Written evidence submitted by Apolitical
1. About Apolitical
1.1. Apolitical is a global network and learning platform for government, with a mission to build 21st century governments that work for people and the planet. Apolitical is used by almost a quarter of a million public servants and policymakers in 160 countries to find and share best practice through communities of peers, and to upskill with short online courses on 21st century government priorities. Members range from mayors and ministers to policy pioneers and digital innovators. Apolitical’s courses are co-developed with leading universities and research institutions, including Oxford, the London School of Economics and Georgetown. Our flagship programmes include climate and sustainability, digital and AI, equity, equality and inclusion, and the 21st century state (spanning topics such as agile governance and change management).
1.2. Apolitical is a trusted and committed partner to the UK Civil Service. Our UK community includes more than 50,000 civil servants; we are an approved supplier on multiple learning procurement frameworks; and we have led the design and delivery of strategic capability programmes with UK departments, including the Digital Excellence Programme.
1.3. In July 2023, Apolitical launched the Government AI Campus. The aim of the AI Campus is to rapidly upskill and prepare public servants for the age of AI through a series of engaging online courses, online exchanges, in-person convenings, and a global community of practice. More than 10,000 public servants globally have engaged in learning on the AI Campus since its launch. Apolitical tracks key trends in how public servants are engaging with AI (specifically generative AI) through community polls, surveys, interviews, and interactions on our platform. This submission reflects some of our key insights on public servants and AI.
2. Focus of this submission
2.1. In response to the call for evidence, this submission focuses on:
● Risks and opportunities of AI adoption in government;
● Data and skills issues in government.
2.2. Artificial intelligence (AI) is not new to government, but generative AI is. ChatGPT was released in November 2022 and unleashed a technological revolution seemingly unprecedented in its speed. Apolitical has been monitoring the emergence of generative AI in government closely. In doing so, we have formed two perspectives on this wave of adoption. The first is at the institutional level: How are government organisations integrating generative AI solutions into operational processes and public services? The second is at the individual level: How are individual public servants and small teams using easily accessible tools like ChatGPT, Gemini and, more recently, Microsoft CoPilot in their workflows?
2.3. The government believes that AI can generate billions in productivity benefits for the public sector.[1] That will require AI deployment at scale across the internal machinery of government as well as front-facing public services. The NAO’s review, however, suggests that the future is still some way off, with less than 30% of UK public bodies either currently deploying, or planning to deploy, AI to provide public services directly.[2] In the meantime, our evidence suggests that it is smaller-scale adoption — at the individual or team level — that is driving the fastest experimentation and use of generative AI in government right now. This is where the majority of our evidence will focus.
2.4. Further evidence of our insights into public servants’ use of generative AI is published in an annexe to this document.
3. Risks and opportunities of AI adoption in government
3.1. Opportunity: public servants are (mostly) optimistic about AI
3.1.1. When polled, a majority (56%) of our community members said they are more hopeful and excited about the rise of AI than fearful and nervous (34%) (May 2023).[3] However, attitudes and concerns vary across different groups of public servants based on a variety of factors, potentially including age, national context, and profession.
3.1.1.1. Age could influence AI adoption amongst public servants: To take a particular anecdotal example of this — a community member in the United States who leads the administration of a small city explained to Apolitical that he had noticed a marked generational difference in AI adoption and attitudes.[4] Apolitical is currently researching with our community to understand whether age is a significant factor in generative AI usage. Preliminary data suggests that significant differences in usage could exist across close age groups. For example, early analysis suggests that the proportion of 45-54-year-olds not using the technology is nearly three times as high as 35-44-year-olds. We expect to publish this data in the summer of 2024.
3.1.1.2. Are emerging economies more optimistic about AI than developed nations? National attitudes towards AI could shape adoption in government. A global general population survey from the Oliver Wyman Forum showed, for example, that regular usage of generative AI is twice as high in India and the UAE than in the UK.[5] Apolitical regularly convenes senior government leaders at fora like COP28 and the annual World Governments Summit. Through these contacts, we have anecdotal evidence that attitudes about AI and government are more optimistic in countries across the Global South, notably South Africa, Kenya, the UAE, and India.
3.1.1.3. Some government professions could embrace AI more quickly than others: Recently published analysis by professional services firm Deloitte shows significant variation in the automation potential of different government jobs. They assess that jobs with high levels of creative difficulty, context variability (i.e. how much a task changes in different scenarios), and moderate accuracy are well suited to automation by generative AI. That could include tasks such as speechwriting, summarising laws or writing reports.[6]
3.1.1.3.1. This is the kind of work that those working in policy or strategy roles tend to do. Yet, interestingly, Apolitical polling data suggests that UK policy professionals are not (yet) experimenting with generative AI to support their daily work.
3.1.1.3.2. In September 2023, in collaboration with the Policy Profession Unit, Apolitical hosted a virtual event at the UK Policy Festival through the Government AI Campus. Nearly 1,000 policy professionals registered. When asked, nearly 75% of attendees said they had never used AI in their work (see figure on page 4 below). This was despite official guidance from the Central Digital & Data Office (CDDO) issued months earlier permitting its use for some policy tasks (like summarising documents). That figure is in stark contrast with most other polls and surveys from our community and the wider knowledge-based workforce. More research would be required to understand this low adoption rate. It could be the result of unique circumstances and an abundance of caution among UK policy professionals. Or it may indicate a reluctance to use tools that are perceived as resulting in automation. Previous enquiries have noted the importance of early employee engagement in facilitating the adoption of new technologies in the workplace.[7] As stated elsewhere in this evidence (3.2), if government leaders want civil servants to use generative AI, they will need to support and encourage them to do so. Those who perceive the highest risk as replacement rather than enhancement will involve a careful and collaborative approach to ensure it is a mutually beneficial process.
Data from Apolitical polling, conducted July-September 2023
Samples: 444 (Apolitical community) 620 (UK Policy profession)
3.2. Risk: Leaders without hands-on experience could hamper generative AI adoption.
3.2.1. As departments roll out new solutions to improve operational decision-making and improve internal processes, leaders will need to provide direction and impetus for individuals to use new AI tools.
3.2.2. The growth of generative AI in government has been led by the experimentation of ‘innovators’ and ‘early adopters’ (including individuals and small teams). However, our evidence shows that there is still a significant portion (around 50%) of public servants not using generative AI — in the language of Rogers’ adoption curve: the pragmatists, conservatives, and sceptics.
3.2.3. The NAO review shows that departments are ready to start piloting new AI solutions to improve internal procedures. A ‘Build it and they will use it’ approach, however, may not be sufficient. The example of UK policy professionals shows that even when new tools and technologies are easily available and permitted for use, public servants may be reluctant to use them. If leaders want new technologies to be adopted and new processes implemented, they need to ensure a holistic approach to adoption, which includes building a culture where experimentation is encouraged.
4. Data and skills
4.1. Risk: Government cannot achieve AI adoption at scale without critical enablers, including skills, capability, culture and an “AI-ready” culture.
4.1.1. Apolitical is leading the design and delivery of the CDDO’s flagship digital training programme for senior civil servants, the Digital Excellence Programme (DEP), alongside the London School of Economics and Ernst & Young (EY). The aim of the programme is to help fulfil the government’s commitment to upskill 90% of senior civil servants in digital and data essentials.
4.1.2. The learning from the programme is centred around two interlocking objectives or themes.
4.1.2.1. The first is to help senior civil servants (not working in digital, data or technology roles) understand what ‘digital’ means in the context of 21st century public administration. We apply the expansive definition of ‘digital’, popularised by Tom Loosemore (co-founder of the UK Government Digital Service): ‘Applying the culture, processes, business models, and technologies of the internet era to respond to people's raised expectations.’[8]
4.1.2.2. Based on that understanding of digital being about significantly more than just technology, we encourage all senior civil servants to see themselves as digital leaders, regardless of their profession. Their decisions can impact key issues in digital strategy, like data sharing, procurement, and legacy technology. In our experience, that awareness is (yet) to be mainstream among the senior ranks.
4.1.3. As the NAO’s review makes clear, there is a range of technological, organisational and policy enablers required for AI adoption at scale in government.[9] Skills and capability — particularly among senior leaders — are part of that mix. However, this is not just about AI-specific competencies. It is about the underlying contextual knowledge required to lead in the digital age and facilitate the transformation required to enable AI adoption.
4.1.4. The government’s current long-term digital strategy rightfully stresses the importance of capability building to achieve wider digital aims (including AI adoption). That commitment should be reaffirmed in future AI strategies. All senior civil servants and future leaders need to understand their critical role in fostering a more systemic approach to digital transformation across government.
4.2. Public servants need to be supported to use new AI tools to mitigate important risks, like data security and bias.
4.2.1. A May 2023 Apolitical global survey of over 1,600 public servants (including 800 from the UK) found that digital transformation and AI are regarded as the most important issues they want to learn more about. By contrast, early data from our ongoing AI research suggests that the overwhelming majority of public servants have not received any training on how to use generative AI tools.
4.2.2. There is academic research to suggest that the effectiveness of generative AI tools is influenced by the competence of the user.[10] To mitigate against the well-documented AI risks around data security, inaccurate information, and bias, governments need to train public servants about what this technology is, how it works, and how it can be used safely and effectively in their particular professional context.
Annexe
More than 10,000 public servants from 116 countries have engaged in learning on the Government AI Campus since its launch. Apolitical tracks key trends in how public servants are engaging with AI (specifically generative AI) through community polls, surveys, interviews, and interactions on our platform. There are three key insights we have learned.
Insight 1: Most public servants are experimenting with generative AI
Regular polling of our community shows that the adoption of generative AI has spread rapidly among public servants. As early as March 2023, around 40% of public servants in our community said they were experimenting with tools like ChatGPT. Our most recent poll, from December 2023, showed that 59% of public servants said they were using it; among those, 42% said they were using it at least once every week.[11] Significantly, this data suggests that generative AI adoption in the public sector is keeping pace with the private sector: around 40% of private sector employees were using generative AI in early 2023, and recent evidence (November 2023) suggests average adoption is at 55%.[12]
Data from Apolitical polling. Average sample size is 520.
Insight 2: Public servants are mostly using AI to help with research and writing
Results from the NAO’s recent review suggest that the forthcoming wave of AI deployments across UK public bodies will be focused on improving internal processes and supporting operational decision-making.[13] That complements evidence from Apolitical polling, which shows that public servants are mainly using generative AI to support research and writing tasks.[14] The AI tools that public servants most commonly use, like ChatGPT and CoPilot, are well-suited to individual-level productivity tasks, like simple data analysis or drafting correspondence.
Data from Apolitical polling, July 2023, Sample 437
Insight 3: Governments around the world are investing in building AI capability
Governments around the world are investing to build their AI capacity. People and skills are critical to this effort and we are witnessing a rapid push to deliver training at scale.
May 2024