Every tender we’ve submitted in the last few months has been delayed. And not just those: I get updates about every tender I enquire about, and they’re ALL running late. The explanation we’re given is nearly always the same: buyers received far more submissions than they expected and didn’t have the capacity to evaluate them on the original timetable.

I don’t blame the evaluators. However, the delay isn’t free, and small suppliers have very little room to absorb it.

When we bid, we commit to a price. We name associates and ask them to keep time free. We plan other work around the potential start date. A larger supplier may have enough other work and people to accommodate a contract slipping by a few months, but we notice straight away. Good freelancers won’t sit around waiting, and the assumptions behind a price can change while a decision is outstanding.

We’re being assessed on a response written for a timetable that no longer exists.

The Procurement Act was meant to make public procurement more accessible to SMEs, and there are some good parts to it, but that needs to extend beyond finding opportunities and submitting bids. If the process depends on suppliers holding prices and capacity for an uncertain period, those who can afford to wait have an advantage.

I really think AI is contributing to this, both by making it easier to bid and by making it easier to produce long responses that say rather less than their length suggests.

Writing a tender response used to take a substantial amount of time: reading and understanding the documents, working out what the buyer wanted, finding the evidence, writing it, cutting it to the word limit and checking it. For a small company, that could mean days of work on something you might not win, and the effort involved was part of the decision about whether to bid at all. No wonder many SMEs wouldn’t bid for public sector work.

You still need to do that thinking. AI doesn’t remove it, but it does make it easier to produce a response without having done some of the creative and critical thinking the work requires. Upload the documents to your LLM of choice, ask for answers against the criteria, add some information about your business, and you can have something that reads like a credible submission remarkably quickly. If a bid feels like a day’s work rather than a week’s, the calculation about whether it’s worth a punt changes substantially.

That can help a capable small supplier who previously struggled to find time to bid. It can also encourage applications where the fit is weak and the supplier would previously have decided the effort wasn’t justified. Making submission easier doesn’t necessarily make the resulting competition more useful. We are seeing this applied to FOI requests. The effort needed to produce and submit a request can fall considerably without reducing the work needed to respond to it.

I can’t demonstrate how much of the increase in the procurements we’ve entered is attributable to AI. But more submissions and pressure on the people dealing with them are precisely what I’d expect when producing the material becomes easier and the capacity to assess it doesn’t increase.

There’s also the material itself. I think AI is contributing to overlong responses, full of plausible commitments, comprehensive descriptions and repeated assurances. A word limit becomes space to fill and the answer covers everything the supplier might conceivably do, while leaving the evaluator to work out what they actually will do and what evidence supports it.

That creates work even where the response stays within the limit. Five hundred words containing a clear commitment, a relevant example and an explanation of how delivery will work are different from five hundred words of polished generalities. Both have to be read. The second may take longer to assess because the evaluator has to establish whether there is anything substantive behind the language.

So buyers may be facing more bids and more effort to extract useful evidence from each one. An evaluator still needs to assess the response against the criteria, moderate the scores and record the reasons. None of those requirements are reduced because the supplier generated the answer quickly.

Using AI to mark the responses might seem like an obvious solution. I can see the appeal, but someone still needs to explain why the evidence justified the score. “The tool gave it a 3” isn’t a reason, and a plausible explanation generated alongside it doesn’t establish that the judgement was sound. If fluent, generic answers are part of the problem, I’d be concerned about using an evaluation approach that might reward those same characteristics.

I’d start with what buyers ask suppliers to produce. “Describe your approach to stakeholder engagement” invites an extensive account of effective practice but doesn’t necessarily tell you whether the supplier can do the work. Ask about your constraints, the decisions the supplier would make and what they have done somewhere comparable. Specific claims still need checking, but they give evaluators something to assess.

Fewer, better questions would help too. Remove overlap and ask for the evidence needed to make the decision. Tighter word limits can discourage repetition, although making them too restrictive just produces dense answers that are harder to read. Suppliers also need to take responsibility for editing. Vendors, a PSA: being able to generate another paragraph doesn’t mean the buyer needs it.

I can see buyers making more use of the competitive flexible procedure to shortlist against published criteria before asking for detailed tenders. More market engagement should also inform evaluation capacity and timetables. Forty interested suppliers won’t necessarily mean forty bids, but it’s a reason to question a process planned around eight.

Raising turnover, insurance or experience requirements to reduce the bidder pile also risks excluding smaller suppliers without establishing that the remaining ones are better able to deliver. Those requirements need to reflect the contract, rather than the buyer’s shortage of evaluation capacity.

And when timetables slip, suppliers need to know early. The people we named may take other work, the proposed start date may pass and the basis of our price may change.

Making bidding easier achieves very little for small suppliers if the resulting process requires us to carry more uncertainty for longer.