Artificial intelligence – perhaps most notably large language models (LLMs) – will change the role of policy professionals across the civil service. Policy work in Whitehall is already being augmented by tools like Redbox, which can summarise the policy recommendations in submissions and other policy documents, and Consult, which summarises and groups responses to public consultations. These sorts of tools are likely to continue to improve, with capabilities extending well beyond the publicly available version of ChatGPT (which for many remains the reference point for what LLMs can do). The improvements in civil service efficiency and so effectiveness they can provide means their use will only grow further.
LLMs have limits and there are some areas where they will have less impact. In particular, officials will continue to use their judgement to navigate the competing interests and idiosyncrasies of Whitehall to get things done. Understanding the very human side of Whitehall – unspoken power dynamics such as which officials and ministers are making the key decisions, what their views are, how those views make their way through the system, and what therefore needs to happen to make sure a policy gets through the Whitehall machine – will be a critical skill.
But increasingly, LLMs can adequately perform some of the tasks that are core to the policymaker’s daily existence. Getting to grips with a topic and producing a written output on it can be done in seconds rather than hours or days. This has substantial implications for the policymaker’s future role.
Going forward, it will be important for policymakers to do the things that LLMs cannot. Officials will need to edit and shape LLM “first drafts”, checking for and correcting hallucinations and biases. But civil servants will also have a far more creative role layering their ideas on top of LLM outputs, which tend to provide “standard” answers, and leveraging the sort of hyper-specific and real-time insights that LLMs can struggle to capture. These could be acquired in new and creative ways, such as spending time on the frontline of a public service, building a professional network which can provide real-time reactions to new developments, or something completely different.
If domain expertise and novel insights are the skills for which policymakers are valued, they need to acquire commensurate experience. But organisations across the world are grappling with a puzzle: by automating the more basic tasks that early career staff typically use to build their first blocks of knowledge, it becomes harder to develop the senior staff of the future.
The existing culture of the civil service can make developing those skills even more difficult: officials can move between jobs too quickly and subject matter expertise is not as highly prized as it could be. Combine policy officials who are too often non-expert with the puzzle posed by LLM adoption and, without change, it is difficult to see how officials can best develop the skills they need in an AI-augmented future.
"The whole ecosystem of the policy profession might have to be re-examined – including the qualities valued in the civil service’s recruitment process "
Solving the LLM adoption puzzle is step one. It could be achieved by ringfencing some of the basic tasks traditionally used to upskill early career officials, while testing new ways to give them knowledge and scaling those up if they work. For example, a relatively junior policymaker might be deployed for a stint on the frontline, giving them personal experience of the operation of the state which they could use in a more conventional policy role in Whitehall once they become more senior.
Step two is to solve problems that have made it harder for officials to acquire subject expertise. Excess turnover of staff is the most obvious, but the whole ecosystem of the policy profession might have to be re-examined – including the qualities valued in the civil service’s recruitment process.
Less tangibly, but no less importantly, a premium should also be placed on developing policymakers’ strategic instincts. OpenAI CEO Sam Altman has argued that “agency, wilfulness and determination will likely be extremely valuable. Correctly deciding what to do and figuring out how to navigate an ever-changing world will have huge value”. The civil service should cultivate these traits in its future leaders, whether through specific incentives in the performance management system, exposure to role models or active training.
The bigger picture is that the policy profession, and the civil service as a whole, must be open to change. AI will be disruptive. How the civil service channels its adoption will determine whether that disruption is positive or negative.
Jordan Urban is a senior researcher at the Institute for Government