What is Artificial Intelligence?
The technologies underlying “Artificial intelligence” (AI) are constantly evolving, which necessitates some elasticity in its definition. For the purposes of this report, “AI” refers to machines behaving in ways that would be called intelligent if seen in a human.1 In other words, these systems can learn, reason, and adapt, performing tasks that typically require human cognition. AI can encompass a wide range of capabilities, varying greatly depending on the context in which it is applied. From simple rule-based systems to complex machine learning models, AI can take many forms to meet the needs of different users.
Other useful terms include:
Generative AI:
Generative AI refers to a category of AI that can create new content, such as text, images, music, or other data, that is similar to what it was trained on. This is achieved using algorithms, often based on deep learning models, to generate new outputs that mimic the style or characteristics of the training data.
Large Language Model:
A Large Language Model (LLM) is a type of AI model designed to understand, generate, and manipulate human language at a sophisticated level. These models are trained on vast amounts of text data and use deep learning techniques to predict and generate text based on input prompts. For example, Meta’s Llama, Open AI’s GPT or Anthropic’s Claude.
Open Source AI:
Refers to AI software whose weights are released publicly with a permissive licence. This allows developers and organisations to collaborate on AI projects, contribute improvements, and tailor the technology to specific needs. For example, Meta’s Llama models.
Closed Source AI:
In contrast, Closed Source AI refers to models whose source code and inner workings are kept proprietary and confidential by the developer. Users typically can only access these AI systems through specific APIs or platforms, without the ability to view, modify, or redistribute the underlying code. For example, Open AI’s GPT or Anthropic’s Claude.
How can AI support the UK’s public sector?
The UK’s public services are under immense pressure. The country faces a challenging fiscal situation and public sector productivity has struggled to improve over the last twenty-five years.2
Figure 1. ONS Measure of Public Service Productivity (Input Productivity)
There is a tremendous opportunity for the UK to transform our public services by using AI to automate and/or improve the efficiency of complex but repetitive tasks. Through personalised interactions, predictive analytics, and real-time decision support, AI can help public sector organisations provide more responsive, accurate, and tailored services to citizens.
The Alan Turing Institute estimates that AI could help to automate 84% of repetitive transactions across 200 government services.3 Meanwhile, research by Public First estimates that greater use of AI to support routine activities and administration in the public sector could create over £12 billion in savings for the public sector by 2030. This could then rise to £17 billion by 2035 – enough to pay the salary for over 330,000 additional nurses.4
At present, public sector use cases for AI cluster into three distinct categories:
1. Processing and Analysing information
AI systems are adept at processing and analysing extensive datasets. This capability includes executing tasks such as data aggregation from multiple sources, identifying patterns through advanced algorithms, and detecting anomalies using statistical or machine learning models.
For example, processing large data sets from traffic sensors to predict congestion and detect anomalies, in order to optimise traffic management in real-time and contribute to longer-term infrastructure planning.
2. Automating Repetitive Tasks
AI can be programmed to execute and automate repetitive and time-intensive tasks by employing rule-based systems or machine learning algorithms. This automation encompasses a variety of operations, such as automated document processing, workflow management, and routine decision-making.
For example, recording notes from a social care visit so that the social worker is free to focus on engaging with service users – and then generating a draft report based on findings from the visit afterwards.
3. Providing Intelligent Assistance
AI can function as an intelligent virtual assistant or advisor, leveraging natural language processing and decision-support algorithms to assist users. As AI systems evolve, “AI Agents” could be designed to autonomously carry out complex tasks, make decisions in real-time, and adapt to changing environments based on predefined goals, reinforcement learning, or continuous feedback loops.
For example, supporting patient triage when emergency calls are made, then directing ambulance drivers to the local hospital that is best equipped to respond to a patient’s specific injury or illness.
If deployed successfully, the UK’s public sector is set to benefit from:
Reduced Bureaucracy
AI technologies can streamline information processing and automate repetitive tasks, thereby enhancing efficiency. Public sector employees would be able to focus on more complex and strategic work, whereas citizens would benefit from faster delivery.
Better Informed Policy Making
AI can analyse large datasets to identify patterns and anomalies, providing critical insights that inform policy-making, budgeting, and resource allocation. This predictive power could support more targeted, and more proactive interventions across various public services – reducing waste, and increasing the chances of success.
Improved Citizen Engagement
AI-powered chatbots and virtual assistants could provide round the clock support to citizens, answering queries and offering personalised guidance through government processes. This could ensure public services are much more responsive, and reduce the workload associated with “failure demand”.
More Inclusive Public Services
AI tools such as speech recognition, translation services and assistive technologies can help make public services more accessible to people with disabilities, non-native speakers, and other vulnerable groups. This ensures that public services are inclusive and equitable.
Direct and Indirect Cost Savings
Though AI technologies require ongoing investment, the long-term cost-savings can be substantial. AI-enabled efficiencies help to reduce costs, whereas AI-powered insights can also directly identify instances of fraud or financial mismanagement.
When we implemented AI in a London Borough Council's contact centre, we focused on two simple but effective changes: enabling self-service via phone and allowing 24/7 access to services. It was the first UK local authority to use voice automation and AI at scale, and these changes significantly reduced the workload on the contact centre. The project has delivered a 5:1 return on investment over three years.
Andy Theedom,
Director at PwC UK
Imagine AI chatbots guiding citizens through applications and even filling out forms for them, reducing failed applications due to technicalities. This could save time and resources for both citizens and government workers, potentially realising substantial productivity gains and improving service delivery.
Jonathan Bright,
Head of AI for Public Services, Alan Turing Institute
Case Study:
Incubator for AI’s Legislation Drafting Tool
The Incubator for AI has recently unveiled Lex, an open-source AI tool designed to improve the process of drafting legislation. Developed in collaboration with the Ministry of Justice, the Government Legal Department, and The National Archive, Lex aims to streamline the complex task of drafting and navigating legislation.
Lex has an advanced semantic search capability, retrieving relevant legal documents with unprecedented accuracy. Complementing this, Lex’s AI-assisted drafting tool can generate explanatory notes for government bills and can also aid in composing the legislative text itself.
To ensure Lex captures the nuances of UK-specific legal terminology, the team has developed specialised open embedding models. These models, along with Lex’s entire codebase, have been open-sourced, inviting collaboration and fostering innovation in the legal tech sector.
Shaped by user research with the Office of the Parliamentary Counsel, Lex has a user-friendly interface tailored to real-world needs. Currently in the prototype stage, plans are underway to expand testing to a broader base within the Civil Service.5
Case Study:
Incubator for AI’s Consultation Analyser
The Incubator for Artificial Intelligence (i.AI) and the No10 data science team (10DS) have developed an AI-powered tool, “Consult”. The tool hopes to streamline the processing of the 700-800 public consultations conducted by the UK government each year.
Using topic modelling and a secure large language model, Consult automatically extracts themes from consultation responses and presents them on interactive dashboards. This allows policymakers to explore data efficiently while maintaining transparency by linking themes to raw responses.
Initially piloted with the Department of Health and Social Care, the tool is now being tested across various government departments. There is a long term ambition for the tool to be available open source on GitHub, fostering further development and scrutiny from the tech community.6
Case Study:
The Intelligent Automation Garage
The Department for Work and Pensions (DWP) serves over 20 million claimants every year. In 2017, it faced a projected increase of 210 million transactions. To address this challenge, DWP partnered with Accenture to develop the Intelligent Automation Garage (IAG), an innovative automation platform designed to handle massive scale while interfacing with legacy systems.
The IAG functions as DWP’s automation centre of excellence, thoroughly scoping problems, architecting solutions, and building scalable automations. It maintains a fully supported in-house service, allowing for continuous improvement as new technologies emerge.
Key achievements of the IAG include:
- Processing over 20 million items and cases;
- Implementing 65 scaled and transitioned automations;
- Saving more than 2 million operational hours.
This automation initiative has allowed DWP staff to focus on complex decision-making tasks rather than repetitive work, improving both efficiency and accuracy in benefit processing. The system’s ability to scale instantly to meet demand while also downscaling when not needed has been crucial to its success.7
Recent government activity:
In 2018
The UK Data Protection Act was passed, implementing GDPR rules that require AI systems processing personal data to adhere to strict standards of transparency, fairness and accountability. The Act mandates that AI systems must be designed to respect privacy and avoid bias8
In 2021
The UK’s National AI Strategy was published, serving as a framework for all AI development. It emphasises fostering innovation, investment, and ensuring the equitable distribution of AI’s benefits as major priorities. The strategy highlights AI’s potential to transform public services, improve decision-making processes, and enhance efficiency, while taking account of sector specific ethical considerations that will be necessary.9
In 2023
The Incubator for AI (i.AI) was established in the Cabinet Office. Its team of technical experts are charged with creating new AI tools for the government and implementing them, often taking an open source approach to development.10
The AI Safety Institute (AISI) was announced during the UK’s global AI Safety Summit. Chaired by Ian Hogarth CBE, AISI is working to test advanced AI systems and inform policymakers about their risks, as well as foster collaboration across the wider AI community.
In 2024
Digital, data and AI delivery across central government – including the Government Digital Service, Central Digital and Data Office and i.AI – were consolidated under one roof in the Department for Science, Innovation and Technology.
The Secretary of State for the Department for Science, Innovation and Technology, Peter Kyle, commissioned an AI Opportunities Action Plan from Matt Clifford, who will also chair a new AI Opportunities Unit to implement its findings. The goal is to better use AI to spur economic growth and improve public services.11
What role can Open Source AI play?
In broad terms, Open Source AI refers to AI software whose weights are released publicly with a permissive licence. In many instances, this approach allows for greater collaboration and customisation, because a wider community of developers and researchers can contribute to AI algorithms, models and applications.
Particular benefits associated with Open Source AI include:
Increased Transparency
Open Source AI allows a broad community to directly analyse and fine-tune the model. This visibility enables thorough scrutiny of algorithms, data processing, and decision-making architecture, which is then crucial for building trust in AI applications used in the public sector. This collective oversight facilitates quick identification and resolution of potential vulnerabilities, reducing the risk of undetected flaws.
Greater Control of Data
Open Source AI models offer significantly more control over data management protocols. These models can be hosted on an organisation’s own infrastructure, giving public sector bodies greater autonomy over (potentially sensitive) data sets. This can also help to better ensure compliance with data security regulations and standards.
Enhanced Flexibility
Open Source AI models can be modified and extended to address specific challenges. This adaptability enables the creation of bespoke AI applications that align precisely with public sector requirements. The ability to repurpose existing models also accelerates innovation, allowing agencies to respond quickly to changing circumstances without being constrained by off-the-shelf solutions or a single proprietary provider.
Reduced Costs
Open Source AI allows organisations to choose the hosting solution that meets their needs and budgets. Smaller models can run on an organisation’s in-house infrastructure, providing predictable and limited costs. The output cost — the expense associated with generating AI responses — for smaller models such as Meta’s Llama 3.1 8B are up to 73% lower than OpenAI’s GPT-4o mini.12 For larger models, cloud hosting and API providers offer pay-as-you-go pricing that are responsive to fluctuations in workload. For example, the output costs of Llama 3.1 405B are around 10% lower than GPT-4o and up to 40% lower than Anthropic’s Claude 3.5 Sonnet.13
These attributes are particularly useful in a public sector context. For example, NHS Trusts are likely to benefit from the flexibility, accountability and affordability associated with Open Source AI. The model can be fine-tuned with their specific data, and code can be audited by the community to ensure the models are safe, reliable and free of hidden biases. Meanwhile, NHS Trusts would be able to maintain control over sensitive patient data, without reliance on external proprietary systems.
Of course, Open Source will not necessarily be appropriate in every use case. There will be occasions where the public sector will be more comfortable with a closed approach, i.e. in specific use cases where transparency is perceived to create vulnerabilities that bad actions could try to exploit. Equally, there may be instances where a “plug and play” proprietary tool will be sufficient for quick adoption – particularly if there is a lack of in-house expertise to support ongoing maintenance.
But ultimately, organisations will want to pick the model that meets their particular needs best; and the best outcomes will be delivered through a diverse and competitive ecosystem of options that the government can leverage at the appropriate moment. The choice may not always be binary: hybrid approaches may often suit complex public sector needs. This will necessitate a firm understanding of the different approaches available to decision makers, and consistent evaluation of long-term goals when selecting AI solutions for public services.
Opening AI offers significant advantages for the public sector. It allows for transparency, collaboration, and shared innovation across government departments and even nations. It can also help avoid vendor lock-in and promote interoperability, which are key concerns for public sector IT. The open nature of these solutions allows for better scrutiny and trust-building with citizens.
Amanda Brock,
CEO, OpenUK
With a government entity, you’ve got a lot of confidential data. You don’t want to have to share that data with external providers, rather, you want the option to download a model onto your own local hardware. That’s why open source models can work so well in a government context, because you can use your data and highly confidential government information, on your own systems.
Farzana Dudhwala,
Global AI Policy & Governance Lead at Meta
As much as possible, we should maintain principles of interoperability - avoiding some of the challenges that we've made in the past with legacy public sector IT contracts where we get locked in to a particular software or provider. You can still have proprietary AI solutions in the right contexts, but the government should consider where more open and interoperable approaches might be more appropriate.
Theo Betram,
Director at Social Market Foundation
When you put something out there under the scrutiny of all, you get the collective wisdom of experts around the world working to improve it. An Open Source approach also helps to democratise access to this technology for a whole range of use cases. It’s a great way to see what other people can create – from better beekeeping practices to helping women access maternity advice in their native language.
Farzana Dudhwala,
Global AI Policy & Governance Lead at Meta
Case Study:
Albert
Developed internally by the French government, Albert is an Open Source AI tool designed to modernise and streamline public services across France. Created within the Etalab department of DINUM, Albert is hosted on French government infrastructure and developed using Llama 2, with plans to incorporate Llama 3.
Albert aims to assist public officials in various departments by:
- Processing tax queries (approximately 16 million annually)
- Pre-processing environmental project submissions
- Assisting with legal proceedings and transcriptions
- Automating medical reports
- Aiding in forest fire detection
- Supporting civil servant management
Albert will handle routine tasks, freeing up officials to focus on citizen interactions. The AI has already shown promising results, detecting 140,000 cases of tax fraud and recovering €40 million in revenue for local authorities.14 15
This follows on from other successful AI projects by the French Government, like “LLaMandement” an AI tool built on Llama-2, designed to summarise lengthy legislative proposals in more readable neutral summaries. This tool has already allowed French government officials and Ministers to work more effectively and saved the hours of human time it would take to generate such summaries.16
What level of AI adoption has been achieved?
AI is permeating daily life for many UK citizens. One in ten of the general population report using an LLM-based chatbot regularly – rising to one in four amongst 18-34 year olds. However, we have not yet reached the stage where most of the population consciously and consistently uses AI-powered tools.17
AI adoption in the public sector tells a similar story. The National Audit office reports that AI has been deployed by government bodies in 74 use cases, representing 37% of the bodies that responded to its latest survey. Just 21% of government bodies had a strategy for AI in their organisation, with 61% still in the planning phase.18 At a grassroots level, around 22% of public service professionals claim to actively use a generative AI system at work.19
This hints that there are pockets of AI adoption across our public services; but efforts must be made to establish a more systematic approach that fully leverages this technology. Furthermore, as this technology is so new, we develop better measures to understand the relative cost saving and productivity it brings, so that we can quantify its benefits in the long term.
A lot of departments and bodies are well established and mature in using machine learning to support their activity. But we’re earlier in the journey when it comes to generative AI and it’s not as common. We want to find that sweet spot where we don’t go too fast or too slow.
Aran Uppal,
Lead Technology, AI at CDDO
ChatGPT helped people, the general public and the government, understand the possibilities presented by AI and get excited about harnessing them. There is now a critical mass of people with the right skills and enthusiasm in government, they need people in leadership positions backing them and making this a priority.
Mallory Durran,
Group Director of Data Science at NESTA
When implementing AI, we have to be careful that we aren’t just creating more work for civil servants in the process. People on the frontline have to trust that these tools will make their lives easier. The key is to consciously use the time saved by AI for more valuable work, rather than just increasing bureaucratic requirements.
Jonathan Bright,
Head of AI for Public Services, The Alan Turing Institute
Case Study:
NHS AI Lab
The NHS Artificial Intelligence Lab brings together government and health care providers with academics and tech companies to harness the benefits of AI “safely and ethically at scale”.20
The Lab runs a range of programs including its AI Diagnostic Fund which aims to accelerate the adoption of cutting edge AI imaging tools that can help to diagnose patients more quickly. To help develop these technologies, the fund is providing £21 million through “imaging networks” made up of 64 trusts across England.21
Alongside trialling the AI Deployment Platform, a “store” for AI medical imaging tech, specifically designed for radiology workflows,22 the AI Lab also manages the AI in Health and Care Award, which distributes £123 million to help accelerate promising new AI tools including: 23
- Healthy.io’s smartphone albuminuria self-testing: using a home test kit and smartphone app, users can self test for kidney conditions.24
- Skin Analytics Ltd’s DERM: uses AI to analyse skin lesions, detecting which are cancerous, pre-cancerous and benign.25
- DrDoctor: A platform for patient engagement, using AI to predict those less likely to attend appointments and targeting customised communications at them to encourage attendance.26
What is holding adoption and innovation back?
Despite the wealth of opportunities offered by AI, significant barriers continue to prevent widespread adoption. This is true across various public sector departments and agencies. Conversations with industry stakeholders emphasised the following:
Data Quality and Integration
AI adoption is significantly limited by the quality and accessibility of public sector data. While AI experts emphasise that many AI programs can work with imperfect data, legitimate concerns remain regarding data silos and incompatible data formats that will inhibit wide-scale adoption across the public sector.
Regulatory & Ethical Challenges
The public sector must navigate complex regulatory and ethical considerations when deploying AI. Concerns over bias, decision-making transparency, and ensuring that AI applications align with public interest can slow down or prevent adoption. Similarly, handling personal and institutional data with AI systems raises significant privacy and security risks. Sensitive information, such as personal or health data, requires stringent protection, with minimal tolerance for mistakes.
Public Perception and Trust
On a related theme, it’s essential that AI adoption within the public sector is accompanied by broad public acceptance of the technology. Overcoming public discomfort or uncertainty requires transparent model design as well as robust safeguards and clear human accountability.
Budget Constraints
AI implementation requires upfront investment in technology, infrastructure, and skilled personnel – as well as a commitment to ongoing maintenance and improvement. Whilst there can be a significant return on investment over time, the UK’s public sector is facing significant financial pressures, limiting its ability to allocate sufficient resources for longer-term AI initiatives.
Lack of Technical Expertise
Most AI technologies require skilled individuals to source, implement, and maintain AI tools in a systematic way. Whilst there are very competent technologists working in government, the public sector fundamentally struggles to compete with the private sector when it comes to securing top talent.
Poor AI Literacy
There is also a lack of confidence across the broader public sector workforce – including many senior decision makers. Whilst not everyone needs a highly technical understanding of AI models, foundational knowledge is necessary to sensibly identify opportunities and make informed decisions about long-term deployment strategies.
Legacy Infrastructure
Public sector organisations often operate with outdated infrastructure, which can result in insufficient access to the high-performance computing resources needed for rapid iteration and testing. Unlike the private sector, which may have more flexibility to invest in scalable cloud computing services or cutting-edge hardware, the public sector may struggle with resource constraints that make it difficult to experiment with and refine AI models and other complex software.
Fear of Failure
Unlike the private sector, public sector organisations can find it harder to embrace iterative deployment of new tools and services – particularly for those that are public facing. This hesitancy stems from accountability pressures, fixed budget cycles, and public expectations of flawless service delivery from the outset. The fear of public scrutiny and potential backlash over ‘failed’ experiments creates a risk-averse culture that clashes with the inherently iterative nature of AI innovation.
We’re constrained by poor data infrastructure. Like any model, when working with AI, it's garbage in, garbage out. You need high quality data to build great tools. There are basic applications of AI we can and should crack on with now – but if we also spend the next few years getting our digital transformation right, it is hard to imagine what the world might look like.
Mallory Durran,
Group Director of Data Science at NESTA
If you’re a health manager, for example, and the choice is between doing something which you fully understand and fully believe is a good idea in the long run - i.e. implementing AI -, and your obligation in the short term to fulfil your critical services that are under immense pressure - i.e. keeping your ward running safely - there’s a fundamental problem. You’re inevitably going to cannibalise long-term budgets to meet short-term urgency.
Theo Betram,
Director at Social Market Foundation
Recommendations
Conscious change is required across the UK’s public sector organisations to ensure that the barriers to AI adoption can be overcome.
Make AI adoption a national endeavour.
The public sector will benefit from clear, consistent leadership from senior government officials regarding AI adoption and innovation. Public sector workers at all levels need the confidence to make long-term investments in the capacity, infrastructure and skills required for AI adoption – and this means a long-term drum beat of support from the most senior policy makers. Much like the government has already done with its Missions, making explicit that AI adoption is priority for the UK across the whole of Government is crucial to its acceleration.
Introduce regular skills training for senior decision makers.
Whilst technical experts are needed to successfully deploy this technology, we can only take meaningful strides if senior decision makers understand the basics. As a starting point, all government ministers and permanent secretaries should therefore receive formal training on the practical principles underlying AI, irrespective of their policy brief.
Deploy the new AI Opportunities Unit as an in-house Consultancy.
Government would benefit from a centralised “AI Implementation Unit” that houses a “pool” of experts. This team could sit within the recently announced AI Opportunities Unit, and consult for other government departments to support the deployment of new AI tools. This would help to mitigate any skills shortages and ensure best practice is shared across projects.
Conduct a mandatory audit of activity every six months.
Information about public sector pilots is extremely limited. Much of the quantitative data regarding AI penetration is supplied by non-mandatory surveys. DSIT should therefore facilitate a more reliable and regular way of collecting usage data about this fast-moving technology. Case studies derived from this audit should then be made readily accessible for others to learn from.
Expand the AI Safety Institute’s remit to support testing and piloting of AI products.
The UK AI Safety Institute (AISI) is already building the infrastructure to test the safety of advanced AI, and to measure its impacts on people and society. As a next step, the UK AISI could usefully establish a “public sector sandbox” that serves as a secure environment where departments and other public sector bodies can test data and AI models.
Create a procurement process fit for the digital age.
Standard government procurement processes still take too long and are too difficult for innovative businesses to access. Government must prioritise procurement reform – perhaps leveraging AI tools as a means of reducing red tape – to allow new and emerging technologies into the public sector more quickly. This should ultimately help to drive a culture of innovation and experimentation across government.
Take advantage of Open Source approaches where appropriate.
The government already has a significant legacy of embracing open data and open innovation. Working with a range of AI experts, DSIT should compile formal guidance that explains the benefits and drawbacks of open vs closed approaches to AI. This should acknowledge that “openness” exists on a spectrum, ranging from basic technical information to entire models, including underlying weights, training data and the code used to run them. Whilst different use cases will necessitate different approaches, it’s important that public services are able to make informed choices to best meet their needs.
Deploy a consistent communication strategy with the British public.
It’s essential that AI adoption within the public sector is accompanied by broad public acceptance of the technology. This will require a concerted effort to communicate clearly with the public about how, where and why AI is being deployed – being honest about its limitations, but enthusiastic about its potential.
Don’t delay.
Full-scale transformation won’t happen overnight. We need a long-term roadmap to fulsome AI adoption across all facets of public service delivery, that will tackle infrastructure, interoperability and investment challenges. But, in the interim, public sector bodies should still continue to implement AI solutions in smaller, manageable projects. This “start small, scale fast” approach will allow organisations to gain experience, build confidence, and demonstrate value before larger-scale deployments are ready to unfold.
I always emphasise the importance of starting with practical, achievable goals rather than waiting for perfect conditions. There’s already significant potential for existing AI tools to improve public services - for example, triaging and prioritising casework. We shouldn't plan out a perfect software program for waterfall delivery. Instead, we should build the smallest thing possible, then iterate and improve it. This approach allows us to make progress and learn, rather than getting stuck in endless planning and preparation.
Paul Maltby,
Director of Public Services at Faculty AI
As a data scientist, you start with a project and work towards what is viable. You have to work slowly towards a final goal and iterate along the way. There’s a sequence of learning and development that has to happen, and you’re not necessarily going to get it right straight away. So AI might give the illusion that it’s more intelligent than it is, and that will set expectations very high - which means we need to ensure we manage expectations appropriately.
Aran Uppal,
Lead Technology, AI at CDDO
The key to successful AI implementation is fostering curiosity and a willingness to learn within the organisation. We always look for champions who are eager to adopt new ways of working, as this allows for the transfer of expertise from external parties back into the organisation. It's not about specific technologies, but about cultivating a new set of skills and an appetite for learning that will drive the development of the public sector.
Andy Theedom,
Director at PwC
Building trust is crucial when implementing AI in the public sector. That's why we've developed Camden’s Data Charter, which is essentially a social contract between us and our residents. We spent months working with a diverse group of citizens, explaining complex data concepts and being transparent about both the benefits and risks of AI. This deliberative democracy approach has been key to gaining public trust and support for our initiatives - it’s also allowed us to be more ambitious with our AI plans.
We're now collaborating with other organisations, sharing our best practices with the United Nations, the Commission for Equality and Human Rights, and the Alan Turing Institute. By working together and being open about our processes, we're able to push the AI agenda forward in a way that's responsible and community-focused.
Tariq Khan,
Chief Digital and information Officer,
London Borough of Camden
Acknowledgements
Public First has worked with Meta and techUK to convene the experts in AI & the public sector. This factsheet reflects Public First’s independent research findings from a roundtable and multiple expert interviews as well as opinion polling. We like to thank the following individuals for their contributions:
Alice Cambpell, Head of Public Affairs at techUK
Amanda Brock, CEO of OpenUK
Andy Theedom, Director at PwC
Aran Uppal, Lead Technology, AI at CDDO
Farzana Dudhwala, Global AI Policy & Governance Lead at Meta
Jonathan Bright, Head of AI for Public Services, Alan Turing Institute
Mallory Durran, Group Director of Data Science at NESTA
Paul Maltby, Director of Public Services at Faculty AI
Tariq Khan, Chief Digital and information Officer, London Borough of Camden
Theo Betram, Director at Social Market Foundation