CHRISTOF STACHE/AFP/Getty Images
show image

The 2020 digital government wishlist

In 2019 the UK government commissioned a review of AI adoption in the civil service, published guidelines for procuring AI-powered technology and launched the latest round of its GovTech Catalyst Fund. But the findings of its AI review remain closely guarded and the appointment of Dominic Cummings as Boris Johnson’s most senior advisor suggests that more radical change could be on its way.

If the government is to embark on a more fundamental transformation programme over the coming years, it will need to lay the groundwork over the next twelve months. We asked digital government experts about the policies, initiatives and approaches they would like to see in 2020.

Double down on transforming procurement of digital services

Daniel Korski is CEO of Public, a venture capital fund focusing on the GovTech sector

In 2020, I want to see government double down on transforming how it sources and procures digital services.

2019 saw a sea change in digitisation with the creation of NHSX, guidance for departments on how to use and buy AI, and the rollout of the Spark Dynamic Purchasing System – all of which are proof that Whitehall isn’t totally afraid of innovation.

Going forward, major investment in both the GovTech ecosystem and the digital skills of officials will be essential: a National School for Government and Technology would give decision-makers the ability and confidence to choose innovative technology that’s fit for purpose.

Right now, though, I want to see government chasing the easy wins: roll out Spark to more buyers, emphasise the cost of obsolete systems in any business case for digitisation, explore opportunities to carry the NHSX model across to other departments and bodies, and help local government to take advantage of the opportunities central government has created.

There’s still a long way to go in 2020, and to lose the momentum that has been built up over the last 12 months would be a tragedy.

Decide what digital “success” means

Dr. Tanya Filer leads the Digital State project at the Bennett Institute for Public Policy, University of Cambridge, and is director at StateUp

Digital government could do with less of many things in 2020 (silos, legacy IT systems, AI controversies, to name a few). But the festive season is no time for restraint, so my wishes focus on what we could do more and better:

More and better evidence of success

I hope that 2020 will be the year of measurement and evaluation. The last UK government, on paper, did too. Responding to the House of Commons Science and Technology Select Committee’s report on Digital Government in October, the government agreed that “there should be clear metrics to measure the progress of government digitisation”. It planned to work with departments to develop “metrics that will enable us to measure and evaluate our success”. Those ambitions are commendable. Clear metrics can help to set ambitious whole-of-government strategy, and to tweak policy when things go awry.

A starting point is to decide what digital “success” means. UK government identifies government digitalisation as a driver of ‘efficiency’, which is vague. Global indices, which help to shape governance here and elsewhere, also focus on efficiency gains. Digitalisation can bolster productivity and reduce costs. But to ensure digital government improves relations between the state and citizens, in 2020 we should be more creative, evaluating and rewarding government digitalisation projects against a holistic range of objectives, including quality (of which efficiency is one part) and accountability.

More publicly available data on GovTech start-ups and SMEs

Last year I wished for “more and better data on the lifecycle of GovTech startups”. I’m repeating the wish for 2020 (they say persistence pays). Good analysis of such data could help us to ensure that entrepreneurs in this space are being supported in the right ways, and at the right moments. Are GovTech start-ups becoming scale-ups? How many? How long does it take them? Under what conditions? Are they pivoting away from the public sector to solve private sector issues instead? How does the quality of services they deliver compare to large, incumbent providers? The Conservative Party manifesto promised an improved R&D budget. This data could help to work out how to direct enough of it, in sensible ways, towards innovation with public purpose.

Become an exemplar for transparency in algorithmic decision-making

Elliot Jones is a researcher at Demos

Government should become an exemplar for transparency in algorithmic decision-making. The Government’s guide to the use of AI in the public sector rightly includes consideration of transparency but it is light on detail and sits uneasily with current failures. The Home Office has been criticised by MPs like Chi Onwurah for refusing to give details of its visa processing algorithms and the Joint Council for the Welfare of Immigrants has been forced to take legal action against the Home Office to get an explanation of how the algorithm works.

In practical terms, the government could start by piloting the Model Cards developed by Google as a shared framework for machine learning “nutrition labels”, perhaps using the post-Brexit immigration overhaul as an opportunity to make things transparent from the start. And where government leads, industry may follow, through procurement requirements and setting standards citizens will come to expect in all walks of life.

Develop a standardised process for data sharing with suppliers

Tom Nixon is head of government practice at Faculty

One thing that would significantly speed up AI adoption across government is a standardised process for data sharing with suppliers. This would take the uncertainty and risk away from individual buyers, who are currently having to work out the rules from scratch.

AI is built using data science, which by its nature is quite experimental: it is often quite hard to know exactly which data will be useful for training an algorithm, or what the benefits will be, until after the data has been analysed. This doesn’t fit well with established data governance processes which expect certainty of use and clarity of impact. And often teams trying to adopt AI are having to break new ground within their departments, working through a myriad of opaque governance and approvals processes. This can lead to delays accessing data, and project risk sat with teams and suppliers.

So we would like to see government developing a set of general guidelines for departments on what data it is reasonable to share for what purpose when building AI, with a standardised approvals process to accompany that. This would give government buyers more confidence, and should speed things up for suppliers.

Exclusive: government blocks full publication of AI review