Every CFO I talk to right now is doing two things simultaneously: signing off on AI spend and quietly wondering if they'll regret it. The worry is a feature, not a bug.

Sixty-five percent of finance leaders are dedicating 10% or more of their 2026 budgets to AI and automation. In the same Billtrust survey, 59% of those leaders said the current wave of AI investment might be a bubble that won't deliver sustained long-term value. Nearly one in five describe themselves as "very concerned" about overinvestment.

Finance is writing the checks and worrying about them at the same time. In the receivables data we see across the market, that skepticism is what makes the spending disciplined enough to deliver returns.

Why the worry is the healthiest signal in the data

In the early stage of any technology cycle, the danger is uniform enthusiasm. Everyone buys, nobody measures, and the vendors with the best stories take the budgets. Finance leaders in 2026 are doing something different. They are spending aggressively while asking hard questions about every dollar. It's not a coincidence that the organizations with the best returns are the ones under the hardest internal scrutiny.

Seventy-nine percent of organizations using AI in finance report tangible returns, including directly reducing DSO or accelerating cash application, significant outcomes from technology that in many cases has only been at scale for 18 months.

The bubble concern asks whether the pace and scale of investment across the market is sustainable. Whether AI works is a different question, and conflating the two leads to bad decisions in either direction.

Why pulling back isn't an option — and what the returns actually require

Eighty-four percent of finance teams are already using AI in financial decision-making. The debate about whether to start is over for most organizations. Instead, the conversation has shifted to which deployments to defend, where to scale what's working, and which new use cases can justify the spend.

The cash environment makes pulling back harder. More than two-thirds of customers are paying slower than six months ago. Thirty-four percent of finance organizations have already cut headcount, and the teams that remain are carrying more. A 2020 operating model running against 2026 payment cycles and margins is how you lose ground quietly.

The deployments actually delivering returns share something in common. They aren't bolt-ons. They're the ones where companies redesigned who makes decisions, how work flows, and where accountability sits, with AI handling the volume work and humans handling the judgment calls. Most organizations are doing a bolt on and calling it a redesign.

The companies that will win are the ones weaving AI into the fabric of how their finance teams already work, so seamlessly that the end user barely registers it's there. Gmail has been AI-powered for years and nobody thinks about it. 

The accounts receivable professional at a lumber company in Ohio shouldn't have to understand machine learning to benefit from it. She should just notice that cash is arriving faster, disputes are resolving sooner, and her workload has gotten lighter. The bubble concern coexists with aggressive investment because the AI most likely to deliver returns is the AI that disappears into existing processes. Tools that get bolted on drift to the periphery, but the ones woven into the architecture stick.

What disciplined AI investment looks like

The organizations getting returns share a few common traits. Every AI investment is tied to a specific cash outcome: DSO, working capital, fraud loss, a financial line item. If it can't clear that bar, it belongs in an experimentation budget, not the operating budget. Deployment timelines are short and reassessed often. 78% of finance teams now revisit forecasts at least quarterly, and AI investments should be evaluated on the same cadence. Annual reviews are too slow for technology this volatile.

The leaders who keep asking whether the spending is justified are the ones building the muscle to redirect investment when something isn't working. And defense gets funded alongside offense. Fraud detection and deepfake defense have rocketed to the top as planned AI investment priorities, a category that barely existed a year ago. The same technology automating legitimate operations is being weaponized against the organizations using it, and finance teams that don't invest in AI-powered defenses will face AI-powered attacks unprotected.

Where this leaves finance leaders

Organizations that emerge strongest from periods of uncertainty tend to have made the uncomfortable investment at the right moment. In 2008, it was cloud infrastructure. In 2020, it was digital operations. The finance teams that get this cycle right won't have done it by waiting for certainty. 

The CFOs who are only spending are going to get burned. The ones who are only worrying are going to get left behind. Those that take a tactical and analytical approach to AI spend will thrive.