In quick succession after the UK Government published its AI Opportunities Action Plan, the National Audit Office (NAO) released its report ‘Government’s approach to technology suppliers: addressing the challenges’ (the NAO digital procurement report). Reading both documents in relation to each other paints a picture of the difficulties and pitfalls in the acceleration of public sector AI adoption desired by the UK Government.
More generally, I think this reflects the tensions faced in most jurisdictions yet to find ways to adapt their procurement practices and programmes to the digital environment and to ‘data first’ approaches, and how important but expensive interventions in ensuring continued investment in procurement skills and systems can have large knock-on effects on the broader functioning of the public sector for better and for worse (an issue I am researching with Nathan Davies).
In short, the AI Opportunities Action Plan seeks to ‘push hard on cross-economy AI adoption’ and places AI procurement at the forefront of that effort. As I highlighted in my hot take on the plan, one of its main weaknesses is the lack of detail on the measures to be put in place to address the large digital skills gap in the public sector— while the extent to which that gap is reduced will be determinative of how far AI procurement can go in contract design, contract and performance management, and other crucial tasks to deliver the plan (see full comments here).
This built on my earlier research, where I have stressed how a risk-based approach to the design and implementation of AI procurement requires advanced digital skills, and how shortcomings in digital skills compound key risks, such as data governance, technological and operational dependency, and system integrity risks (see here ch 7, and here).
My research, and that of others such as the Ada Lovelace Institute (see here and here), has also stressed how current guidance and best practices are insufficient to support the procurement of AI, and how this compounds the issues arising from shortfalls in digital skills. It is also clear that these issues are bound to especially affect particularly resource-constrained areas of the public administration, and that local authority procurement is in a uniquely challenging situation (which I am researching with colleagues at Careful Trouble).
All this research raises significant questions on the deliverability of plans to accelerate AI adoption in the public sector in ways that align with the public interest and do not generate unacceptable risks of mass harms (see here) and, in my view, advocates for a different approach that focuses on putting regulatory stopgap solutions in place while investment in the required fundamentals (data, skills, processes) is addressed, and provides a source of independent oversight of this high stakes process of public sector digital transformation. There are also environmental and other reasons to favour a ‘frugal AI’ approach (see eg here).
The main issue with such cautious (or I would say, realistic) approaches is that they do not convey a politically popular message, and that they are exposed to criticism for being excessively pessimistic or over-prudent, or/and for slowing down the adoption of AI-based solutions that (with the right technosolutionist lenses on) will unlock massive changes in resource-starved public service delivery.
In my view, the NAO digital procurement report makes for grim reading, but it is a strong endorsement of the need for such alternative, slower approaches.
As summarised in its press release, based on its recent investigations into different aspects of government digital transformation programmes, the NAO extracted the following lessons for the UK government to consider:
At central government level
-
There are not enough people with digital commercial skills in government.
-
Government procurement guidance does not address all the complexities of digital commercial issues.
-
Government struggles with the breadth of issues that affects its ability to engage effectively with suppliers.
At department/ministerial level
-
Departments do not make full use of their digital expertise when procuring for technology-enabled business change.
-
Digital contracts are awarded with insufficient preparation.
-
Approaches to contract design can negatively impact successful digital delivery.
This leads the NAO to formulate related recommendations
‘The NAO is recommending that the centre decides who should take ownership for addressing the problems identified in our report. It should produce a sourcing strategy to include improvements in how it deals with ‘big tech’ and strategic suppliers. It should also create a digital skills plan to plug recruitment shortfalls and to better equip and train decision-makers responsible for digital commercial activities.
For departments, the NAO recommends departments strengthen their ‘intelligent client function’. They need to identify and develop key requirements before tenders and bid processes commence, and improve how policymakers and technical specialists work together with procurement specialists. Departments should also improve their capability to collect and use data to inform a pipeline of supply and demand. This would help the centre of government build a more strategic approach to suppliers.’
In my view, the NAO’s findings and recommendations stress the crucial importance of addressing the public sector digital skills gap (both at central and departmental/contracting authority level), so that shortcomings in procurement guidance and in subsequent procurement planning and design, and contract management, can take place. They also stress the urgency in creating workable sets of guidance that provide much more detail and support than the existing generic documents.
What is worth further highlighting is that, unless and until these issues are addressed, digital procurement cannot be successful and, what is more troublesome, in the current context, an acceleration of AI procurement is a very bad idea because it will aggravate the problems identified in the report and potentially create situations that will be impossible or exceedingly costly to fix later on.
In my view, the NAO report should be a wake up call to the UK Government — and to other governments operating in comparable contexts — to do things more slowly and to find ways to fix technological debt, skills shortcomings, and lock-in and other problems associated with high concentration in digital markets. It is difficult to fix them now, but it will be more difficult to do every year from now. Given the nascent state of AI procurement, it seems to me that there is still a window of opportunity to change tack. I am not optimistic that this will happen, though.
