I wrote for @civilserviceworld.bsky.social about the skills the policy profession should cultivate in an AI-augmented future
www.civilserviceworld.com/professions/...
I'm in @computerweekly.bsky.social talking about how LLMs might change the way the civil service makes policy
www.computerweekly.com/opinion/Tech...
If we can get really good surface level material from an LLM in seconds, humans themselves will recognise they need to change to add value. And it’s that intrinsic motivation which has to be part of the trigger to drive any real change I think
I think you’re basically right, but I think my emerging view is that LLMs could be a sufficient game-changer to force the whole eco-system to change and developing SME to become a priority (as @teodorgrama.bsky.social has just posted after we discussed your thread!)
Perhaps widespread LLM adoption (in the drive to meet ministerial expectations, increase productivity and cut costs) might be what finally forces the civil service to more radically rethink how it approaches subject matter expertise in the policy profession.
Thanks for the thoughtful response! Probably should have been more explicit in the piece but I agree churn is a big problem to developing subject matter expertise. Hopefully I was clear that officials having that expertise is important to avoid the sort of 'missed hallucinations' you describe
Correct. Which is why the policymaker’s role will increasingly include adding specificity & context to LLM outputs & developing strong networks who can feed real-time insights into policymaking. Maybe comes out more in the full piece than my thread
www.instituteforgovernment.org.uk/publication/...
It's been a lot of fun writing this piece. Read the whole thing here www.instituteforgovernment.org.uk/publication/...
But they need the skills to do so & the civil service must not overlook the workforce challenges that rolling LLMs out more widely will bring
If adopted well LLMs will create a richer policymaking process. Automating more basic tasks means policymakers will have the chance to acquire new insights & develop new capabilities, leading to more complete policy & better results for citizens
But that's pretty far outside the existing paradigm. Perhaps the best approach would be for the civil service to start by ringfencing tasks, but actively commission test & learn projects to explore more imaginative approaches & scale those where they work
2) sounds interesting. Maybe the type of official who is currently a junior policymaker could instead be deployed to the frontline, giving them personal experience of the operation of the state which they can use in a more conventional Whitehall policy role once more senior?
The civil service basically has two options
1) Preserve some basic tasks for junior officials so they can build domain expertise needed to intelligently use LLMs
2) Reinvent the way policymakers acquire expertise, reducing reliance on the (now AI-augmented) traditional methods
This also has consequences for policymakers’ ability to gather all-important insider insights. All very well freeing up time for policymakers to collect info in new ways, but if they don't have a baseline level of expertise they won't know where to look for it & how to interpret it
LLMs might not only make domain expertise more important to possess, but also harder to acquire - they’ll automate things that policymakers traditionally have used to acquire the building blocks of knowledge, like putting together the first draft of a submission
Performing this role well relies on policymakers having the capability to make it work. If domain expertise & 'insider' insights are the things for which they're valued, they must have the commensurate knowledge. But this presents a paradox...
As @tomwestgarth.bsky.social told me, “a core value add for policymakers will be – what valuable information do I have that is not written down on the internet?”. Understanding the hidden ways in which the state is dysfunctional (or works) will be crucial
The time freed up by using LLMs to perform traditionally time-intensive but more basic tasks could also give policymakers the chance to gather & deploy new types of info. Particularly important will be hyper-specific or real-time insights - things LLMs struggle to capture
2) Layering new ideas on top of LLM outputs. This includes being prepared to push them in a more radical direction - LLMs tend to give the 'standard' answer, when sometimes a more creative one is needed
1) Using expertise to edit & shape LLM ‘first drafts’ (e.g. of submissions) – checking for & correcting hallucinations & untoward biases. This also means interrogating the assumptions buried within an LLM’s output & checking alignment with the elected govt's political values
LLMs can produce a written output (like a ministerial submission) in seconds rather than days – rapidly performing the tasks traditionally core to a policymaker’s daily existence.
So in an AI-augmented policymaking process, how does the policymaker still add value?
How will LLMs change the way policy civil servants work?
New @instituteforgov.bsky.social paper finds that:
1) The role of policymakers will increasingly be to add expertise & new insights to LLM outputs
2) LLMs might make it harder for policymakers to acquire important skills
Much discussion about AI in government - @jordanurban.bsky.social brings some balanced clarity in this piece which is well worth a read
www.instituteforgovernment.org.uk/publication/...
💼Are you or have you ever been on the civil service fast stream? Are you applying this year?
The @instituteforgov.bsky.social wants to know your views on the scheme - and how it can be improved.
Fill out our survey (and get the chance to win a £30 Amazon voucher): tinyurl.com/5fvh7wz8!
Generative AI & particularly LLMs will have a big impact on the civil service. For @instituteforgov.bsky.social I’ve zoomed into how it could change policymaking in particular.
What are the opportunities & risks? Read my piece to find out!
www.instituteforgovernment.org.uk/publication/...
Some key takeaways from our chapter on AI, digital and data in Whitehall Monitor 2025 [🧵]:
Last chance to apply for this great role!
On a personal level, delighted to see Pat McFadden's speech yesterday focus on improving civil service recruitment & bringing more outsiders into government. Something I've been urging for years & could be an important set of reforms. www.instituteforgovernment.org.uk/publication/...
What is the state of diversity in the civil service? My latest @instituteforgov.bsky.social analyses the most recent data and picks out the biggest stories, trends and changes from the last two decades. Quick thread on where the civil service stands in 2024 – and how it still needs to change. 1/5
This will be a great event and we’ll learn a lot about what public sector reform under the new government might mean. Sign up below to come along!
