In March 2024, the Cabinet Office addressed the use of artificial intelligence (AI), machine learning and large language models (LLMs) in public sector procurement for the first time. New guidance outlines the potential risks inherent to using AI in procurement exercises, in addition to ‘best practice’ approaches for mitigation measures.
Over the past 15 years, bid and tender consultancy Executive Compass have had direct involvement in over 7,000 PQQ, SQ and ITT submissions, and have published guidance around the use of AI when drafting public sector bids and tenders. Consequently, they are ideally positioned to provide commentary on how PPN 02/24 will impact future public sector procurement.
AI in bid writing – impacts on the tender process
For clarity, the guidance states that AI is not currently prohibited from use during SQ or tender process, and this is unlikely to change. Equally, the guidance issued from the cabinet office is only applicable to central government contracting authorities – local government are permitted to form their own guidance.
However, there are considerable impacts stemming from its use for suppliers and contracting authorities alike, including:
- Streamlined bidding processes, allowing for a greater number of overall bidders and new entries into markets for organisations new to tendering
- Greater volumes of tender submissions due to efficiencies and streamlined processes introduced by the use of AI
- Increased ‘hallucination’ from AI-generated quality responses, which can include misleading or outright incorrect content when not properly checked.
Understandably, dramatic structural changes will be necessary to ensure the tender is awarded to the most suitable bidder. Consequently, the PPN gives contracting authorities free rein to implement control measures as they see fit – so long as it does not impact the transparency and impartiality of the tender process.
Controlling AI in central government tenders
To account for increased use of AI and LLMs in central government bids and tenders, the PPN also issues guidance and control measures for ensuring its use is proportionate and appropriate. Essentially, AI will be integrated into existing bid processes, but stronger due diligence and time for evaluation committees will verify the authority are making the right choice.
One such control measure may include the use of longer evaluation timetables following the submission deadline. This will ensure sufficient time for evaluators to conduct due diligence in accordance with increased bidding activity and suspected AI- or LLM-generated content. To supplement, there may also be greater use of post-tender clarifications, bidder presentations and site visits to establish the veracity of a bidder’s individual submission.
These measures are illustrative examples only, and further measures could be introduced to check suppliers have the capacity and capability to meet the requirements of the bid.
Example ‘use of AI’ disclosure questions
Annex B of PPN 02/24 outlines example disclosure questions which authorities may consider using in future tender exercises – for instance, as part of the Selection Questionnaire or Invitation to Tender document.
- Have you used AI or machine learning tools, including large language models, to assist in any part of your tender submission?
- Please detail any instances where AI or machine learning tools, including large language models, have been used to generate written content or support your bid submission.
As it stands, questions are optional, non-scored and information-only, but give key clues and insight into how procurement will be structured alongside AI tools.
How you should use AI in your bid strategy
To summarise, the central government is taking a proactive approach to the use of AI and LLMs within the tender process. While there are efficiencies which can be introduced when completing a bid submission or drafting quality responses, it is not a panacea, and unadulterated AI content is unlikely to result in a contract award.
Furthermore, authorities are becoming increasingly adept at spotting content written exclusively using AI and LLM tools. Evaluators read hundreds of submissions each year, and can identify AI content in a short period of time. Although not a formal part of the marking criteria, this could end up exercising a negative influence on the marking process.
Consequently, in-house bid teams should adopt the following approach to AI in central government procurement:
- Recognise the limitations to AI-generated content, including overly generic content which is not specific to the opportunity
- Avoid ‘copy and paste’ approach to AI content by tailoring outputs to the opportunity, including additional levels of detail to ensure a bespoke response
- Review any content produced by AI or LLMs to ensure it is fully accurate, avoiding the potential ‘hallucination’ effect AI can occasionally produce
- Utilise AI as a research or ‘idea generation’ tool when drafting quality responses, to include content or differentiators which may be overlooked.
By using the above as a sensible, best-practice approach, you can profit from the efficiencies and streamlined processes resulting from AI, while ensuring suitable control measures are in place to mitigate any negative effects.
Also Read: Building a Robust Procurement Training Strategy: Essential Steps and Considerations
