In the recent PPN 02/24, ‘Improving Transparency of AI Use in Procurement‘, the Cabinet Office has provided guidance on the use of AI in procurement for central government departments, executive agencies, and non-departmental public bodies. This guidance has significant implications for both English public sector buyers and suppliers.

Before delving into the implications of PPN 02/24, it is crucial to acknowledge the dynamic landscape that AI introduces to procurement within the public sector. The guidance provided by the Cabinet Office not only signifies a step towards modernisation but also underscores the importance of transparency in leveraging AI technologies.

Both buyers and suppliers are poised to encounter a change in basic assumptions in their approach to procurement, necessitating a comprehensive understanding of the implications outlined in the policy. Let us review what these implications are, starting with what they mean for suppliers to the public sector.

Implications for Suppliers

For suppliers, the implications of PPN 02/24 extend beyond mere procedural adjustments; they signify a fundamental transformation in their ability to respond to a tender, and the pace of that response.

As AI increasingly permeates bid writing processes, suppliers must adapt to stringent disclosure requirements and ensure the integrity of AI-generated responses. Moreover, heightened competition underscores the importance of maintaining credibility amidst an evolving procurement ecosystem.

Key implications include:

• Disclosure of AI Use: If you use AI to write your bids, you will likely have to say so. Suppliers using AI to develop their bids must be prepared to disclose this information. Failure to do so can undermine their credibility and negatively affect their bid score, effectively removing them from the process.

• Manual Proofreading: It is vital for suppliers to manually proof their AI-generated responses to ensure that they can fulfil the promises made on their behalf by the AI. This is to avoid the risk of the AI adding in misleading statements or information that the supplier cannot provide. You don’t want your bid to make promises you just can’t keep.

• Training AI on Own Data: To minimize the risk of misleading statements or hallucinations, suppliers should consider using AI that is trained only on their own data. That way, it’s more likely to be right and safe. This can also help in ensuring the accuracy and reliability of the AI-generated responses. In addition, suppliers must ensure that confidential client data is protected, and should not be added to a LLM (Large Language model) generative AI, unless they can guarantee that there is no possibility of that data being subsumed into the LLM’s public-facing corpus, which can then risk exposure at a later date to an non-authorised party, including the organisation who created the LLM.

• Competition and Credibility: Because more sellers will use AI, you might face more competition. With an expected increase in the use of AI for bid writing, suppliers should be prepared for heightened competition and the possibility of remarkably similar bids. Anecdotal evidence suggests that such instances may lead to all affected suppliers’ bids being marked down, effectively discounting them from the process due to undermined credibility. If all bids end up looking the same, don’t be too surprised if no-one is appointed.

• Ethical AI Use: Make sure your AI is fair and honest. Suppliers must prioritise ethical considerations in AI deployment, ensuring that AI systems adhere to principles of fairness, accountability, and transparency. This entails not only disclosing the use of AI but also demonstrating efforts to mitigate biases and uphold ethical standards throughout the bid process. Failure to do so could result in reputational damage and jeopardise future bids.

Implications for Buyers

On the buyers’ end, the implications of PPN 02/24 usher in a new era of diligence and scrutiny in procurement practices. Integrating AI into the evaluation process necessitates a cautious approach, emphasising the importance of retaining, and in some instances increasing, manual due diligence to maintain the quality of their supply base, and by extension, the products and services they procure.

By incorporating specific questions about the use of AI in tender documents, buyers can proactively assess the use of, and by extension accuracy and reliability of, AI-supported bids. Furthermore, understanding how suppliers plan to integrate AI technologies into service offerings enables buyers to anticipate the potential impact on end-users and residents, fostering a more informed and transparent procurement environment.

Five key concerns for buyers are:

• Disclosure Section in Invitation to Tender: Ask sellers if they used AI to write their bids. Buyers are advised to include specific questions in the invitation to tender regarding the use of AI in the procurement process. This includes asking suppliers whether they have used AI or machine learning tools to assist in their tender submission and if so, whether the resulting bid, and the tool that was used to create it, have been checked and verified by the supplier for accuracy before bid submission.

• Protecting Data Confidentiality: Buyers should be clear about their expectations with regards the addition of any confidential information supplied by them to LLMs by suppliers.

• Due Diligence: Even though AI can help, always double-check important decisions yourself. While AI can support decision making, it should never replace a buyer’s own due diligence when assessing the supply chain. Buyers should be aware that AI may not always be able to determine factual correctness and should exercise caution when relying solely on AI-generated information.

• Integration of AI into Service Offerings: Buyers may also inquire about how suppliers plan to integrate AI or machine learning technologies into the service being procured. This information can help buyers assess the potential impact of AI on the services they are seeking and ensure that they are fully aware of the experience for end-users and citizens.

• Monitoring AI Performance: Buyers should establish mechanisms for ongoing monitoring and evaluation of AI performance throughout the procurement lifecycle. This includes tracking the accuracy and effectiveness of AI-supported decision-making processes, identifying areas for improvement, and implementing corrective measures as necessary to ensure the integrity of the procurement process and outcomes. Fix any problems to make sure everything goes smoothly.

In conclusion, the implications of PPN 02/24 for both English public sector buyers and suppliers are significant. Suppliers must be prepared to disclose their use of AI in bid writing and ensure the accuracy of AI-generated responses. Buyers, on the other hand, should exercise caution when relying on AI-generated information and include specific questions about AI use in the tender process.

By understanding and addressing these implications, both buyers and suppliers can navigate the use of AI in procurement more effectively and transparently.