‘Not enough technical people work in the civil service, meaning civil servants don’t understand technologists or tech,’ says ex-head of AI incubator Laura Gilbert.
Image:
Legacy tech on Whitehall
Last week the government announced its ambitious industrial strategy, promising investment in AI, quantum and semiconductors. But these lofty promises risk running into the sand unless a culture clash between technology and government can be addressed.
Dr Laura Gilbert has seen this clash from the heart of government, having headed up the Incubator for Artificial Intelligence and the 10DS data science team at the Cabinet Office. The cautious culture of the civil service can be a major barrier to technological innovation, she told Computing.
“The civil service doesn’t take risks,” she asserted, comparing the attitude towards taking a chance between government and business. “In business if you hire someone who refuses to take risks you’ve hired the wrong person. They will never deliver an extraordinary service. They will never change the way anything works.”
There is little incentive built into the machinery of government to deliver extraordinary services or to change how things work, and a lack of clear accountability means the same mistakes are repeated over and over. Added to which a lack of institutional enthusiasm for exploring new solutions means that many promising ideas fall at the first hurdle, failing to attract sufficient support.
“It’s not driven by losing a market share or losing out to a competitor, and failure is very indirect. So people are able to waste hundreds of millions of pounds on things that don’t work, and then continue doing it again and again.”
See also: Public sector IT is all plans, no progress, says ex-DWP CDIO
The situation has improved somewhat in recent times, she said – with AI all the rage politicians have been keen to have something tangible to point to. Happily, resistance to tech-driven change is gradually lessening too, but core issues remain, not least a dearth of technical expertise at the heart of government.
“Not enough technical people work in the civil service, meaning civil servants don’t understand technologists or tech,” said Gilbert.
The technically minded civil servant who want to make a difference therefore needs to be prepared to go it alone, build a MVP and demonstrate potential benefits rather than waiting for permission.
“If you wait for the system to agree to do that as a whole, you’ll be waiting a long time. So what you do need is visionary people. You do need brave people, and you do need very intelligent and professional people who will go in and lead the way. It’s not going to do it by itself.”
Procurement problems
Gilbert also raised issues with the government’s procurement process. Many public sector IT projects are very large and complex. Understandably, procurement tends to be handled by officials who lack a deep understanding of the technology they are purchasing. Suppliers can take advantage of this knowledge gap, charging multiples of the real cost with layers of secondary providers also creaming off their share. “[Suppliers] don’t necessarily have the best interests of government in mind,” Gilbert cautioned. “You see this over and over again with some of these big tech companies.”
To address this issue, she suggested, the civil service should hire more technical people with experience of building systems themselves – people who know roughly how long projects should take and how much they should cost. These tech-savvy individuals should be made available to procurement professionals as contracts are drawn up. “There should be an upper limit as to what it should cost, and we should be not be allowing suppliers to run off for four years then come back for something that’s no longer state-of-the-art. They should be sitting next to the user, ensuring they are delivering the right thing quickly and in stages if needed.”
Getting it right
Things could be different, she insisted. While in charge of the AI incubator for government she implemented a policy of “radical transparency”, with engineers encouraged to write blogs to demonstrate what they were doing and why, the purpose to engender trust and understanding among their more policy-oriented colleagues. Open sourcing the code performs a similar function among the technical.
The fruits of the incubator’s labours include Redbox, a tool to enable civil servants to interact with official documents, and Connect a solution to optimise connections to the National Grid, with the potential to save up to £10 billion over 10 years. Importantly, unlike many government tech projects, if they fail to deliver they will be mercilessly culled.
Asked for examples of governments getting tech right, Gilbert mentioned Ukraine, a special case perhaps, but a country that has managed to stand up dozens of factories to make drones and prosthetic limbs in a very short space of time. Then there’s Albania, which has built a multilingual AI assistant to help people fill in forms and to interact with every part of government. Albania has been bold, she said, and it has paid off. “What they’re not doing is worrying that it might get the odd form wrong. What they are doing is offering really good service, and they’re doing it incredibly well.”
Gilbert left the AI incubator at the start of the year and is now head of AI for government at the recently opened Ellison Institute of Technology Oxford, and also a senior director of AI at the government advisory not-for-profit Tony Blair Institute (TBI), where “we will be building tools that make governments better, get them to make better decisions.”
Laura Gilbert was speaking to Computing at at ServiceNow’s public sector event in London.
