The Bank's artificial intelligence (AI) strategy

Executive summary

This strategy sets out the future direction of the application of artificial intelligence (AI) internally at the Bank and elaborates on how we plan to fulfil the AI-focused mission that forms part of the Bank’s overall data and analytics (D&A) strategy. We have created this standalone strategy so our staff and the public have a clear understanding of the what the future of AI will look like at the Bank, and how this aligns with the Bank’s overall objective to promote the good of the people of the United Kingdom by maintaining monetary and financial stability. It does not discuss the Bank’s work on understanding the wider implications of AI on our regulatory responsibilities and our monitoring of developments in the financial system and broader macroeconomy.

Our AI strategy emphasises the importance of making AI tools and services accessible to all our staff, and of ensuring that the establishment of the Bank’s Enterprise Data Platform, on the cloud, allows our staff to fully engage with the Bank’s data using AI tools. At the core of our strategy is a mission statement, which describes our overall objective. There is an accompanying vision that sets out our long-term aspiration to drive AI growth at the Bank. Alongside this are a set of key goals and priorities. Our goals focus on what we are aiming to achieve and centre on productivity, experimentation, responsible use of AI, collaboration and learning, and supporting our staff. Our priorities set out in more detail the actions we will take to achieve our goals.

In our strategy, we address the importance of having key foundations in place to ensure we are being transparent and so that everyone understands what is expected of them in their role. These foundations include:

• having strong governance controls in place, so our staff understand how to use AI tools and services safely and responsibly;
• creating a robust AI skills curriculum so staff understand the opportunities and limitations of AI and have the appropriate level of AI fluency; and
• building partnerships with other central banks, government bodies and private companies to share knowledge and best practices in AI.

This is a new and exciting time for the Bank and we aim to use this AI strategy as a springboard to drive innovation, growth and change at the Bank, as well as provide clear direction and define what success looks like. In due course, we aim to create a strategy implementation plan in close consultation with the business. We will be looking towards everyone at the Bank to get actively involved to support the design, development and implementation of the AI strategy. We believe AI has the potential to have a profound impact on how we work at the Bank going forward, particularly in finding solutions to problems that were previously difficult or impossible to solve.

We are also conscious that AI is a rapidly changing field and, as such, this strategy will need to be kept under regular review to keep up to date with a changing world. We will review at least annually to make sure it remains fit for purpose.

What is artificial intelligence?

Artificial intelligence is a set of algorithms that allow machines to perform tasks that traditionally required human intelligence. Machines, or AI systems, can perceive their environment and use learning to achieve defined goals. Many AI systems generate predictions or outputs that can be used to automate or augment output that would typically require human understanding. Common applications of AI systems in everyday life include advanced web search engines, recommendation systems (eg through streaming media services) and generative tools (eg AI assistants).

AI is an umbrella term encompassing various sub-disciplines such as machine learning (ML), deep learning (DL) and generative AI (GenAI). Much of AI is built upon machine learning, a set of techniques that enable computers to learn from and make inferences based on data. Machine learning models are typically trained on data to recognise certain patterns. A subset of machine learning is deep learning which uses neural networks, with multiple layers of processing, to analyse data. Deep learning models are particularly good at using unstructured data such as text and images to do complex tasks such as speech and image recognition. More recently, a capability within deep learning has emerged called GenAI. GenAI creates new or original content in the form of text, images, videos or other data types, often in response to prompts. GenAI tools can produce output that closely mimics human-like thought and reasoning (though it doesn’t ‘think’ or ‘reason’ in the same way a human does). Advances in computing power and data availability have made generative AI possible and now tools such as ChatGPT are easily accessible to all.

Machine learning, deep learning and GenAI represent the interrelated landscape of artificial intelligence, as shown in Figure 1 below, with each contributing to the development of systems that can independently perform an increasingly wide range of complex tasks.

This strategy is not focused on the most wide-ranging definition of AI, therefore traditional econometric or statistical models will not be considered as AI in this context. 

How can AI help us at the Bank?

In recent years, we have developed some AI capabilities at the Bank that have allowed us to work more efficiently. For example, in the Prudential Regulation Authority (PRA), we have pioneered the application of machine learning and deep learning models to categorise and extract unstructured information, and therefore allow reliable querying of management information, board packs and disclosures we receive from the firms we supervise. We are also currently piloting off-the-shelf GenAI tools to support our staff with tasks such as summarising meetings and documents, and coding assistance.

However, the range of activities that AI can do has expanded rapidly over the past few years, with newer and more sophisticated AI tools and services becoming more accessible. We recognise the need to capitalise on this and utilise such AI tools and services to solve our most complex business problems. AI has the potential to have a profound impact on how the Bank works going forward, particularly in finding solutions to problems that were previously difficult or impossible to solve.

Our aim is to make AI tools and services easily accessible to all our staff, regardless of their role. All staff can access MS Copilot to support their work and integrated Copilot tools, for use with MS Office and for coding, are available to all staff, cleared to work with sensitive data. An Enterprise Data Platform is currently being built on the cloud and this will allow staff to fully engage the Bank’s data with AI tools, to obtain insights from it more effectively and efficiently. For all staff, access to new and emerging AI assistants will mean improved productivity and less need to carry out administrative work. For data professionals, the ability to experiment with advanced AI tools and services will provide opportunities to develop and create bespoke AI products to answer the most pertinent business questions, as well as automate routine and repetitive tasks.

AI principles

To support a consistent and coherent approach to AI across the Bank, our AI strategy leans on a set of principles that set the tone for how we approach AI. The principles are embodied in the ‘TRUSTED’ framework, outlined in the Bank’s Chief Data Officer, James Benford’s speech TRUSTED AI: Ethical, safe, and effective application of artificial intelligence at the Bank of England.

Mission and vision

Our mission: To enable the safe, ethical and effective use of AI in the Bank.

Our vision: We want everyone, regardless of their role in the Bank, to have the opportunity to use AI tools and services to empower them to excel in their roles. We will enable our technical staff to access AI tools and services to create innovative and ethical AI products for the Bank, and we will foster a culture where all staff understand how to use AI tools effectively and responsibly.

Goals and priorities

Starting from our mission to enable safe, ethical and effective use of AI in the Bank, we have developed five key goals that describe what we are aiming to achieve and how we will fulfil our mission. Alongside these are our priorities which set out in more detail the actions we will take to achieve our goals.

Governance, ethics and compliance

At the Bank, we have established an AI governance committee which is co-chaired by the Chief Data Officer and the Chief Information Officer. The committee is responsible for:

• developing and running the Bank’s AI governance framework;
• identifying and managing potential risk areas and ensuring there is a mechanism for staff to flag these;
• developing policies and guidelines on AI implementation;
• providing guidance to business areas in relation to AI ethics and safety; and
• ensuring clear communications to disseminate new policies and guidelines.

We have developed a robust AI governance framework to promote consistent practices Bank-wide. It delineates the differing responsibilities for different types of staff. For example, people managers have different responsibilities compared to colleagues who are not managers. Additionally, the responsibilities will differ between non-technical staff, who may predominantly use off-the-shelf AI tools, and technical staff. We will ensure all staff using AI tools and creating AI products take responsibility for managing risks, as set out in the AI governance framework. Alongside this, we have developed an AI ethics framework which will provide a clear set of foundational principles to guide staff on how to work and innovate with AI in an ethical manner. Given the pace at which AI is advancing, it will be paramount to regularly revisit and, where necessary, revise our AI governance and AI ethics frameworks to ensure they remain fit for purpose.

Talent and skills

The effective use of AI requires investment in our talent. We need an AI skills curriculum to help staff in all roles understand the opportunities and limitations of AI and have the appropriate level of AI fluency. We are committed to investing in our staff, so the curriculum will include a range of upskilling, learning and development opportunities for all colleagues at the Bank, and will be focused on their needs and the requirements of their job. For data professionals, an enhanced offering will be available, and we plan to provide bespoke training and opportunities for self-study where appropriate. By adopting a flexible approach, staff will have more autonomy to shape their own training and development needs around the latest and most relevant AI tools and trends. We are keen to empower staff in their role so they can support the Bank to advance its AI capabilities effectively. The AI governance committee has responsibility for maintaining the AI skills curriculum.

We aim to attract the best and brightest AI talent to the Bank. To achieve this, we will clearly communicate how the AI strategy complements our D&A and cloud strategies and highlight the excellent learning and development opportunities available. We want potential recruits to see the Bank as a great organisation to grow and develop their careers in AI.

Links with the Bank’s wider data and analytics strategy

AI has the potential to transform our ability to use data effectively. But we require the right fundamentals of data governance and management, and a computing environment that can make the most of our use of AI. Delivering this is an important part of the Bank’s investment in data and IT. This is informed by a unified technology and data architecture which brings all parts of the organisation together.

The Bank has published its D&A strategy. Our investment in AI forms part of this and will be supported by the overall strategy. In particular, there are two foundations: the establishment of an Enterprise Data Platform and investment in training and culture, as discussed above. But the development of improved governance, including metadata, and consistent data management are crucial to allow the Bank to unlock AI’s potential. Information and records management is particularly important, especially in light of generative AI’s rapid improvements in capacity to process text and other forms of unstructured data.

AI use case prioritisation

We need to effectively identify and prioritise high-value AI use cases and initiatives, and build a strong portfolio of work across the Bank. To do this, we will adopt a federated approach, where individual business areas will have the freedom to define and organise their work, but with central support and controls where it facilitates the safe and effective development of AI services.

The selection of AI projects that require particular central support will be based on a robust and transparent process, and on criteria including value-add, impact, feasibility, complexity, resources, and alignment with Bank objectives. Having a standardised approach to selecting AI projects will ensure we are being transparent and can deliver a wide range of new and innovative AI products. We will also create a central repository of all AI projects enabling the sharing of experience and best practices across the Bank.

We will first focus on building a Bank-wide portfolio of work incrementally and identifying the capabilities needed to support this (eg people, governance), as well as demonstrating business value. We will adapt our approach as these supporting capabilities mature and develop over time.

Measurement and improvement

We will define an approach to identifying the benefits from our investment in AI. This will measure our progress in delivering the AI strategy and develop a clear plan to make improvements where needed. Similar to our approach for the D&A strategy, we will design and develop a strategy implementation plan which will set out in more detail how we plan to identify and measure benefits as well as track and evaluate progress in achieving our goals. We will use both quantitative (eg KPIs) and qualitative measures (eg staff surveys) to assess and track our performance. To promote openness and transparency, we will report our progress to the Bank’s Data and Analytics Board, an internal executive level group, who are responsible for the implementation of the overall Bank-wide D&A strategy.

Partnerships and collaboration

The Bank benefits greatly from partnering and working with other central banks, academia, government bodies and private companies to stay in touch with what they are doing in the AI space. We are keen to work with a variety of organisations, so we are exposed to a diverse range of thoughts, experiences and ideas. Where appropriate, we will partner with external organisations to progress our AI work.

Ways of working

Our aim is to adapt how we work at the Bank and deliver AI products that have high business value; this will help us move our AI agenda forward at pace. To achieve this, we will continue to encourage multidisciplinary collaboration across the lifecycle of AI projects. Furthermore, strengthening our approach to how we scale AI projects, from proof of concept to production, will mean more AI products can be used widely across the Bank. Alongside this, defining our approach to how we will build a culture of experimentation means staff will be able to safely trial AI tools and services, and share their learnings Bank-wide.

Our colleagues are at the heart of all the change we are working towards. We want them to be fully sighted and on board with the strategy and any changes being proposed, and will work towards creating a coherent communications plan so understanding, usage and adoption of AI tools and services is successful across the Bank.

Horizon scanning

The power and range of AI tools are increasing rapidly. Keeping pace with AI trends, tools and services is key to ensure we stay up to date as an organisation; we need to actively incorporate changes into our work whilst also keeping a firm eye on future risks. In the PRA specifically, we will need to monitor and assess how the firms we regulate and supervise are utilising AI and make sure supervisors keep on top of changes, as well as understand how this impacts their work. We aim to design and develop an internal framework to be confident that the firms we supervise are adhering to high standards with their use of AI, and so we are clear on how to supervise these firms effectively with our objectives in mind.

More broadly, we will aim to keep abreast of technological advancements, thinking beyond AI, focusing our attention on transformative trends such as quantum computing. The Bank needs to understand how the use of AI in the broader economy will affect our regulatory responsibilities and how it will change the working of the macro economy and financial system. This type of horizon scanning is underway but is not discussed in this strategy which is focused on the application of AI within the organisation.

Operating model

AI is likely to affect all areas of the Bank. We want to ensure that AI tools are well-suited to business needs and that staff are empowered to develop solutions to the challenges they face. To achieve this, we will operate a federated model where as much autonomy as possible is devolved to local business areas. The clear governance, ethics and compliance framework mentioned above should support local areas’ ability to innovate safely and effectively.

The use of AI needs to be consistent with the Bank’s wider technology and data architecture, and there will be areas where synergies can be achieved by strengthening our central support functions. This support will take several forms. We will investigate the best way to ensure that expertise is available to all areas of the Bank, perhaps by the establishment of an AI Centre of Excellence. And our Technology directorate will take the lead in ensuring the right technology environment is in place to effectively develop and run AI solutions across the Bank.

Buy versus build

The rapid improvements in AI, particularly generative AI, have led to a wide range of commercially available tools, many of which are highly relevant to the Bank. But there are also situations where more bespoke tools may be required that need to be developed internally. The Bank will take a pragmatic approach to this, using external tools where appropriate but ensuring the in-house expertise to build where that is preferred.

Glossary

For the purpose of this strategy, we have defined the terms as follows: