AI is in the bidding room now, and it isn’t leaving
Love it or hate it, the box is open. AI is being used to write quality submissions, and increasingly it is being used to score them. There is no putting that back.
A number of major framework hosts have now said openly that AI will play a part in evaluating quality responses. At the same time, almost everyone bidding the big frameworks is using it to write those responses. Both sides of the table have picked up the same tool. That changes the game in ways the market is only starting to feel.
The playing field just got a lot flatter
The most immediate effect is levelling. Highly technical, demanding quality questions used to be where the well-resourced bidders pulled away. A tier 1 with a full bid team, a library of past submissions and the people to polish every answer could score consistently where a smaller firm could not. That gap is closing fast.
An SME can now produce a quality response that scores 100% on a question it would previously have struggled to answer well, let alone consistently across a full submission. When everyone can hit top marks on quality, quality stops being the differentiator it was. And when the technical scores bunch at the top, the decision gets pushed onto price. The firms with lower recovery costs win that race, and that is rarely the tier 1.
So you get two things happening at once. Scores compress at the top, and commercials get driven down underneath them. One loose mark, half a point dropped on a single answer, can now cost you a place on a multi-year, multi-billion-pound framework. That is four to eight years of regional, sometimes national, position gone over a margin that used to be comfortable.
CPS2 is the warning shot
You do not have to look far for evidence. Government Commercial Services’ Construction Professional Services 2 procurement tells the story plainly.
That framework was awarded late, and it was late for a reason. The volume of applicants and how highly they scored across the board created a logjam, with significant push-back from bidders along the way. Look at lots 1 to 4. They were set up for 15 places each. They have ended up awarding close to double that on each of those lots, because the scoring simply did not separate the field the way the model assumed it would.
Then look at some of the names that landed places. No offence intended to any of them, but are they all realistically national suppliers, with the depth of resource to service a framework of that size and reach? Some clearly are. Others scored their way onto a national framework on the strength of a written submission, and that is a different thing entirely from being able to deliver against it.
That is the crux of it. When the written submission can be made near-perfect by anyone, the written submission stops telling the client who can actually do the work.
A year of renewals, and some of them will hurt
We are heading into a period with a lot of major contractor-led frameworks up for renewal. Apply what CPS2 showed us and the outcome is not hard to predict. There will be casualties. Incumbents who walked into renewals confident of holding their place are no longer in the driving seat.
With the value of work involved, that means some very disgruntled suppliers and some very nervous framework hosts, with their legal counsel on speed dial. Procurement challenges follow upsets, and upsets are now far more likely when the scoring no longer behaves the way it used to.
Add the wider backdrop. Local government reorganisation is redrawing who commissions what. Central government keeps turning over. If enough major upsets land in a short window, the tried and tested routes to market may stop looking so safe, even to the players with large business development and marketing functions behind them. When the certainty goes out of a framework strategy, people start asking whether the framework was ever the moat they thought it was.
It is worth asking why so many hosts are happy to let AI mark the work. The honest answer is that it saves them money. Many can no longer charge suppliers an annual subscription to hold a place, so margins on running a framework have thinned. Several have pushed project levies up to compensate. An evaluation process that runs faster and cheaper is an easy yes in that context.
Where this is actually driving us
Step away from frameworks and look at individual schemes and programmes, and you see the counter-move already forming.
Clients procuring a specific project have a much more vested interest than a framework host. They are not assembling a panel, they are choosing the team that will actually deliver their building. They can see what is happening. It is too easy now to talk a good game through a polished technical response, so they are asking for more air time with bidders. More interviews. More scored and non-scored presentations. More chances to test whether the experience on the page is real and deliverable by the named team in the room.
Who can blame them. The written word has lost some of its power to tell them apart, so they are reaching for the one thing AI cannot fake on a supplier’s behalf: the people.
We think the human element rises from here, and rises fast. Suppliers will compete harder than ever for the best talent in the market, especially the people who present well and carry strong client networks. The edge will not be how high your AI can score or how far your rates can fall in the short term. It will be how well you know a client team as individuals, and how much work you have put in to pre-position with them long before the bid lands.
That, in our view, was always the real driver of bidding success. AI has just stripped away everything that used to disguise it.
Which is where early intelligence earns its place
If the human relationship is the edge, the question becomes a timing one. You cannot build genuine standing with a client team in the weeks after a tender drops. By then everyone is in the room and the conversation is a competition.
The advantage goes to whoever started earliest, knew the scheme was coming before it was visible to the field, and used that lead time to get in front of the client and the consultant team while there was still a relationship to build. That is the work AI cannot do for you, and it is exactly the window we exist to open.
Early means early, and in a market where everyone can write a perfect answer, early is most of what is left.