UK Procurement Updates: AI and the New Procurement Regime

The UK government has continued its updates to publicly available information about the new procurement regime, due to come into force later this year. In recent weeks, we’ve not only received three different procurement policy notes (within the space of ten days!), but we’ve also seen the publication of the Procurement Regulations 2024.

On 25 April, APMP’s UK Chapter hosted a panel discussion – the second of a series which began with a similar discussion back in February. The discussion covered the recent updates made by the government, and focussed particularly on the role of AI in the new procurement regime.

Read on for a summary of the key updates from the UK government and particular points of interest from the panel discussion.

It’s likely that the new regime will commence from 28 October

The new draft regulations provide the legislative detail for the Act, and will bring some elements of the new regime into effect on 28 October. This date is set to be formalised this month.

During February’s panel discussion, there was some scepticism around the feasibility of this ‘go-live’ date. The timely publication of the secondary legislation and the consultation report, however, indicate that this commencement date is likely to go ahead.

The obligation to provide assessment summaries explaining reasons for loss has been watered down

The Act’s introduction of ‘assessment summaries’ – notices informing bidders of the outcome of the procurement, as well as giving feedback on their bid – has been altered slightly by the draft regulations. The previous iteration of the Act asserted that teams would receive explanations for both why they were awarded the grade they got, and why they were not awarded the next grade up.

The new regulations have removed the latter obligation on contracting authorities. While this is a disappointing omission from the regime given the potential this information had for helping bidders to improve on future submissions, teams can still hope to receive more consistent information from contracting authorities than they are currently. One panellist noted that a guidance note is expected to be released between now and October which should give more information on assessment summaries.

The new regime reduces some barriers for SMEs, but that AI could give advantage to bigger companies

The new regulations will bring into force a ‘tell us once’ principle, whereby organizations can enter all of their business information and answers to generic questions once onto the new digital platform, rather than completing these details for each bid. One panellist stated that this introduction reduces the disproportionate burden on time and resources for SMEs, as it significantly cuts down the amount of fresh writing that is needed for each procurement.

There were some concerns about AI potentially creating an uneven playing field for SMEs due to the high price point for innovative, complex AI systems. Panellists acknowledged that larger companies who can afford to be early adopters of the technology may have an initial advantage in bidding at this stage. However, as one panellist pointed out, the price point for such AI will likely come down as the technology evolves and its use becomes more commonplace. They also pointed out that open AI platforms such as ChatGPT can still offer SMEs the opportunity to reduce research workload and generate ‘starting points’ for writing.

Panellists were largely positive about increased transparency over AI use

A new procurement policy note, Improving Transparency of AI use in Procurement, was published in March, and put forward the idea that procurement teams should begin to ask bid teams for disclosures around the use of AI in their proposal development. This was met with a mixed reaction from our community. While some applauded the move for increased transparency, others shared concerns over the implications of being expected to disclose the role AI plays in their processes.

The panel acknowledged that there is a grey area in the policy note when it comes to whether or not procurement teams will be biased against the use of AI. The policy note gives the impression that there is some nervousness on the government’s part around false information entering into bids.

The panel’s AI expert stated that this idea of ‘hallucination’ by AI – i.e. where generative AI essentially fabricates plausible-sounding (but inaccurate) information – is a problem that has been more or less solved throughout the technological evolution of the past few months. Many AI systems are now geared towards discerning meaning from the user’s search teams, pulling sources from the internet that match that ‘meaning’, and citing said sources so that humans have the opportunity to verify them.

Another panellist stated that the problem of potentially inaccurate information appearing in a bid puts the onus on the contracting authority to conduct adequate due diligence to ensure accuracy. They also mentioned that the policy note reaffirms the non-discrimination principle in explicit relation to AI. This means that contracting authorities could face repercussions, such as damages claims, if it is found that they have allowed bias against AI use to interfere with fair evaluation of a bid.

A member of the panel stated that, from the vantage point of a government contracting authority, the use of AI could actually be beneficial. They pointed out that bid teams can sometimes stray into answering the question they feel the RFP should have asked rather than the question directly in front of them. AI, in contrast, will only answer the exact question as written, giving bid teams the opportunity to tailor that answer to include necessary information in a way that keeps true to the question’s intent.

There is not currently an appetite among procurement teams to use AI for bid evaluation

While panellists acknowledged the benefits of using AI in evaluation of more simple requirements in bids, e.g. clarifying whether specific thresholds are met, AI is simply not yet advanced enough to accurately evaluate more complex and human requirements. Indeed, one panellist stated that, even if AI were so advanced, evaluation of requirements that relate to humanity and the human experience are unlikely to ever be delegated to AI evaluation.

There was some discussion around how using AI in the evaluation process could ease potential capacity burdens on procurement teams. This issue is particularly significant given the AI policy note’s assumption that increases in the volume of bid submissions are expected to come as a result of widespread AI adoption. However, the panel generally agreed that, for now, AI use in evaluations is unlikely.

It was also generally agreed that, due to the principle of transparency in the Act extending to both the supplier and contracting authority side of things, evaluation teams will be expected to exercise transparency around any AI use in their evaluation process.

The point of AI in procurement is not to replace humans, but to speed them up

The panel’s AI expert pointed out that recent data collected internally at their organisation showed that teams who used a blend of human and artificial intelligence to produce bids saw an 8x speed improvement and a quality improvement of 10-15%. A member of the audience supported this conclusion, stating that a client of their organisation had engaged two teams in writing the same bid submission – one with the assistance of AI and one without – and found the AI-assisted bid to be substantially superior.

While it is too early to have any substantive data on whether AI use can help boost win rates, the panel was hopeful that such data may become available soon.

The panel concluded that, while AI cannot write or comprehend better than humans can, it can do both significantly faster. This means it can free up some time for humans to home in on the specifics and perfect the details of the bid.

APMP members can watch the full panel discussion on APMP TV here.

Join the Conversation