Can AI increase your chance of winning a bid?

Yes, but not by using the same AI as the buyer. Three things move the odds: what their machine can read about you, what your machine knows about you, and whether your bid survives a machine before it reaches a person.

Every day I hear brilliant and I hear absurd.

“One team swapped entirely from ChatGPT to Copilot, as that’s what the authority uses. They paid a consultant, as they were told that using the same AI would make them in sync and give them more chance of winning.”

I call it the same-model fallacy. Using the same AI as the person you are selling to is not going to make any respectable difference. But can we use AI to increase our chances of winning a bid?

With that said, here are three levers I would hang my hat on. Each one kills a bad habit.

Lever 1: Be legible to the AI your buyer actually uses

There are three moments when a buyer’s machine reads about you and reports back.

1. Before the notice, when someone asks an AI “who does this technology in the North West” and a longlist forms.
2. During evaluation, when due diligence and competitor comparison get run against whatever is public about you.
3. The Central Digital Platform share code, the one structured record of you they actually hold, which gov.uk says is “not accessible in any other way”.

A buyer’s AI can’t see your AI. It can see your website. So here is the check I would run today. Ten minutes. No developer.

• Test first, to get a baseline. Ask a few AI chats a question a buyer would ask: “Who supplies X in Y region?” If you aren’t in the answer, that’s the longlist you’re not on.
• Two URL checks: yoursite.com/robots.txt, to see whether an AI can get in, then yoursite.com/sitemap.xml, to see whether there is a map for machines.
• As a rough rule of thumb, Copilot reads Bing, so Bing Webmaster Tools and IndexNow matter. Copilot’s web search can be turned off by an admin, and then it only sees what is already in the user’s tenant: past submissions and the share code.
• And remember, there is a lot of talk about GEO. That does not mean SEO is dead!

Market analysis is increasingly being done by AI agents, before anyone speaks to you. If the agent can’t read about you, you aren’t on the longlist it’s building. If it can read about you, you get to shape what it can say.

None of this will win a bid on its own, but it can decide whether you are in the room.

I checked 500 transport companies this month; 92% have never written a line for an AI to read. The full picture is in The Invisible Industry.

Want your own company’s score? Message me and I will send it.

Lever 2: Stop chatting to it, give it your brain

The habit this kills is the fresh session. Someone pastes a tender question into a blank chat, gets a confident answer, and it goes in the bid. AI is great, right?

Well, that chat most likely doesn’t know everything about your company, and it has filled in its memory gaps with something plausible, not factual.

A true story. A commercial director asked ChatGPT for an average bus driver salary in a specific region. It gave him an answer. It was confident, and it went in the draft. The answer was guessed. It was 7% away from the truth and it would have cost them over £500k. He caught it, and it became the trigger point for us here at economAIcs.

When you are working with your AI model, retrieval and memory are everything. Your past bids, KPIs, method statements, engineering specs. You want the model to cite your own material, not invent something to give you an answer. Hallucination is a real risk in the transport bidding world. Protect against it.

Three questions to ask an AI today:

1. “What was our on-time performance on our last contract of this type?” If it answers without your own document as the source, it’s guessing. If it doesn’t answer at all, it has no brain. Either way, that is the gap.
2. “Where did that number come from?” Ask it after every figure. If it can’t cite a source you trust, delete it.
3. “Are you training on this?” Then go and check the setting yourself!

In the bidding world, I see a point where your agent will read the ITT, map every question to your evidence library, flag gaps, compare against your competitors, identify the route to win and leave an audit trail that an evaluator can follow.

That point isn’t in the future. It is what we are building now. YourTransportBrain: one brain, bespoke agents. Designed to put you ahead.

Lever 3: Write for two readers, because your bid will be parsed before it’s read

The habit this kills is writing to impress. Bid teams polish prose for an evaluator they assume is tired and grateful for a good read. Fluent writing has never scored a point on its own, and Tussell said as much in January: AI-generated content does not affect scoring, and fluent writing alone does not improve outcomes.

PPN 017 already asks you to disclose AI use, “for information only” (see How PPN 017 affects transport bids). My honest stance is that buyers are preparing to handle AI-assisted bids at volume. That also means checking them at volume. Evaluation stays human. My expectation is that what reaches the human is changing.

A machine first pass doesn’t read your bid. It checks it. Four things, roughly:

1. Does every question have an answer where it was asked?
2. Does every claim have evidence next to it?
3. Do the numbers agree with each other across all parts of the bid?
4. Does what you claim about your business match the public record?

Structured, evidenced, cross-referenced and consistent survives that. Beautiful prose, with a number that disagrees with itself on page 40, does not.

So run it yourself. Take your last submitted bid and give it to the AI you built as part of lever 2. Ask it three things:

1. “List every question in the ITT and where in our response it is answered.” Anything unmapped was a compliance gap.
2. “List every number that appears more than once and tell me if they agree.”
3. “List every claim we made about a contract, KPI or accreditation and tell me if it can be checked publicly.”

That is red-teaming your own bid against a machine reader. My expectation is that by 2027 it is standard practice on the supplier side. I don’t see anybody doing it this way.

But we do. Send me the ITT. I will red-team it for free.

One structured truth

I have spent my whole career saying “make it easy for someone to buy from you”, and it has never rung truer than now.

In the bidding world, every company has two descriptions. One is your website, written by marketing, and increasingly read by the buyer’s machine. The other is your bid library, read by yours. In most companies I meet, the two have never been introduced.

The next real edge in bidding is not a better model. It is one structured truth: a single machine-readable record of what your company is, what it has won and what it can prove, from which both descriptions are built, so the two can never contradict each other. Every KPI claim, every site, every accreditation, every contract, in one form. Publish the public half. Feed the whole of it to your own AI.

Every public contract you have ever won is already on the record, through award notices. Your buyer’s machine can read that record today. The only question is whether your bid and that record describe the same company.

My expectations, plainly. By 2027, red-teaming your own bid against a machine reader is standard supplier practice. The award-notice record becomes the layer an evaluator’s assistant checks a bid against before a person reads a word, and the scoring stays human. And the Central Digital Platform share code quietly becomes the buyer’s canonical picture of who you are, which means most bidders are being judged on a record they have never read back.

Bids will still be marked by people. My expectation is that they will be checked by machines first, and the bids that make checking easy are the ones that reach the evaluator’s desk with nothing to explain.

That is what economAIcs is for. One brain for your bid team, built on your own material, that makes your company the easiest in the sector to verify. Because the easiest company to verify is the easiest company to award a contract to.

Make it easy. Be the easiest company in the sector to award a contract to.

Benjamin Ross. Practical application of AI in public transport. Founder, economAIcs. Benjamin Ross is the founder of economAIcs, the company behind YourTransportBrain.

Related: How to Win Public Transport Tenders: Lessons From the Bid Table, The Invisible Industry, How PPN 017 affects transport bids: AI transparency in central government procurement and AI in UK Public Transport: what’s actually real in 2026.