The policy-industrial complex
Senior Researcher
One of the ironies of the labour market disruption caused by AI is that, for all the forecasts of economy-wide job losses, those so far most at risk are also closest to the technological frontier, in software development and coding.
At the same time, it’s often argued in Whitehall that AI will sweep away the majority of ‘back-office’ work, leaving policy professionals to perform at the top of their license in a newly slimmed down corporate centre. Less commonly argued, but just as plausible, is that the policy teams responsible for drafting the next government white paper, action plan or industrial strategy, will be amongst the first to see their jobs go (rather than, for example, staff in HR, audit or finance).
As a senior official told a Reform think tank roundtable, once you acknowledge that policymaking is fundamentally about developing and translating complex sets of rules (as well as managing stakeholders, keeping tabs on developments in the sector, and so on) you realise much more of the policy process itself could occur through code. Which would have profound implications for the ~35,000 civil servants in the policy profession, but also for many others employed in policy, external and public affairs roles in the private and third sectors.
Of course, the interface between the most senior civil servants and ministers is defined by human interaction and built on a strong relationship of trust. Roles that are defined by these characteristics may become even more valuable.
And in some quarters, there’s a belief that policy, as an area of expertise and profession, is a uniquely creative endeavour – an ‘art’ – based on advanced problem-solving and a rare ability to synthesise ideas from across disciplines and sectors (what Matthew Syed calls “recombinant thinking”). If that is true, perhaps we should be sceptical of how quickly AI will disrupt policy jobs?
Reality suggests this is an optimistic view of how many policies are developed, gain attention and are taken up by decision-makers.
Take the prevalence of the “posting-to-policy pipeline”: simple, often common-sense ideas that begin their life on X or Substack and through sheer virality are forced into the mainstream and eventually shape policy (YIMBYism, for example, or the recent focus on public sector R&D and capital investment).
The increasing success of these ideas suggest policy may be more susceptible to meme than detail. Despite the trope there are “no silver bullets solutions”, catch-all policies recur, like clockwork, in particular debates – that the Government should urgently enlist the support of a citizens’ assembly, for example, or set up hubs to co-locate certain services, or initiate a Royal Commission.
These ideas are produced, on mass, by what might be called the ‘policy-industrial complex’ — a sprawling network of policy thinking (involving the Civil Service, think tanks, local government, charities and advocacy groups) which prioritises easily digestible ideas that can be packaged up and sold to a popular audience. The Ford Model T of the policy economy.
In this industry, which (rightly) emphasises the value of cognitive diversity and disruptive thinking, there are strong incentives to champion ideas (some of which will soon be produced by AI) not because they are the correct, nuanced answers to complex problems – or strike the right balance between difficult trade-offs – but because, in effect, they signal that you’re in the know and make for a good elevator pitch.
Policy remains, as permanent secretaries themselves have reflected, one of last bits of Whitehall to be truly professionalised. Instead, policy generalism, disconnected from the realities and challenges of implementation, thrives – with very few incentives to develop expertise in a particular domain. Why spend years developing in-depth knowledge of housing and planning policy, for example, when the “build more/build less” dividing line almost exclusively sets the parameters of the debate?
This is where the ‘craft’ of policymaking matters most and brushes up closely with politics: how, as a minister, do you reconcile priorities, at a constituency, parliamentary and national-level, that are rarely aligned? And where competing values and interests militate against easy, five second answers?
Against this backdrop, policymaking is likely to add most value in a world of automation where it operates more as a ‘cottage industry’ – human, placed-based and connected to the needs of specific communities – than mass industry.