Michael Jary, who served for two years as the primary external director advising government, says that automated tools could improve operations, and also help support a more connected civil service
The government’s former lead non-executive director has spoken of his belief that artificial intelligence systems capable of operating with limited human supervision could offer departments productivity savings of 30% or more.
Michael Jary, who was one of the founders of OC&C Strategy Consultants and served as government lead NED from 2022-2024, made his comments about the potential of “agentic AI” at a recent event hosted by the Institute for Government.
He added that the use of such AI power could underpin a more networked, less siloed future for government.
“We must imagine the new model of digital-era government,” he told the event. “I would guess that the application of agentic AI in government could lead to productivity improvements of at least 20% to 30%. And also make government faster, more accessible and more transparent.”
He added: “This moment also allows us to imagine new organisational models. Rather than government being siloed around departments and agencies and on separate central and local levels, it could be inherently more networked.”
Jary said that common data sets and components, linked by a network of rules and APIs, could transform fragmented services into “seamless journeys”. He said that because government services differed greatly, the imposition of “monolithic or monopolistic structures” should be resisted. However, he said common infrastructure could be shared.
Jary said such a future would require changed ways of working, with agile development teams delivering for real user needs.
“The organisational and cultural distance between policy and design can be shortened, indeed policy is design,” he said. “It’s about testing solutions with stakeholders, learning, and getting something into production, fast. Most successful corporates already operate this way.”
Jary said India’s delivery of the Aadhaar digital ID system to serve a population of 1.4 billion with a team of just 100 people was proof of the opportunities at hand.
Related content
- AI could save £5bn a year and take workload from ‘tens of thousands’ of jobs, government ops chief says
- Government LLM pilots to assess potential ‘employee satisfaction’ uplift
- Failing to secure public trust could derail government’s AI agenda
He said that while compliance, accountability and evaluation could be designed into systems, there would be “important debates” about what parts of the future tech infrastructure need to be in public ownership, and what is done by the private sector.
Jary said the future would require not only a smaller civil-service headcount but also a very different skillset, with designers, product managers, engineers, data scientists, and citizen-insight specialists in even more demand.
“I’m quite confident that government can win the war for talent if it articulates the opportunity,” he said. “Would a young technology graduate prefer fine-tuning ad placement at Facebook or tackling the most important societal challenges?”
Jary also spoke of the need for government to refocus itself away from departmental boundaries, with small, multi-disciplinary teams that are empowered to make decisions. He said the Starmer government’s move to organise around missions was “promising”, but warned it may get “bogged down in the silos of the organisation structure and cumbersome coordination”.
Jary added that his time as a NED at the Ministry of Housing, Communities and Local Government had made him recognise the need to reverse the “extreme centralisation” of government in the UK. He said the growth of funding initiatives that required local authorities to bid to central government for resources was a particular concern.
He said local government needed to be empowered to deliver, with deeper devolution to big cities particularly important.
Jary said such devolution should be funded by “radically streamlining and simplifying central government to do what only it can do”.
