Most bid content stops at submission. Guides tell you how to find the tender, how to answer the question, how to price the work. Almost nothing tells you what happens on the Monday after you win, which is a strange omission, because that is when the promises you made stop being persuasion and start being obligations.

The Procurement Act 2023 changed the shape of that Monday. Two of its sections matter more than the rest, and they are frequently run together as though they were one rule. They are not.

Two duties, not one

Section 52 is about setting the indicators, before anything is signed. It applies to a public contract with an estimated value of more than 5 million GBP. For those contracts, at least three key performance indicators must be set before the contract is entered into, and the indicators must be published. It came into force on 24 February 2025.

Section 71 is about assessing performance against those indicators, afterwards. Performance must be assessed against the indicators set under section 52, at least once every 12 months, and again when the contract is terminated. Information about the assessment must be published. It came into force on 1 January 2026.

The dates are the part worth pausing on. Section 71 has been in force since the start of this year, which means the first full annual assessments under it are being made and published right now, against indicators that were set and published under section 52 as far back as early 2025.

If you won a contract over 5 million GBP in 2025, somebody is currently writing a public assessment of how you have performed. That is not a future compliance problem. It is this quarter.

What this actually changes for a supplier

The published assessment is the part that changes behaviour, and it changes it in a way that is easy to underestimate.

A private contract review is a conversation between two parties who both want the relationship to continue. A published assessment is a durable record that your next buyer can read. It sits alongside the notices already published about the award. Anyone evaluating your next bid can look at how the last one went, and increasingly they will, because it is now there to look at.

So the incentive shifts. It is no longer enough to write a strong bid and then manage the relationship well. The evidence has to exist, in a form somebody can point at, at the moment the assessment is written.

The gap where it goes wrong

Here is the pattern we see most often, and it is not a compliance failure. It is a filing failure.

The bid promised something specific. Perhaps a number of apprenticeships, a local spend proportion, a response time, a named methodology. That promise was written eighteen months ago by a bid team, under deadline, and it was good. It won.

Then delivery started. The delivery team is not the bid team. They have the contract, and possibly the specification, but they very rarely have the bid. The commitments are sitting in a submitted document nobody has opened since the award, in a folder named after a tender reference that means nothing to the people now doing the work.

When the assessment comes round, somebody has to reconstruct what was promised, from a document they have to go and find, and then evidence it retrospectively from records that were never kept with that question in mind.

The commitment was real. The delivery may well have been real too. What is missing is the thread connecting them, and that thread is what an assessment is.

What CrowMark does about it

This is the part of the problem CrowMark was built to close, and it is worth being precise about which bits are automated and which are not.

The commitments come out of the bid and into the contract record. When a bid is won, the commitments it made become tracked items on a supplier profile rather than sentences in a submitted PDF. They carry the wording that was actually offered. The delivery team gets the promise without having to find the document.

Evidence attaches to the commitment it belongs to. Post-award delivery evidence is filed against the thing it proves, not into a general folder. When an assessment is due, the question is which commitments have evidence and which do not, which is a question you can answer in an afternoon rather than a fortnight.

The section 52 and section 71 duties are checked as duties. They are two separate checks in CrowMark because they are two separate duties in the Act, with different triggers. Merging them is exactly the mistake that leaves an organisation confident it is compliant because it did the first half.

Concerns and corrective actions hang off the commitment. When something slips, it is recorded against the specific indicator it affects, so the eventual assessment has a history behind it rather than a verdict arriving from nowhere.

On the buyer side the same record works in the other direction. Scorecards, periodic review sessions and a risk radar all read from the commitments made at tender, so a review meeting starts from what was offered rather than from what somebody remembers being offered.

Where the AI helps, and where we stop it

CrowMark uses AI throughout the bid, and the design constraint is the same everywhere: it is grounded, and it cites.

Drafting is grounded in your own material. Answers are drafted from your previously submitted bids, your answer library and the evidence you have uploaded, not from a general model impression of what a good answer sounds like. Each suggestion cites what it drew on, so you can check the source before you accept the sentence. The model proposes. You decide.

A long tender pack gets briefed before anybody writes. An invitation to tender running to hundreds of pages is read and summarised into what is actually being asked, what is mandatory, what the deadlines are and what the evaluation weightings say. That is a reading task with a right answer, which is the kind of task this technology is genuinely good at.

The AI is not allowed to invent a number. That is enforced rather than hoped for. A figure that appears in a drafted answer has to come from something you supplied. It matters most in social value, where a plausible invented figure is worse than a blank, because a blank gets noticed and a plausible number gets submitted.

And where the ground does not exist yet, we say so. PPN 026, the new edition of the Social Value Model, was published on 5 August 2026 and applies to central government procurements commenced on or after 1 January 2027. Its detailed sub criteria have not been published and are expected in autumn 2026. CrowMark offers PPN 026 and makes it the default on new work, because that is what future tenders will ask for, and it does not score against sub criteria that do not exist. Anyone selling you a finished PPN 026 scoring engine today is scoring against criteria nobody has published.

That last one is the same discipline as the first three. A tool that answers when it does not know is not saving you work. It is moving the error somewhere you will not find it until an evaluator does.

What to do this quarter

If you hold a public contract over 5 million GBP awarded under the Procurement Act 2023, three things are worth doing before your next assessment falls due.

Find the indicators. They were published. Go and read the ones attached to your own contract, in the words they were published in, rather than the version in your internal summary.

Map each one to where its evidence lives. Not who is responsible for it. Where the proof is, today. Any indicator where that answer is a person’s name rather than a location is an indicator you cannot currently evidence.

Read your own winning bid again. All of it. The commitments in it are the ones you will be assessed against, and it is very common to find one or two that the delivery team has never seen.

None of that requires software. It requires somebody to spend a day on it. The software is what stops you having to spend that day again every twelve months, for the life of every contract you hold.

Frequently asked questions

Which contracts have to publish key performance indicators?

Section 52 of the Procurement Act 2023 applies to a public contract with an estimated value of more than 5 million GBP. For those, at least three key performance indicators must be set before the contract is entered into, and the indicators must be published. Below that threshold the section does not bite, though an authority may still set indicators as a matter of contract.

How often is performance assessed and published?

Section 71 requires performance to be assessed against the indicators set under section 52 at least once every 12 months, and again when the contract is terminated. Information about the assessment must be published. Section 71 came into force on 1 January 2026, so the first annual assessments under it are being made and published now.

Are section 52 and section 71 the same duty?

No, and treating them as one is a common mistake. Section 52 is about setting and publishing the indicators before the contract starts. Section 71 is about assessing performance against them and publishing that assessment afterwards. They have different triggers and different dates, and section 71 came into force nearly a year after section 52.

What happens to the commitments made in the winning bid?

Legally they form part of what was offered and, depending on how the contract is drafted, part of what was agreed. Practically they become the thing somebody has to evidence later, often a different somebody from the person who wrote them. That gap between the writing and the evidencing is where most delivery disputes start.

Sources

The statutory positions above are taken from the legislation itself rather than from summaries of it.

Every claim about what these sections require is on our sources page, with the duty stated in the instrument’s own terms.