AI & Procurement in UK Public Services

Using AI in Bids and Tenders – Setting the Example?

Rowena Osborne

August 11, 2024

The UK Public Sector is trying to balance caution with this AI while embracing innovation.

In March 2024 the Alan Turing Institute, working with UK Research & Innovation (UKRI), held webinars around AI regulation in the UK and EU. Around the same time the Procurement Policy Note: Improving Transparency of AI Use in Procurement Information Note 02/24 was launched. I doubt these happened in isolation and reflect the increased interest in generative AI.

UK AI Regulation

Currently this is in development with work to date focusing on collating standards to inform both the UK’s AI white paper and provide practical steps in the meantime on how to build compliant systems. There are five principles:

Safety, security and robustness

Appropriate transparency and explainability

Fairness

Governance and accountability

Contestability and redress.

ISO/IEC 42001: AI Management System has been closely reviewed and supports all five principles particularly around accountability and governance principles, and stronger physical and psychological well-being where mapped in. The UK white paper however requires stronger remits on fairness in regards to unfair market impacts and what it take for an AI system to be fair.

EU AI Act

The EU is set to be the first in the world to put formal legislation in place around AI. Drafts of the Act show it will focus on:

A harmonised framework for development and marketing of AI tools

Rules to apply to all types of AI and all sectors including fundamental rights protection.

Having extraterritorial application.

There are key provisions on the risk-based approach to AI models which applies to providers and deployers of AI systems, and data must be ‘representative’.

The key difference between the two approaches are the EU is more rigorous on safety and the UK is more focused on innovation.

How does this inform bids, tenders and proposals?

It gives us insights on what some of our clients and suppliers will have to take into consideration. It also provides a temperature check on the thinking around AI products which we may need to address in upcoming narratives. Much like the growth in cyber security requirements and questions, AI has the potential to become it’s own sub-topic especially in public sector opportunities.

PPN Info Note 02/24 sets out key points for civil servants to use when issuing tenders including:

Making it clear bidders cannot use confidential contracting information (not available in the public domain) to train AI systems e.g. tender documents to train AI or Large Language Models (LLMs) to create future tender responses.

If suppliers use AI tools to create tender responses, additional due diligence is required to check they have the appropriate capacity and capability to fulfil the contract and to establish the accuracy, robustness and credibility of suppliers’ tenders.

Plan for increases in activity e.g. more clarification questions and tender responses as AI streamlines processes.

It highlights that AI has the potential to support decision making processes but it is critical decisions do not rely on AI. Also, that LLMs may include inaccurate or misleading statements as they are trained on ‘statistically plausible’ strings of text. It recommends adding specific questions to disclose the use of AI in tender responses e.g.:

Have you used AI or machine learning tools, including large language models, to assist in any part of your tender submission? This may include using these tools to support the drafting of responses to Award questions.

Where AI tools have been used to support the generation of Tender responses, please confirm that they have been checked and verified for accuracy.

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.

Conclusion

Does this make you more or less likely to use AI in UK bidding? I think in the right bid phases – yes – with caution and honesty. Making sure the bid timeframe allows for the accuracy checks before taking first drafts into the review process and evolving the narrative will be vital.

Main Image: “Machine Learning & Artificial Intelligence” by mikemacmarketing is licensed under CC BY 2.0.