The CFO Is Now the Most Important Person in Your Funnel

G2 asked more than a thousand B2B software buyers this year whether they had watched a CFO reverse a purchase their company had already approved. Nearly half said yes, inside the past twelve months.

The finding sits in the 2026 Buyer Behavior Report, which also drew on interviews with over fifty sales and marketing leaders. It describes something most marketing teams have no way of seeing. A champion does everything the playbook asks. They find you, shortlist you, sit the demo, drag security review over the line, build the internal case and get a yes. Then finance reopens a decision everybody thought was closed.

In the CRM, that deal is logged as stalled, or slipped, or lost to budget. In the loss review it becomes a story about timing. Nobody writes down that the buyer had already picked you.

We have spent a year on this site arguing that the funnel is blind at the front. Buyers arrive with a shortlist already formed, 92% of them before any formal purchase process begins, and 41% with a single vendor in mind. The G2 data says the blindness runs the other way too. The decision gets made somewhere you cannot see, and then it gets unmade somewhere you cannot see. What your funnel has clear visibility into is the middle, which is the part that no longer decides anything.

Three things landed on the same desk this month

The 2027 planning cycle opened a few weeks ago and the mood going in is expansive. Forrester's budget planning research has close to nine in ten B2B marketing decision-makers expecting their investment to grow over the next year, across technology, headcount and programs. Bigger asks are being drafted right now.

Q4 is also the densest renewal window on the calendar. Procurement platform Tropic puts more than 30% of software renewals in the final quarter, and its data shows what lead time is worth: companies that open renewal talks six months out see up to 39% better savings, against 22% for those starting at sixty days. Every buyer you are courting this quarter is being coached, at the same time, to squeeze every vendor they already pay.

Then there is the budget line that just blew up. Finance leaders in the SaaS community report that 2026 AI budgets, most of them set last October, ran out well before the year did. The open-checkbook period for AI spend is closing during the same weeks your champion is asking for a signature.

The CFO reversing your deal in November is the person writing the budget that decides whether you get another shot in March.

Finance spent a decade losing this fight

Software buying decentralized for fifteen years, and that was the commercial premise of SaaS. Get a card number from a department head, land, expand, let procurement find out later. It worked well enough that Zylo's 2026 SaaS Management Index now puts 81% of software spend under the control of business units, with individual employees accounting for a little under 4%. The average company runs close to three hundred applications.

One number in that same index explains the swing back. Average SaaS spending rose 8% year over year while portfolios grew 2.2%. The tool count barely moved and the bill did, which means companies are paying considerably more for roughly what they already had. Zylo sells software management, so read the framing accordingly, though the gap tracks with what procurement teams are reporting elsewhere.

Show a finance leader that number and software stops being a growth input and becomes a leak. Per-tool caps are appearing. ROI cases are becoming a precondition of approval rather than something assembled afterwards. For anyone selling AI capability, buying authority is moving out of the technical teams that championed it and into procurement.

Gartner handed finance the analytical cover. Its forecast that more than 40% of agentic AI projects will be cancelled by the end of 2027, driven by escalating costs, unclear business value and weak risk controls, is now furniture in board decks. Gartner analyst Anushree Verma has been direct that much of the market is early-stage experimentation dressed as production capability, and has named "agent washing", the rebranding of chatbots and assistants as agents, as a distortion in how the category is sold. The forecast dates from mid-2025 and has been recycled hard since, which has done nothing to slow it down. It gives every finance leader in the market a defensible reason to wait.

The approval meeting is a brand test

Eight in ten buyers in G2's research have pulled software recommendations from AI tools in the past two years. Discovery collapsed into a prompt. Evaluation has since passed research to become the longest stage of the buying journey for the first time, and buyer concern about internal resistance to AI adoption climbed from 16% to 29% in a single year.

Watch what happens to the decision itself. When an AI assembles the shortlist, everyone on it arrives pre-validated, and all three or four finalists clear the functional bar. Your champion can no longer justify a choice on strengths the alternatives appear to share, so they quietly swap the question. They stop asking which tool is best. They start asking which one is hardest to get burned by.

This is the argument we keep making, arriving from an unexpected direction. A risk review is a brand problem. When the specs converge and the demo proves nothing the competitor's demo did not also prove, what remains is whether the buying committee believes your company will still be operating, supported and worth defending in three years. That belief was built long before the meeting, out of things no attribution model captured. Nobody has ever been fired for choosing the vendor everyone in the room had already heard of.

The unglamorous stall points are worth naming too. Research from Cleverbridge, reported by MarTech, found roughly three quarters of buyers say purchases get badly delayed or abandoned over internal approvals and back-and-forth with vendors. Thirty-six percent point at waiting for a price. Thirty percent cite negotiation over terms. Those are seller failures, arriving at the exact moment a deal is most fragile.

What your champion actually carries into the room

Take an inventory of what your team shipped last quarter. Case studies written for a practitioner. Pricing that requires a call. A security overview behind a form. Proof of value measured on a horizon that suits your renewal cycle rather than the buyer's fiscal year. All of it assumes an audience that already believes you, and that assumption holds until the moment somebody has to defend the spend to a person who does not.

A CFO needs total cost modelled at realistic volume, including the production costs that demo economics leave out. Compliance documentation they can download instead of schedule. Proof of outcome on a timeline finance will accept, which is almost never the timeline in your case studies. A quote inside a day, given that a third of stalled deals are sitting on paperwork nobody thought to measure.

None of that will look like a campaign, and it will not show up on a Q4 dashboard. It is still the job.

Where this argument runs out

Seven in ten buyers in G2's data say the pace of AI innovation is pushing them toward shorter contracts. Only 9% are comfortable letting an AI agent execute a purchase inside approved guardrails, and just 2% would allow it without pre-approval. Buyers are holding back for good reasons. Committing to three years in a category that reinvents itself every six months is a bad trade, and control over spending is the last thing a finance function surrenders.

No volume of enablement content fixes a pricing model that asks a CFO to underwrite that risk. If your champions keep dying at the same gate, the honest read may be that your packaging is what's being rejected. That is a marketing problem as much as a finance one, and it belongs in the plan being written this month rather than a QBR next summer.

Whatever gets built between now and December is what your champion carries into that room. Most of it currently stays on their laptop.