Your buyer is writing your business case without you.

Somewhere right now, a director at a company you would like to sell to is pasting a prompt into a chat window. Something like: build me a business case for replacing our current process with AI tooling, including expected savings, implementation risk and a rollout plan.

The output is mediocre and coherent, which is a much lower bar than good, and enough to circulate. It will go into a slide, then into a meeting you are not attending, and it will set the number your product gets compared against for the rest of the cycle.

This is the part of AI-assisted buying that GTM teams have not adjusted to. The visibility problem gets all the attention. The bigger shift is that the buyer’s internal justification is now assembled early, quickly, and from whatever happens to be publicly available.

What the research shows

The behaviour is documented. Forrester’s January 2026 buying research, drawn from a survey of business buyers worldwide, describes generative AI as the starting point of the process, with buyers then leaning on internal colleagues and external influencers specifically to justify and de-risk the decision. Its title on the pressure buyers are under is blunter than most analyst prose: leaders face mounting pressure to justify every dollar spent.

Secondary reporting of the same survey puts roughly 47% of buyers using AI to build internal business cases before contacting any vendor, alongside 55% using it to compare vendors and 54% to research products. I have not been able to open the underlying report to confirm those three splits, so treat the exact percentages as indicative and the behaviour as established. The direction is corroborated by Gartner, which found 45% of buyers using AI during a recent purchase and 67% preferring to buy without a rep involved.

Why this is worse than losing a click

A lost click costs you attribution. A business case drafted without you costs you the frame.

Three specific things go wrong.

The model invents a baseline. Asked what a process currently costs, it will produce a plausible industry figure. If that figure is low, your value story is now competing against a phantom starting point that nobody in the room will question, because it arrived looking like research.

The cost model comes from someone else’s pricing. Whichever vendor has the clearest public pricing shapes the budget line, and if yours is hidden behind a contact form, the machine uses a competitor’s numbers to size your deal.

The risks get generic. A model asked about implementation risk returns change management, data quality and integration effort. Those are real, and they are also not the risks that actually kill projects in your category, which means the mitigation plan is aimed at the wrong things and the project fails for a reason nobody planned for.

By the time you are in the room, you are arguing against a document rather than starting a conversation.

Write the business case yourself and publish it

The response is uncomfortable for most enterprise marketing teams because it involves publishing things they prefer to keep in a controlled conversation.

Publish a real cost baseline for the problem, with your method shown. Not a benchmark report with a gate on it. A page that states what this process typically costs, how that figure was derived, and where it varies. If your number is the retrievable one, your number becomes the baseline. This is the highest-return page you can write this quarter.

Publish enough pricing to be modelled. Full transparency is not required. A shape is: what drives the price, the range for a mid-sized deployment, what pushes a deal to the top of the range. Vendors who refuse this hand the modelling to whoever did not refuse.

Publish the failure modes. A page describing the four reasons implementations of this kind fail, with what to do about each, will get retrieved and quoted more than anything else you write. It also arrives in the buyer’s risk section with your framing attached, which is worth more than a case study.

Publish the internal document your champion needs. Not a brochure. A five-slide structure they can put their own logo on: the problem in their language, the options including doing nothing, the cost comparison, the risks with mitigations, and the recommendation. Make it downloadable and editable.

That last one feels like giving away control. Your champion is already building that deck. Your only choice is whether they build it from your material or from a model’s best guess.

The number your buyer actually needs

Most vendor ROI content answers the wrong question. It calculates value created. Business cases get approved or rejected on a different basis: what happens if we are wrong.

An approver is weighing personal exposure. If this fails, how visible is the failure, how much is unrecoverable, and how quickly can we stop. A business case that addresses those three directly gets signed faster than one promising a larger return, and it is the part almost no vendor writes.

Which is why the evidence discipline matters here as much as the arithmetic. Forrester’s 2026 predictions flagged 19% of buyers feeling less confident after using AI tools because of unreliable information, and forecast that B2B companies would lose more than $10 billion through ungoverned use of generative AI. Some of that loss is going to be business cases built on numbers that were never true, discovered at renewal.

If your figures are the ones that hold up under checking, you win twice. You shape the case going in, and you are the vendor still standing when somebody audits the results twelve months later.

Start here

Take your last three lost deals where the reason was recorded as no budget or no business case. Ask the buyer, if you are still on speaking terms, how the internal justification got built and what numbers it used.

You will usually find the numbers came from somewhere other than you, and that they were wrong in a direction that cost you the deal.