Finance now sits in nearly half of software deals. How to get an AI purchase past the CFO

What G2 found

G2's 2026 Buyer Behavior Report, published on 22 July 2026, surveyed more than 1,000 B2B software buyers (data period and region not stated). It also interviewed more than 50 sales and marketing leaders. Finance involvement in software decisions rose from 31% to 46% in a year, in G2's words, and nearly half of buyers said their CFO had vetoed an approved deal in the last year (period and region not stated).¹ G2 runs a software review marketplace, so it has an interest in how software is bought.

The report also links that pressure to contract terms. Seventy percent of buyers said the pace of change pushes them toward shorter contracts, and buyers who had a deal vetoed pushed for terms under 12 months at 40% against 18% for others (data period not stated).¹ Preference for outcome-based pricing rose from 11% in 2025 to 23% in 2026, on G2's figures, with the fieldwork dates not stated.¹

The same buyers lean on AI to find software: 82% said they had sourced software recommendations from an AI chatbot in the last 24 months (fieldwork dates not stated).² AI shapes which vendors reach the list. Finance decides which one gets paid.

Why do AI purchases draw extra scrutiny?

AI purchases carry a cost finance cannot see in advance: usage. A seat licence is a fixed line. An agent or an AI feature priced per task, per token or per resolution moves with volume, and most companies are not yet watching it closely.

In KPMG's AI Quarterly Pulse, a survey of 204 US leaders at firms with $1 billion or more in revenue run from 28 April to 25 May 2026, 53% said they were deploying AI agents. Only 26% had full real-time visibility of their AI operating costs.³ If large firms with finance teams to match cannot see the running cost, a mid-sized company should assume its CFO will ask about it first.

Our view: most late vetoes on AI are not about the price on the order form. They are about an open-ended running cost and a result nobody has defined. Both can be fixed on paper before the vendor shortlist is final.

The one-page business case

A CFO needs five things on one page, in this order. Write them with the budget holder before you talk to vendors, then share the page with finance while there is still time to change the scope.

One page, five lines

1. Today's cost

   Volume a month, minutes per item, loaded hourly cost, rework. Your numbers, not a benchmark.

2. The result you will accept

   One metric with a threshold and a date, signed before the build: hours saved, exception rate, share of answer.

3. Total cost, capped

   One-time fee plus running cost per task at expected and peak volume, with a monthly ceiling.

4. The exit point

   The first commitment ends at a review. If the result is missed, you stop with what you paid for.

5. The owner

   A named person who watches running cost and results every month after launch.

Line one is the hardest and the most persuasive. Time a sample of the work, multiply by loaded cost and add the rework you already pay for. Our scoring model for choosing the first process shows how to get those inputs in an afternoon.

Line two should name one metric, not five. Finance will hold you to it, which is the point.

Short terms and outcome pricing, read carefully

G2's buyers want shorter contracts and more outcome-linked pricing, and both help a case through finance when they are designed well. A short first term caps exposure. An outcome link makes the vendor carry some of the risk.

Read the outcome clause closely, though. Ask who measures the outcome, on what sample and over what period. A vendor-measured metric with no method is a variable price, not a shared risk. Our survey of published GEO prices sets out what agencies and tools put on their price pages, which gives finance a reference point before a proposal arrives.

Failure risk is part of the case too. Gartner predicted in June 2025, more than a year before this post, that over 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls.⁴ A forecast, not an outcome, but your CFO may have read it. Lines two, three and four on the page answer it directly.

What this means for how we sell

Sigzen AI publishes fixed prices for each rung on our pricing page, and each first engagement ends at a defined deliverable. The AI readiness assessment ends with scored use cases. The automation sprint has acceptance criteria agreed before the build and hours saved signed off at the end. That structure is what this post recommends you ask any vendor for.

Our view: if a vendor will not put the running cost and the acceptance metric in writing before signature, finance is right to veto the deal. Ask early and save everyone the late surprise.