Is UK Public Procurement Getting Confused About AI?

This is a big year for public procurement in the UK. In October (probably) the new Procurement Act becomes law, finally replacing the EU public procurement legislation with a new set of regulations designed by and for UK organisations.

Generally, I feel the Cabinet Office policy team did a good job steering the consultations and proposals into a set of new rules, although there are some issues that concern me. But the team has now tied itself in knots somewhat over a different issue that has suddenly leapt to the forefront of everyone’s minds – how AI might affect public procurement (and many other aspects of our life of course!)

A PPN (Procurement Policy Note) was issued the other day that has caused some controversy and confusion. I must say, PPNs are usually clear and helpful, whether or not you agree with the underlying policy they are communicating, but PPN 02/24 is a mess.  Having read it a few times now, I think the problem is that it tries to cover too many issues, all AI related but really quite different, in one note.  I can see the following all mentioned in the note:

1. Concerns about the use of AI in writing tender responses and proposals, in particular whether AI responses are likely to be inaccurate in terms of reflecting the actual capability of the supplier or how they will deliver the contract. In other words, the risk of AI generated bullsh*t showing up in bids. 
2. Confidentiality or even national security issues in terms of firms using government documents connected with the procurement process to train AI systems and models.
3. Worries that AI becoming ubiquitous and cheap is going to lead to many more suppliers putting in bids in response to opportunities, putting stresses and strains on procurement (and other) resources in public bodies.
4. Issues around the actual purchase of AI solutions.

It seems to me that these are totally different issues. For instance, even if there was an outright ban on any use of AI in developing bids (which would be daft), there would still be legitimate security and confidentiality issues around the use of government documents in “training” AI.  That needs to be considered, but really has very little to do with procurement.

Similarly, advising people how to buy AI technology well is fine, but that surely is no differ relay to “category-related procurement advice” around energy, laptops of anything else. It is not really a procurement policy issues.

The first point – on use of AI in writing bids – has probably gained the most comment and criticism. The PPN suggests that buyers should ask suppliers to disclose whether AI has been used in bid construction, but that the answer “should not be scored” as part of the evaluation process. However, if the supplier says “yes” , might that mean their scores for other questions will be reduced if buyers know AI was involved? This could be a legal minefield.  And as others have pointed out, asking questions “for information only” in tenders is not good practice, only increasing bureaucracy and cost for bidders and indeed buyers.

The PPN also mystifyingly mentions the “risks” inherent “if a bid writer has been used by the bidder”. Sorry? I mean, someone always writes the bid. I assume they mean an “external” bid writer, but in my experience such individuals usually take more care to reflect the organisation accurately than some poor sales person who gets landed with the task of writing the document! 

The key point surely is that any bid should reflect the organisation’s capability and experience accurately, and provide a proposal that is meaningful and realistic about the actual goods and service that will be delivered if that bidder is chosen. That applies whether AI was involved or not. Indeed, humans are just as capable as AI of making up nonsense to put into bids – in fact, I suspect humans, being more creative, are more likely to write lies or nonsense than AI.

Anyway, this is a badly thought-out PPN, written in haste I assume, and further clarification and development of the very different points discussed within it will surely be necessary.