapidly build (and test in a handful of areas) a foundational prototype of the LNA that acts as an information support system for citizens, signposting them to the right services and ingesting information on citizens’ requests and needs at first contact. As its architect, DAIS would have full oversight of and sovereignty over the LNA. This would avoid ongoing licensing fees and the possibility of vendor lock in. This would also allow for interoperability across all local councils.

  • The tool should be trained on information collected via websites using LocalGov Drupal open-source code, which will include data on available community services. This prototype would also collect request data at the front door. DAIS should also work towards augmenting and automating the six key functions around which the LNA is designed by creating standards for the key types of data required. For instance, DAIS should codify rules data on eligibility for transactional benefits, which would enable the pre-approval of services. To ensure quick piloting, DAIS should build out this pre-approval function first by focusing on services that require only local data or publicly available national data.

  • DAIS should also work with MHCLG, the Department for Business and Trade and the Department for Environment, Food and Rural Affairs to integrate public data sets on population, geospatial information, environmental information, public health and community safety to enhance information around local services, following a once-only principle of data sharing. As TBI set out in its paper Governing in the Age of AI: Building Britain’s National Data Library, this can be done via the National Data Library.

  • DAIS should also work with GDS to incorporate central-government information into the LNA, so users can reach the services they need regardless of their starting point.

  • DAIS should support councils to create a single version of each person’s personal data to reduce the reliance on asking customers for information. Underpinning this should be a digital ID for citizens. This would require DAIS to support CMSs across the council to adopt the digital ID and work with MHCLG on legislative changes to give councils the power to enforce adoption. As part of this, DAIS should investigate how local-government systems can be integrated with the OneLogin identity-verification system that the government is rolling out elsewhere.

  • As CMSs adopt a single digital ID for interactions with citizens, DAIS should fund a research stream to assess the feasibility of introducing a single CMS. This single CMS would enable a view of citizens across local services, reducing fragmentation and cost. The single CMS could then be integrated with additional specialist modules (for instance, a care-needs assessment module), which can be decided in the procurement process.

  • DAIS should enable visibility of the individual’s data by both enabling a “citizen’s view” of an individual’s own data through the LNA and contributing to the emergent National Data Library by feeding in locally collected, high-value, suitably anonymised data.

  • DAIS should build out the functionality of the LNA, releasing more modular components as data become available. As the LNA conducts more types of interaction that require personal data, DAIS should adopt the “earned-autonomy” framework, outlined in TBI’s paper Governing in the Age of AI: New Model to Transform the State, as guidance for local authorities adopting the LNA. DAIS should adopt TBI’s PEARS framework (set out in the same paper, which says that tools should be predictable, explainable, accountable, reversible and sensitive) to ensure that the LNA fosters public trust and secures broad support.

  • DAIS should set an expectation that all local governments use the LNA, implemented across all six key functions, to navigate citizens by 2030. DAIS should build and manage open APIs to allow local governments to procure or build additional capabilities if desired, with the potential for this to be rolled out as an integral part of the LNA. DAIS should monitor progress towards a fully realised LNA in each area.

  • DAIS should implement robust security safeguards and robust information-governance practices, including compliance with the Data Protection Act (2018) and the Public Sector Equality Duty.

  • In cases where the navigation function of local-government services is contracted out to private call centres, the introduction of the LNA may require bringing this function back in house – especially if the LNA is deemed to reduce the need for call handlers. DAIS should work with councils to commit to a timetable to sunset these contracts.

DAIS should monitor the effectiveness of the LNA through agreed key performance indicators (KPIs). KPIs should include participation rates, accessibility, user satisfaction, service-delivery times, the reduction in backlogs, community outcomes and trust in government.

The LNA would streamline access to services and personalise interactions with citizens. More than this, it would drive data-led insights which would enable a proactive, rather than reactive, approach to emerging need. For the first time, demand data could be analysed at an aggregate (regional or national), segmented (group) and individual level. This would allow councils to understand issues as they arise and intervene preventatively, helping with resource allocation, revealing patterns in demand to inform service design and nudging users to the right services.

3. Modernise Local Planning with a Decision-Making Platform and an AI Assistant

Local government is critical to delivering the homes that are needed. However, planning capabilities have been squeezed or redirected (where unitary councils exist) to other priorities. In England alone, spending on planning and development fell by 58 per cent in real terms between 2010 and 2020 as constrained budgets forced difficult choices.[_]

In TBI’s paper The Urgent Need to Build More Homes we outlined national-level policy changes to facilitate housebuilding, such as allocating high-priority national sites for development, and clarifying, streamlining and simplifying planning rules and building regulations. More can be done at the local level to make the system faster, better and cheaper.

Ostensibly, England has a plan-led system.[_] Currently, each of the 337 Local Planning Authorities (LPAs) in England creates a “local plan” which outlines a strategic vision for a local area and its development policies. The plan allocates areas for housing, infrastructure and commercial development, and LPAs approve applications in line with this plan.

Yet this system is broken. Fewer than one-third of local planning departments have an up-to-date plan,[_] while many are of poor quality and have limited influence over the application process. Planning applications are beset by long delays. A reimagining of local plans and the planning-application process is required to support sustainable economic growth.

Recent national-level changes are set to impact the future of planning departments and their responsibilities. The recent proposals to change local-government boundaries will reorganise LPAs and likely aggregate planning functions across two or more local areas. These changes will make reform even more essential. At present, local authorities have different processes for creating plans, collecting, storing and sharing data, and processing applications, which present organisational barriers to the successful delivery of planning services.

Challenges in the Local-Plan Process

Local plans should provide a blueprint for transforming local areas, outlining the infrastructure required to support development and setting out clear rules around which applications should be accepted and which should not.[_]

When creating the local plan, LPAs must refer to national policy priorities and targets – including the National Planning Policy Framework, which outlined the target for 1.5 million new homes during the current parliament and the requirement to use lower-quality “grey-belt” land when necessary[_] – but have discretion on the size, type and tenure of the housing stock.[_] While national government is legislating to improve housebuilding, the success of these measures hinges on LPAs.

The local plan is created through a sequential process,[_] from initial scoping through to public engagement on strategy, evidence gathering, document drafting and examination by an appointed inspector before adoption. Guidance on a “30-month” local-plan process was issued by the previous government in 2023, including Gateway Reviews to help catch problems early. Currently, the expectation is for LPAs to update their plans every five years. LPAs are expected to monitor outcomes of the plan’s policies and send yearly reports to central government.

The system only works if LPAs have the will, capabilities and capacity to keep plans up to date. Yet some plans date from 2003 – about four years before the iPhone was released – and so are unlikely to meet the infrastructural or housing needs of the current local economy.[_] Even when plans have been updated, the process can take around seven years from scoping to publication and many are of variable quality. There are three reasons for this.

First, there are lengthy and costly data-collection processes. More guidance is required on what a local plan should include, and what evidence should be locally or nationally collected, without which there is a wide variation in data.[_] Many LPAs outsource local-area research to consultants, which can be resource- and time-intensive. This, combined with labour-intensive processes to make decisions on plans (most councils analyse consultation responses manually) limits the quality of the initial plan.

Second, LPAs can struggle to maintain coherent strategies due to changing local demands or national policies. Changes to national policy can require LPAs to dynamically adjust their plans, but many do not have the mechanisms to do this well. This can lead to the plan process being postponed or abandoned.[_] Should LPAs fail to get efficient central-government course corrections or go ahead with publication without the policy changes, the local plan is at risk of being out of date. Changes in national policies are expected; the inability of local governments to rapidly incorporate changing priorities should not be.

Third, consultation can take several months or even years, and can be dominated by obstructive local organisations representing a minority. Representations from the public can be provided in multiple formats (email, postal document or in person), which makes them difficult to analyse manually. Most citizens feel excluded from the plan-making process due to the complexity of engagement and protracted timeline.[_]

The lack of a data-driven, dynamic and fast-paced process can result in a plan that is inaccurate and irrelevant, leading to delays or even complete abandonment of local developments,[_] as well as stoking mistrust between the public and different levels of government.

Delays and Strains in Planning Applications

The planning-application process involves several steps. Citizens and organisations can ask for pre-application advice for complex projects, or submit applications to the LPA or via the National Planning Portal. Manual checks are conducted on the documentation and a case officer is assigned. The application is given a determination date (eight weeks for minor applications and 13 weeks for major developments) and is publicised for community engagement. Policy officers (or the full planning committee for complex developments) approve applications, subject to any conditions. All applications must be compliant with the local plan. Rules around “permitted development” allow some builds to take place without going through this process. [_]

However, councils have struggled to make fast and good-quality decisions on planning applications. Backlogs are endemic. In the two years to April 2024 just half (49 per cent) of minor applications were processed within the mandatory period – falling to 19 per cent for major applications. A significant proportion of applications were extended.[_] [_] [_] [_] Meeting housing and planning targets will be difficult when these delays exist in the application process.

Figure 7

Many planning applications fail to meet initial statutory timeframes (figures represent the 24 months to the end of March 2024)

Source: Department for Levelling Up, Housing and Communities[_] [_]

Note: Percentages are rounded

These delays are apparent at every stage of the application process. Pre-application advice for complex projects can take three to six months, while document validation can often “take weeks and sometimes months”.[_] Delays in major applications can also be informed by local politics.[_]

The pressure to clear backlogs can also lead to poor-quality decisions, including administrative errors, which can mean starting the process over again.[_] Planning officers sometimes struggle to monitor changes to planning proposals throughout the process, meaning that what is built often does not resemble what was presented at local-plan examination.[_]

Two reasons explain these delays and poor-quality decisions.

First, most council workforces are overburdened and suffer from a lack of support. Staff shortages are pervasive. More than half of LPAs find it difficult to recruit planning officers and more than a third have challenges retaining them.[_] The largest skills gaps for LPAs are in digital capabilities.[_] Only one in ten councils have enough staff in their planning department to meet demand[_] [_] To scale operations in the current model, national government has had to provide additional funding to increase staffing levels and capabilities,[_] [_] which is expensive and time-consuming.

Second, planning officers rarely have the tools needed to make good decisions. With more considerations to factor into planning decisions, such as biodiversity and sustainable-drainage requirements, the quantity and diversity of information required to make good decisions can be overwhelming and time-consuming.[_] Often planning officers must sift through sheets of data before making decisions.

Transforming Local Plans and Planning Applications With AI

AI can speed up the development of good-quality local plans and their use in a reformed planning-application process without compromising the quality of decisions.

AI can analyse complex data sets and make recommendations, accurately summarising national policy. It can constantly monitor changes in local conditions or national policy, flag relevant legislation, simulate the impact of policies on local areas and streamline the consultation process from months to weeks. For instance, Helsinki has developed a virtual twin of the city to model the impact of policies.[_] In the UK, West Oxfordshire and Cotswold LPAs have moved local-plan consultations to a new digital platform with the support of AI, reducing time spent on consultation by 85 per cent.[_]

In the planning-application process, there are a growing number of tools to dramatically reduce the time citizens spend filling in forms and support planning officers to validate documents, consult and make decisions.

Spotlight

How are organisations innovating?

Several organisations have developed AI tools to transform aspects of the planning-application process

  • Searchland has an interactive map of local policies which can be navigated by citizens using an AI assistant, enabling users to find specific land opportunities efficiently.[_]

  • Genie AI uses AI to draft Section 106 agreements (documents used for larger developments to mitigate risks to the local area). Genie AI reviews documents by examining each clause, comparing it to standard or expected provisions, and engages with stakeholders. This tool could save eight to ten hours per week, improve the speed of applications by 50 per cent and save up to £12,000 per contract.[_]

  • The city of Sydney uses AI to provide immediate feedback to citizens on non-compliant segments of planning applications.[_] This has eliminated the need to deal with 30,000 applications manually, many of which had errors which contributed to delays.

  • InformedDECISION is an AI-powered support platform that streamlines the planning process in Scotland. It has led to a 50 to 60 per cent reduction in case triage time, and a 40 to 60 per cent reduction in time spent evaluating and responding to uncomplicated statutory consultations.[_]

  • The UK government’s Incubator for AI, working in partnership with MHCLG, is developing a tool that can turn analogue planning documents or PDF scans into machine-readable geospatial data in under a minute compared to the 1–2 hours it currently takes.[_]

Where these or similar tools are used in the UK, they are limited to pockets of good practice across local government. Two barriers limit their potential.

First, many of these tools are incompatible with existing CMSs or with each other, leading to low take-up and increased effort for citizens and planning officers to use them. There are few incentives for incumbent technology providers to open their systems while long-term, rigid contracts facilitate vendor lock-in. These incumbents offer products or services to compete with third parties; given the barriers to change, it can be easier for councils to accept them even if products are inferior.

Second, most local governments do not have the data infrastructure required to implement AI. For instance, to facilitate quick decisions in planning applications, planning rules need to have been translated into machine-readable data. Historical data, which can be used to train new tools or inform policy decisions, are locked in proprietary CMSs, requiring expensive and time-consuming data migration.

The government has made some progress towards better planning-data infrastructure. For instance, through its national Planning Data Platform MHCLG hopes to index planning data and make them available in a standardised format, including data on listed buildings and flood risk.[_] Powers in the Levelling-Up and Regeneration Act (LURA) allow MHCLG to mandate data standards, but to date only 43 local governments are currently implementing them.[_] [_] MHCLG is currently working on creating more standards for planning applications.[_] This will be an important first step towards challenging incumbent technology providers to improve.

In addition, community-led practices, including through the Open Digital Planning Community[_] (funded by MHCLG), have led to the introduction of innovative digital planning software. This software includes PlanX, a citizen-facing rules-based system that includes guidance and submission services.[_] In recent years, MHCLG has provided funding to support more than 100 LPAs to join the Open Digital Planning Community, to assess and plan how to improve their digital maturity, and to publish data sets on the Planning Data Platform.[_]

Finally, Development Corporations (DCs) are being established by mayors to deliver regeneration and property development, and will likely be the vehicle to create new towns. DCs are empowered to make key decisions related to land acquisition, infrastructure development and planning, often by becoming the LPA for their area, with the goal of regenerating areas, attracting investment and promoting sustainable development. As new organisations, they are not encumbered by legacy systems and are solely focused on delivering the planning functions, which provides them with the perfect opportunity to test AI innovations.

The impact of AI on planning and economic development

TBI estimates that using AI for planning and economic development roles could lead to a 35 per cent time saving across local government, which would free up almost two days per week for housing and economic planners.[_]

More can be done to improve the quality and dynamism of local plans, the speed of planning-application decisions, and how these processes interact with each other. DAIS should introduce a new data and decision-making platform for local-plan development and an AI Planning Assistant to streamline the planning process.

Introducing the data and decision-making platform for local-plan development will require the following steps:

  • DAIS should create a platform for local plans that hosts tools to produce them in a consistent, machine-readable format.

  • As a first step, DAIS should work with MHCLG to create a taxonomy on how to classify and organise planning information, based on shared national standards for the structure and content of local plans.

  • DAIS should create an AI tool that can transfer existing plans to this platform. The standard way of presenting plans would enable DAIS to visualise policies and progress at a local, regional and national level, supporting housing and infrastructure planning at every level.

  • DAIS should work with MHCLG to link nationally collected, locally relevant data hosted on the Planning Data Platform to the local-plan platform. This should be expanded to include social and behavioural data, such as migration patterns, demographic and family data. DAIS should use pattern-recognition algorithms to identify key trends in these data to anticipate future needs of the local area, as has been trialled in Singapore.[_]

  • DAIS should create guidance on the type and quality of additional data that could be collected locally relevant to the creation of local plans. MHCLG should consider enforcing existing data standards (under LURA 2023) for the collection of this input data, and DAIS should create or modify existing data-cleaning tools to support compliance.

  • Given the potential disruption to LPAs in the government’s Devolution Priority Programme, DAIS should support the early piloting of AI planning tools through Development Corporations. These areas should be set up as regulatory sandboxes to accelerate innovation and test new rules, and as guidance for plan-making. DAIS could develop an AI tool that can create an initial draft of the local plan using the data and standards described earlier or trial a digital-twin tool to model different planning scenarios and improve decision-making.[_] [_] As more LPAs join the platform, DAIS could use pattern-recognition technology to identify LPAs with similar priorities and initiatives, and facilitate knowledge-sharing.

  • DAIS should incorporate consultative tools on the platform, such as those created by i.AI,[_] to gain a holistic and accurate view of community needs and preferences, as well as expert opinion, helping to speed up the process and enhance co-production.

  • DAIS should create an AI-enabled dynamic process of automatic checks for local plans against national legislation, flagging issues for further consideration. Such tools are being built in other areas – for instance, i.AI is trialling a tool to interrogate existing laws, called Lex,[_] while the Department for Business and Trade is developing a platform to provide access to UK regulation for businesses as machine-readable data.[_] Once this is in place, MHCLG should remove static and reactive Gateway Reviews (periodic assessments by the Planning Inspectorate to ensure the local plan is on track) inherited from the previous government in favour of a dynamic checking process.

  • The current approach to incentivising more frequent updates to plans can be improved, as the current statutory targets of five-year reviews are rarely met.[_] With better tools available via DAIS, MHCLG can move to a more proactive system where, as the context changes (either through modifications to national legislation or because of changes in underlying data), LPAs should instead receive input from MHCLG detailing where and when the plan should be tweaked. Rather than creating a supplementary document, LPAs should be able to update the existing plan quickly via the platform. MHCLG should offer guidance on which changes require consultation and hold LPAs to account for delivering these changes by publishing tracking updates to plans, which this digital system would make significantly easier.

  • DAIS should publish the standardised local-plan data to spur innovation in citizen-facing tools.

Figure 8

How the local-plan process would change

Source: TBI analysis and Gov.uk[_]

Introducing an AI Planning Assistant for the end-to-end planning-application process will require the following steps:

  • DAIS should set a goal for an end-to-end AI Planning Assistant to be used by every LPA, streamlining tasks and offering access to the process at any time of day. It could also be integrated into the LNA.

  • DAIS should define modular components for stages of the planning process (for example, ecological requirements or local-plan checks). Innovation competitions should be set up in areas where there is a dearth of tools. The LPA should have choice in selecting the AI tool, including the components it wants to buy. DAIS should encourage SMEs to engage by allowing them to bid for contracts in a consortium.

  • DAIS should map out the “input” data sets vital to the performance of each of these tasks. The Planning Data Platform already brings together much of these data in a standard fashion, but many local authorities do not use it. MHCLG should use the powers granted by LURA to ensure all LPAs adhere to these data standards. DAIS can support with compliance for these standards by providing expert advice, assessments on key changes and helping to build/procure compliance software to conduct data cleaning or transfer tasks.

  • DAIS should work with MHCLG to codify development rules, already started in PlanX, including machine-readable rules on permitted development, planning conditions and case law. Recent examples include the Brownfield Passport.[_] These clear rules will help improve the pace of planning and facilitate the integration of AI.

  • DAIS should build repositories of current and historical CMS data to support policymaking, operational decisions and tool development. In the short term, DAIS could procure or build a generative AI tool to extract and analyse historical data held in CMSs. To ensure long-term improvements, DAIS should explore new data standards for “output” data stemming from the planning-application process, enhancing interoperability between CMSs.[_] [_] A “Smart Data” scheme[_] – modelled on Open Banking – would enable secure, standardised data exchange via APIs, reducing barriers to innovation and improving service delivery.

  • DAIS should allow citizens to view relevant operational and decision-making data in the process and be notified about progress. Protocols should be created to enable requests for human support at every stage. The document-validation and approval processes should be combined and streamlined using AI, allowing for the provision of feedback to citizens, including the likelihood of the application’s success, and the option to proceed. DAIS should adopt TBI’s “earned-autonomy” framework, outlined in the paper Governing in the Age of AI: A New Model to Transform the State, allowing the tool to prove its accuracy before moving to the next level of autonomy.

  • Codified rules and data standards enable AI tools to approve “default” decisions in simple applications, such as building a single home on a brownfield site. DAIS should adopt a human-near-the-loop approach, also referenced in A New Model to Transform the State, with staff checking a selection of automatically approved applications. More complex developments, or those with a lower confidence of being approved, should be batched using robotic process automation according to the type of application and issues, and intelligently assigned to officers for investigation.

  • An AI consultation tool should summarise public feedback on planning applications.

With major housebuilding and energy ambitions, the creation of AI growth zones[_] and forthcoming changes to planning-permission rules, now is the time to fix the underlying processes for local plans and planning applications, so that these crucial decisions to boost economic growth are not stalled by delays and poor-quality operational decisions.

Introducing AI can transform how local plans are created and managed, and speed up the planning process. Good-quality, dynamic local plans that are better integrated with the application process will improve the quality and accountability of decisions, and provide a shared vision that can spur investment. Citizens would experience seamless approval for simple applications, allowing planning officers to spend more time on complex cases. Finally, central government would have visibility over alignment with policies at a local level and could track the progress towards these goals in real-time.


Chapter 4

Recommendations for Councillors

The creation of DAIS would help to build a shared vision and governance framework, pilot new tools and services, curate the environment for scaling innovation and monetise international access to innovative tools. However, councillors do not need to wait for this institution to be set up. They can lead the charge and drive change locally. Many of these recommendations can also apply to chief executives and directors. There are five areas in which councillors can make progress:

1. Champion AI innovation

  • Serve as an advocate for AI, recognising it is a fundamental tool, and push for it to be on the agenda of cabinet and scrutiny committees. For instance, Hammersmith and Fulham Council has a cabinet member responsible for AI and another for digital innovation. Updates on AI projects are regularly given to the Policy Oversight Board.

  • Request a baseline audit of repetitive, admin-heavy tasks across services such as social care, planning and benefit transactions within the local authority to identify low-lift automatable tasks.

  • Bring together service leads, digital teams and finance departments to identify a small number of high-volume, high-cost pain points that could be improved with AI.

  • Continually communicate with the workforce about the introduction of AI and what it means for them, such as the potential to improve their working conditions and shift workflows to more productive and fulfilling tasks. Sharing successful case studies and clearly linking the use of these tools to problems workers face can help reduce cultural barriers to adoption.

2. Adopt tools that have been proven to work

  • Ask officers to review existing AI pilots in other local areas (such as those identified in this paper for case-note transcription, service navigation or planning) and identify where these could be safely tested. Work with teams in service departments and with citizens to identify the key problems that could be addressed through existing tools.

  • While regulation is under consideration for emerging AI capabilities and frontier models, local governments could create interim policies that balance risk with innovation. The example of Seattle, which – in the absence of national guidance – introduced guidance when LLMs were released, shows how councils can be proactive in creating the conditions for experimentation.[_]

3. Use convening power to de-risk adoption

  • Promote collaboration and shared learning across local government by engaging with networks such as LOTI and LGA, and connect with other pioneering local authorities to inform best practice.

  • Where councils are being restructured under the local government reorganisation plans, they should start to work with their neighbouring local governments on data-sharing and AI innovation in preparation for the move towards unitary authorities.

  • Set up collaborations with local universities, SMEs or other civic technology organisations to co-design local testbeds for AI tools that encourage responsible adoption.

4. Adopt existing standards and platforms

  • Adopt existing data standards such as Open Referral UK (for service directories), the Planning Data Platform and LocalGov Drupal (for local government websites).

  • Require procurement teams to use government procurement frameworks such as G-Cloud to help de-risk the adoption of tools.

5. Adapt the workforce and attract the right talent

  • Conduct a review of AI-literacy training for councillors and officers.

  • Request a skills audit to identify roles most impacted by automation and explore how they could evolve.

  • Urgently review pay for AI and technology roles (in our earlier paper A New Model to Transform the State, TBI recommended benchmarking salaries for AI-related roles to at least 75 per cent of the private-sector market rate, later mirrored in the government’s AI Opportunities Action Plan).

  • Review career-progression pathways for AI-related professionals.[_]


Chapter 5

Conclusion

It is challenging for local government to offer citizens the quality, dependability and speed of service they deserve. At the heart of this is a bureaucratic, labour-intensive and unresponsive operating model that prevents local government from scaling services to support citizens and businesses when they need it.

Advances in AI can help local government deliver more for less. Local areas could become testbeds of innovation and ideas.

However, at present they are held back by risk-averse organisational cultures, innovation siloes and limited market power. A wholesale change to the institutional landscape is required. The introduction of the Devolved AI Service will help councils to create a shared long-term vision for AI innovation, improve capabilities, build and scale AI tools and services, and create a sustainable independent funding stream for local innovation. In this paper we have outlined three opportunities to introduce:

  1. An AI co-worker in high-volume, high-cost services, starting with an end-to-end tool to support social workers.

  2. A Local Navigation Assistant, to efficiently and effectively navigate citizens to the right services first time.

  3. A data and decision-making platform to create dynamic and high-quality local plans and an AI Planning Assistant to streamline the planning process.

Through these interventions, local government can regain its ability to transform communities and drive economic growth. Instead of firefighting crises or simply serving as the delivery arm for central government, councils can focus on long-term priorities – building vibrant, resilient local economies and delivering meaningful improvements in residents’ lives. In the age of AI, local governments now have the tools and must use them to deliver on this vision.


Chapter 6

Acknowledgements

We would like to extend our thanks to those who offered their advice and guidance in the development of this report (while noting that contribution does not equal endorsement of all the points made and does not reflect the views of respective employers).

Jay Bangle, TPXimpact

Bury Council

Calum Campbell, Social Work Scotland

Hammersmith and Fulham Council

Dani Ismailov-Law, Beam

Leeds City Council

May-N Leow, Faculty

Local Government Association

Kingston Council

Paul Maltby, Faculty

Michael Mander, Beam

Nathan Nagaiah, UK Centre for AI in the Public Sector / Newham Council

Sam Nutt, LOTI

Tom Rintoul, Social Finance

Peter Skinner, NHS England

Stefan Webb, TPXimpact

TBI staff:

Calum Handforth

Tone Langengen

Dr Laura Gilbert


Chapter 7

Methodology

To estimate the potential time savings from using AI in local government, we partnered with a council. This council provided TBI with a list of employment roles across the council, the departments in which they were located and the salary bands for each role.

Using this information, we matched the roles to the O*NET database via the Warwick University Computer Assisted Structured Coding Tool (CASCOT). As O*NET is based on US occupational data, CASCOT was used to identify the closest UK Standard Occupational Classification (SOC) codes for each role, which were then matched to their nearest equivalents in O*NET. To ensure accurate alignment, we cross-referenced both the 2010 and 2019 SOC codes to find the best possible match for each position. Where the role provided by the council was ambiguous, we used descriptions of the role found online to match with core tasks in the O*NET database.

Once suitable matches were identified, we used TBI’s bespoke tool to calculate the estimated time savings for each occupation. This tool draws on LLMs to assess how well AI can perform 19,000 distinct tasks listed in the O*NET database and merges this information with UK Labour Force Survey data to estimate time savings based on how important each task is to the overall role. (A full description of how this database was created can be found in TBI’s paper The Potential Impact of AI on the Public-Sector Workforce.)

Our analysis uses this database to find an aggregate time saving for full-time equivalent staff and the potential cost saving using the mid-point of salary bands.

We then extrapolate the figures for this local government to the national level by dividing by the number of people in this council and multiplying by the number of people employed in local government, a figure provided by the Office for National Statistics (ONS). For the potential cost saving, we adjusted for potential pay differentials across local government using ONS pay data. The key caveat to this national figure is the potential change in the composition of roles across local governments, which reflect key demographic and economic situations in local areas.

Footnotes

  1. 1.

    In this paper we focus on “principal authorities” (London borough councils, unitary councils, metropolitan borough councils, county councils, and district councils). Combined authorities are also included. Other types of local government such as fire and rescue authorities, police and crime commissioners, and parish and town councils are not included in this paper. https://lgiu.org/resources/local-government-facts-and-figures/local-government-facts-and-figures-england/

  2. 2.

    https://www2.local.gov.uk/publications/local-government-white-paper

  3. 3.

    https://www.nao.org.uk/wp-content/uploads/2012/06/1213187es.pdf

  4. 4.

    https://www.local.gov.uk/parliament/briefings-and-responses/public-health-grant-allocations-local-authorities-202526

  5. 5.

    https://www.pwc.co.uk/industries/assets/future-of-local-government-report.pdf

  6. 6.

    https://www.adass.org.uk/wp-content/uploads/2024/07/ADASS-Spring-Survey-2024-FINAL-1.pdf

  7. 7.

    https://www.ageuk.org.uk/latest-press/articles/2024/new-age-uk-analysis-finds-that-28655-older-people-died-waiting-for-social-care-in-the-most-recent-year-for-which-figures-are-available/

  8. 8.

    https://www.gov.uk/government/statistical-data-sets/live-tables-on-planning-application-statistics

  9. 9.

    https://news.sky.com/story/most-areas-set-to-increase-council-tax-by-at-least-4-99-from-april-13312588

  10. 10.

    https://obr.uk/docs/dlm%5Fuploads/E03004355%5FNovember-Economic-and-Fiscal-Outlook%5FWeb-Accessible.pdf

  11. 11.

    https://www.instituteforgovernment.org.uk/explainer/local-authority-section-114-notices

  12. 12.

    https://ifs.org.uk/publications/how-have-english-councils-funding-and-spending-changed-2010-2024

  13. 13.

    https://ifs.org.uk/publications/how-have-english-councils-funding-and-spending-changed-2010-2024

  14. 14.

    https://www.ageuk.org.uk/siteassets/documents/reports-and-publications/reports-and-briefings/health--wellbeing/age_uk_briefing_state_of_health_and_care_of_older_people_july2023.pdf

  15. 15.

    https://webfs.oecd.org/Els-com/Affordable%5FHousing%5FDatabase/HC3-1-Population-experiencing-homelessness.pdf

  16. 16.

    https://www.local.gov.uk/sites/default/files/documents/Resident%20Satisfaction%20Polling%20Round%2038%20-%20Final.pdf

  17. 17.

    https://www.local.gov.uk/sites/default/files/documents/february-2015-polling-ful-de7.pdf

  18. 18.

    https://www.historyandpolicy.org/policy-papers/papers/a-central-role-for-local-government-the-example-of-late-victorian-britain

  19. 19.

    Altogether, 1,003,528 hours of time were estimated to be saved each year through the introduction of AI, representing a potential productivity improvement of £29,808,690

  20. 20.

    Total time saved per year is not FTE-adjusted and so figures are likely to be a slight overestimate of the hours saved.

  21. 21.

    https://www.newham.gov.uk/newham-data/uk-centre-ai-public-sector

  22. 22.

    https://www.local.gov.uk/case-studies/hertfordshire-county-council-ai-road-maintenance-robot

  23. 23.

    https://www.ft.com/content/29afb50c-60c0-4f0f-90b9-8a3a3f397690

  24. 24.

    https://www.lbhf.gov.uk/news/2024/09/hf-tackling-damp-and-mould-tech

  25. 25.

    https://www.local.gov.uk/publications/state-sector-artificial-intelligence

  26. 26.

    https://mhclgdigital.blog.gov.uk/wp-content/uploads/sites/222/2024/04/FC-Pilot-Report.pdf

  27. 27.

    https://www.local.gov.uk/publications/state-sector-artificial-intelligence

  28. 28.

    https://mhclgdigital.blog.gov.uk/wp-content/uploads/sites/222/2024/04/FC-Pilot-Report.pdf

  29. 29.

    https://www.gov.uk/government/publications/state-of-digital-government-review/state-of-digital-government-review

  30. 30.

    https://www.computerweekly.com/news/450303822/A-quarter-of-local-council-procurement-policies-do-not-support-use-of-G-Cloud-research-shows

  31. 31.

    https://www.local.gov.uk/our-support/cyber-digital-and-technology/cyber-digital-and-technology-policy-team/ai-consultation#-ai-use-in-local-government

  32. 32.

    https://www.gov.uk/government/publications/state-of-digital-government-review/state-of-digital-government-review

  33. 33.

    https://www.gov.uk/government/publications/state-of-digital-government-review/state-of-digital-government-review

  34. 34.

    https://mhclgdigital.blog.gov.uk/wp-content/uploads/sites/222/2024/04/FC-Pilot-Report.pdf

  35. 35.

    https://www.gov.uk/government/publications/state-of-digital-government-review/state-of-digital-government-review

  36. 36.

    https://chiefdigitalofficer4london.medium.com/further-thoughts-on-a-new-local-government-digital-service-24fce01be865

  37. 37.

    https://mhclgdigital.blog.gov.uk/wp-content/uploads/sites/222/2024/04/FC-Pilot-Report.pdf

  38. 38.

    https://openreferraluk.org/

  39. 39.

    https://www.localdigital.gov.uk/fund/

  40. 40.

    https://www.localdigital.gov.uk/declaration/

  41. 41.

    https://chiefdigitalofficer4london.medium.com/towards-a-new-data-for-london-library-d95a062f30d9

  42. 42.

    https://www.gov.uk/government/news/experimental-ai-could-help-councils-meet-housing-targets-by-digitising-records

  43. 43.

    https://gds.blog.gov.uk/2025/01/27/same-name-new-ambitions/

  44. 44.

    https://www.gov.uk/government/publications/a-blueprint-for-modern-digital-government/a-blueprint-for-modern-digital-government-html

  45. 45.

    https://www.gov.uk/government/collections/local-government-reorganisation-policy-and-programme-updates

  46. 46.

    https://questions-statements.parliament.uk/written-statements/detail/2025-02-05/hcws418

  47. 47.

    https://www.gov.uk/government/news/pat-mcfadden-vows-to-make-the-state-more-like-a-start-up-as-he-deploys-reform-teams-across-country

  48. 48.

    https://www.localdigital.gov.uk/transforming-the-housing-sector-through-improved-data/

  49. 49.

    https://www.gov.uk/government/publications/a-blueprint-for-modern-digital-government/a-blueprint-for-modern-digital-government-html

  50. 50.

    https://www.aitdp.eclab.uel.ac.uk/

  51. 51.

    https://www.opengovpartnership.org/ogp-local/join-ogp-local-call-for-expressions-of-interest-2023-concluded/

  52. 52.

    https://www.pd-g.de/assets/Gesellschafter/PD-About%5Fus.pdf

  53. 53.

    Principal councils include London borough councils, unitary councils, metropolitan borough councils, county councils and district councils. A combined authority is a legal body enabling a group of two or more councils to collaborate and take collective decisions across council boundaries: https://www.local.gov.uk/sites/default/files/documents/combined-authorities-plai-fb6.pdf

  54. 54.

    https://www.pwc.co.uk/industries/assets/future-of-local-government-report.pdf

  55. 55.

    https://www.nhs.uk/conditions/social-care-and-support-guide/help-from-social-services-and-charities/getting-a-needs-assessment/

  56. 56.

    https://www.adass.org.uk/wp-content/uploads/2024/07/ADASS-Spring-Survey-2024-FINAL-1.pdf

  57. 57.

    https://www.cqc.org.uk/publications/major-report/state-care/2023-2024/access

  58. 58.

    https://news.sssc.uk.com/news/another-increase-in-social-worker-numbers - these figures are for all social workers in Scotland

  59. 59.

    https://www.skillsforcare.org.uk/Adult-Social-Care-Workforce-Data/Workforce-intelligence/documents/Social-Worker-Headline-report-Feb-2024.pdf - these figures are purely for adult social care workers

  60. 60.

    https://www.communitycare.co.uk/2024/02/26/use-of-agency-social-workers-up-25-year-on-year-in-adults-services/1`

  61. 61.

    https://www.nao.org.uk/wp-content/uploads/2023/11/Report-reforming-adult-social-care-in-England.pdf

  62. 62.

    https://www.skillsforcare.org.uk/Adult-Social-Care-Workforce-Data/Workforce-intelligence/documents/Social-Worker-Headline-report-Feb-2024.pdf

  63. 63.

    https://edemocracy.coventry.gov.uk/documents/s37629/Appendix%202%20-%20Social%20Work%20Research%20-%20Retaining%20Social%20Workers.pdf

  64. 64.

    https://www.communitycare.co.uk/2024/09/30/watchdog-increasingly-finding-fault-with-councils-on-social-care-assessments-and-plans/

  65. 65.

    https://democracy.havering.gov.uk/documents/s74429/139.%20AI%20Case%20Work%20Software%20Pilot%20Non%20Key%20Decision%20Paper%20Signed%20Version.pdf

  66. 66.

    https://www.kingston.gov.uk/news/article/525/kingston-council-launches-ai-tool-to-free-up-time-for-frontline-social-workers

  67. 67.

    https://techhq.com/2023/05/generative-ai-in-social-care/

  68. 68.

    https://www.nextgen.com/company/newsroom/press-release/nextgen-healthcare-unveils-latest-ai-driven-advancements-to-transform-the-patient-provider?utm%5Fsource=website&utm%5Fmedium=organic%5Fsocial&utm%5Fcampaign=701Do000000kjfPIAQ&utm%5Fcontent=facebook

  69. 69.

    These insights draw from the local-authority analysis used in this paper.

  70. 70.

    There are around 18,500 social workers employed by local authorities, who each spend on average 29 hours per week on paperwork or computer tasks. To understand the total time required to perform an assessment, we have attributed the following time to each task involved in social-work assessment: validate care request and prepare for assessment (90 minutes), travel to place, assess individual, consult with other professionals, review findings and make decision, type up assessment notes (180 minutes). There are 227,000 assessments waiting to be conducted as of March 2024.

  71. 71.

    https://www.skillsforcare.org.uk/Adult-Social-Care-Workforce-Data/Workforce-intelligence/documents/Social-Worker-headline-Feb2023-FINAL.pdf

  72. 72.

    https://basw.co.uk/sites/default/files/2023-08/FINAL%2080-20%20report.pdf

  73. 73.

    https://ai.gov.uk/projects/minute/

  74. 74.

    https://assets.publishing.service.gov.uk/media/66190f98679e9c8d921dfe44/smart-data-roadmap-action-the-government-is-taking-in-2024-to-2025.pdf

  75. 75.

    For instance, standard care assessments mean councils could adopt a trusted-assessor model, allowing others to perform assessments for care (like care homes), similar to the model used in hospital discharges.

  76. 76.

    https://www.cqc.org.uk/sites/default/files/20180625%5F900805%5FGuidance%5Fon%5FTrusted%5FAssessors%5Fagreements%5Fv2.pdf

  77. 77.

    https://www.local.gov.uk/publications/responsible-buying-how-build-equality-data-protection-your-ai-commissioning

  78. 78.

    https://www.cqc.org.uk/publications/major-report/state-care/2023-2024/access/asc

  79. 79.

    https://patterns.wearefuturegov.com/apply

  80. 80.

    The net promoter score ranges from -100 (very unlikely to recommend) to +100 (very likely to recommend).

  81. 81.

    https://assets.kpmg.com/content/dam/kpmg/uk/pdf/2023/04/uk-local-government-cee-report.pdf

  82. 82.

    https://www.local.gov.uk/sites/default/files/documents/Resident%20Satisfaction%20Polling%20Round%2038%20-%20Final.pdf

  83. 83.

    https://www.wandsworth.gov.uk/housing/benefits-and-support/housing-benefit/make-a-benefit-claim/

  84. 84.

    https://www.gov.uk/government/statistics/housing-benefit-statistics-on-speed-of-processing-for-2024-to-2025/statistical-release-october-to-december-2024-quarter-3-financial-year-2024-to-2025

  85. 85.

    https://www.ipsea.org.uk/moving-to-a-new-local-authority

  86. 86.

    https://www.ons.gov.uk/peoplepopulationandcommunity/populationandmigration/migrationwithintheuk/bulletins/peoplewithadifferentaddressinmarch2020tocensusday2021englandandwales/census2021

  87. 87.

    https://policyinpractice.co.uk/missing-out-2024/

  88. 88.

    https://policyinpractice.co.uk/missing-out-2024/

  89. 89.

    https://policyinpractice.co.uk/missing-out-2024/

  90. 90.

    https://mortar.works/case-studies/find-digital-help-in-salford/

  91. 91.

    https://www.themj.co.uk/pwc-hillingdon-leading-ai-residents

  92. 92.

    https://foundationsweb.s3.amazonaws.com/4210/foundations-dfg-foi-report-nov-2015.pdf

  93. 93.

    https://www.ey.com/en%5Fgl/industries/government-public-sector/using-data-to-improve-social-services-in-london-s-most-deprived-borough

  94. 94.

    https://www.mca.org.uk/consulting-case-studies/ey-with-the-london-borough-of-barking-and-dagenham

  95. 95.

    https://openreferraluk.org/

  96. 96.

    https://developers.openreferraluk.org/Guidance/

  97. 97.

    https://mhclgdigital.blog.gov.uk/2024/03/06/driving-adoption-of-open-referral-uk-to-deliver-millions-in-annual-savings-for-councils/

  98. 98.

    We included in our analysis all workers who have clearly defined citizen-facing roles that would involve the types of tasks that the LNA would help the council perform. These include staff with responsibilities for services such as waste collection, potholes, street cleaning, libraries, parks and green spaces, leisure and sports facilities, and council tax and benefits administration.

  99. 99.

    These insights draw from the local-authority analysis used in this paper.

  100. 100.

    https://ifs.org.uk/publications/how-have-english-councils-funding-and-spending-changed-2010-2024

  101. 101.

    https://www.gov.uk/guidance/local-plans

  102. 102.

    https://www.cpre.org.uk/about-us/cpre-media/only-third-councils-have-up-to-date-local-plan

  103. 103.

    https://www.miriamcates.org.uk/news/importance-having-local-plan

  104. 104.

    https://www.gov.uk/government/news/planning-overhaul-to-speed-up-and-simplify-local-plans

  105. 105.

    https://www.gov.uk/government/publications/reforms-to-national-planning-policy-report-government-response/reforms-to-national-planning-policy-report-government-response

  106. 106.

    https://www.gov.uk/government/consultations/plan-making-reforms-consultation-on-implementation/levelling-up-and-regeneration-bill-consultation-on-implementation-of-plan-making-reforms

  107. 107.

    https://www.hbf.co.uk/documents/13042/Delayed%5FLocal%5FPlans%5F-%5F8th%5FNovember%5F2023.pdf

  108. 108.

    For instance, the Housing and Economic Land Availability Assessment (HELAA),a technical study that identifies potential land for housing and economic development over a local plan period. However, this varies between local governments which can lead to different assessments of available land for development.

  109. 109.

    https://www.gov.uk/government/publications/reforms-to-national-planning-policy-report-government-response/reforms-to-national-planning-policy-report-government-response

  110. 110.

    https://www.rtpi.org.uk/media/15867/what-does-a-good-local-plan-look-like-discussion-paper-oct-2022.pdf

  111. 111.

    For instance, Natural England on water quality in an area in Somerset has put the delivery of 11,000 homes at risk – infrastructure investment programmes not aligned with local plan making.

  112. 112.

    https://www.planningportal.co.uk/planning/planning-applications/the-decision-making-process/introduction

  113. 113.

    The key reasons for extensions: planning performance agreement, an environmental-impact assessment or an agreed extension of time. The data does not differentiate between these categories.

  114. 114.

    “Other” cases are most likely applications not decided within the timeframe and not extended.

  115. 115.

    In the year to March 2023, one in 12 cases referred to the national planning body, the Planning Inspectorate, resulted from LPAs not making a decision on an application within the allotted time.

  116. 116.

    https://researchbriefings.files.parliament.uk/documents/SN06790/SN06790.pdf

  117. 117.

    https://www.gov.uk/government/statistical-data-sets/live-tables-on-planning-application-statistics

  118. 118.

    In the two years to April 2024, English LPAs received 21,997 major development and 610,016 minor development applications.

  119. 119.

    https://urbanistarchitecture.co.uk/recent-planning-application-delays/

  120. 120.

    https://planning-by-design.co.uk/planning-permission-in-decline-a-growing-crisis-for-the-uks-housing-sector

  121. 121.

    https://formedarchitects.com/planning-permission-delays-in-2023/

  122. 122.

    https://www.rtpi.org.uk/policy-and-research/planning-for-a-better-future/#M-1.1

  123. 123.

    https://www.local.gov.uk/sites/default/files/documents/LG Workforce Survey 2022 - Final for Publication - Tables Hard Coded.pdf

  124. 124.

    https://drive.google.com/file/d/1%5FHBNGOYWy2tzIFWyvfDkdEzLTqO01liO/view

  125. 125.

    https://www.lgcplus.com/services/regeneration-and-planning/revealed-capacity-and-churn-issues-facing-planning-teams-16-05-2023/

  126. 126.

    https://www.rtpi.org.uk/media/16015/state-of-the-profession-2023-final.pdf

  127. 127.

    https://www.gov.uk/government/publications/reforms-to-national-planning-policy-report-government-response/reforms-to-national-planning-policy-report-government-response

  128. 128.

    https://www.gov.uk/government/news/planning-overhaul-to-reach-15-million-new-homes

  129. 129.

    https://www.localgov.co.uk/Leveraging-data-to-reverse-the-planning-staff-decline/60244

  130. 130.

    https://www.hel.fi/en/decision-making/information-on-helsinki/maps-and-geospatial-data/helsinki-3d

  131. 131.

    https://lgiu.org/blog-article/ai-and-the-council-case-studies-and-resources-for-local-government/

  132. 132.

    https://searchland.co.uk/blog/local-plan-policy-tool

  133. 133.

    https://www.genieai.co/blog/step-by-step-guide-to-drafting-a-section-106-agreement

  134. 134.

    https://www.dlapiper.com/en/insights/publications/2024/01/how-ai-can-be-used-in-local-government-in-2024

  135. 135.

    https://www.informed.com/news/naturescot-and-informed-solutions-win-digital-category-at-scottish-planning-innovation-awards/

  136. 136.

    https://www.gov.uk/government/news/experimental-ai-could-help-councils-meet-housing-targets-by-digitising-records

  137. 137.

    https://www.planning.data.gov.uk/about/

  138. 138.

    https://www.planning.data.gov.uk/about/

  139. 139.

    https://www.footanstey.com/our-insights/articles-news/planning-data-under-the-levelling-up-and-regeneration-act-2023/

  140. 140.

    https://mhclgdigital.blog.gov.uk/2025/04/17/digital-planning-get-involved-validating-the-baseline-planning-application-data-specifications/

  141. 141.

    https://opendigitalplanning.org/

  142. 142.

    https://www.planx.uk/How-it-works

  143. 143.

    https://mhclgdigital.blog.gov.uk/2025/03/18/funding-boost-to-support-digital-planning-innovation/

  144. 144.

    These insights draw from the local-authority analysis used in this paper.

  145. 145.

    https://www.ura.gov.sg/Corporate/Resources/Ideas-and-Trends/AI-in-Urban-Planning

  146. 146.

    These digital-twin tools could eventually be federated into a National Policy Twin, which TBI has proposed in a previous paper Governing in the Age of AI: A New Model to Transform the State

  147. 147.

    Bradford is already trialling a similar tool: https://www.local.gov.uk/case-studies/bradford-district-council-digital-twin

  148. 148.

    https://ai.gov.uk/projects/consult/

  149. 149.

    https://ai.gov.uk/blogs/improving-legislative-drafting-with-lex/

  150. 150.

    https://www.public.io/case-study/open-regulation-platform-orp

  151. 151.

    https://www.housingtoday.co.uk/news/five-yearly-local-plan-review-requirement-not-operating-effectively-says-lichfields/5127329.article

  152. 152.

    https://www.gov.uk/government/consultations/plan-making-reforms-consultation-on-implementation/levelling-up-and-regeneration-bill-consultation-on-implementation-of-plan-making-reforms

  153. 153.

    https://assets.publishing.service.gov.uk/media/66ef1962d82c72546b9a8cc0/PLANNING%5FREFORM%5FWORKING%5FPAPER%5F-%5FBROWNFIELD%5FPASSPORT.pdf

  154. 154.

    This could include building on existing standards, such as the LGA’s minimum data set for planning applications for residential properties.

  155. 155.

    https://data.london.gov.uk/dataset/residential-completions-dashboard

  156. 156.

    https://api.startupcoalition.io/u/2025/02/FOR-RELEASE-Startup-Coalition-X-TBI-Smart-Data-Report.pdf

  157. 157.

    https://www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-action-plan

  158. 158.

    https://digitalgovernmenthub.org/examples/city-of-seattle-interim-generative-ai-policy/

  159. 159.

    In a recent survey, 53 per cent of local authority respondents thought that career-progression pathways were poorly defined:https://www.gov.uk/government/publications/state-of-digital-government-review/state-of-digital-government-review

Newsletter

Practical Solutions
Radical Ideas
Practical Solutions
Radical Ideas
Practical Solutions
Radical Ideas
Practical Solutions
Radical Ideas
Radical Ideas
Practical Solutions
Radical Ideas
Practical Solutions
Radical Ideas
Practical Solutions
Radical Ideas
Practical Solutions