The Procurement Act 2023 took effect last February to improve how public bodies buy goods and services. The act aims to simplify the process and help government bodies become more flexible and innovative in their purchasing. As AI is rapidly being adopted across all sectors, the government has been gearing up for its use, too. The AI Playbook, released in February, provides guidance for government employees on using AI safely and effectively, while the roadmap for harnessing AI's benefits, the AI Opportunities Action Plan, was launched a month earlier.
This step change in government’s preparation for AI provided Civil Service World and Capita with the opportunity to convene senior civil servants and industry experts to discuss how the technology could be harnessed. CSW editor Suzannah Brecknell chaired the event under Chatham House rules, so attendees are not named in this report. The discussion kicked off with the question of what AI could achieve in the world of procurement.
How AI could be used
Participants felt AI could help with public sector staff shortages by getting “rid of some of the noise” and freeing up its “scarce” but “skilled” employees. It could help save their efforts on activities such as diligence and answering emails to suppliers who haven’t read the spec properly. “The new Procurement Act obviously gives an awful lot of commercial freedoms, both in the new commercial processes that have been established and the ability to amend and adapt your process and respond to market changes,” one attendee pointed out. Freeing up procurement staff would allow them more time to focus on being innovative and creative with that freedom, it was suggested.
AI can help commercial teams be “at the top of their game” through speeding up searches to understand what they are buying, as well as the procurement process. One participant mentioned an AI agent could possibly help commercial and product experts develop specifications together.
AI can also help in areas where standardisation and consistency can improve public sector procurement. One participant noted: “We shouldn't be rewriting laptop specifications every single time across 20, a dozen, two dozen organisations each year to go out to the market.” Another attendee suggested it would be a “step change” once AI could ensure the procurement process was “consistent” and “risk averse”, and could “manage risk out”.
The high stakes in the public sector often drive significant risk aversion. However, one participant said AI could help address fraud by identifying suspicious patterns. “One of the areas I think that procurement could benefit is the use of algorithmic tools to look at behaviour like cartels.” AI could help commercial teams understand markets and identify where supply chains might break down, another added.
Challenges with AI
One key challenge with AI – as with many emerging technologies – is the speed at which it is changing and improving. This not only means that public sector organisations need to be agile with procurement, but also in the change management of new systems and solutions. As one participant noted, organisations can be like "supertankers," making it hard to get everyone on board with new developments.
As AI evolves quickly, both the technology and its terms and conditions are complex and unpredictable – and technology suppliers can rise and fall rapidly. As one participant noted: “We don't know what's good and what's not.” However, the roundtable recognised “it's emerging technology, so nobody else really understands it ”.
The roundtable also noted that AI was often ‘black-boxed’, meaning products were not always fully understood – both in the terms and conditions and the processes they use. This presents ethical as well as operational risks as one participant warned: it might mean you do “not detect some errors that will creep in when you're processing 1000s of things”.
Some AI providers were failing clients by not explaining their products properly, one participant claimed. Sometimes you have to “really hunt” for answers to fully understand the products, they said.
Solutions and what needs to change
One attendee suggested that to address the challenges of fast-changing and complex technology, public sector buyers should see working with their tech partner as “going on a journey”. You should be able to learn together as AI is a fast-changing world.
Attendees agreed it was necessary to have a good relationship with your technology partner with “strong alignment” of objectives, values, and ethical standards throughout. Tech companies would benefit from strong collaboration to develop useful products, while the public sector would benefit from the consistency of this shared alignment, one participant said.
A successful way to find the right partner, one attendee explained, is through small, targeted competitions – a method already used by a large government department. Once the partner is selected, the focus should be on working together to achieve outcomes, rather than sticking rigidly to a single solution.
Going forward
The Procurement Act 2023 brought in a new central digital platform, enhancing the old Find a Tender platform and making it easier for public bodies to meet new transparency requirements around their contracts. One participant said the new platform had a lot of potential to improve procurement through better data availability. “I’m really hopeful that we're going to start to see some really fantastic outputs,” they said, as more and more data is uploaded, allowing for greater analysis and transparency.
Participants also emphasised the need to not “reinvent the wheel”. It was important to look at “what can we do better that we're doing at the moment with the tools we've already got,” one explained, and to learn from others’ experiences.
It was better to not rush into AI, they agreed. It was better to test something less risky and make it work, rather than “boom, give it everything, and off we go”, as one put it. There are ways of working with the partner to ensure you test systems in a safe environment, one participant said. Doing this “doesn’t mean you’re a dinosaur, it means you're just really making sure the parameters and the boundaries have been properly considered,” one participant said.
However, participants agreed a culture change was needed in the public sector to help harness AI effectively, and failure should be seen as learning – not as something to fear. One participant said: “Failures are to be celebrated, because it's a shared learning of helping everyone not to go down the wrong path.”
“I think what stands out is we need people who are commercial AI champions,” one attendee suggested, describing the need for colleagues across commercial teams who “can shine a light on things, can push things, can drive things, can encourage things and showcase what can be done”.