AI adoption may be widespread across U.S. finance functions, but a new study from Rillion suggests that adoption is creating an illusion of progress. Behind high usage and expectations, finance teams continue to struggle with limited trust, manual work, and emerging skills gaps.

The 2026 AI in Finance Report is based on a survey of 250 CFOs and finance leaders across the U.S. It was produced by Rillion, a provider of accounts payable automation software, and examines four gaps between AI ambition and reality: trust, skills, automation, and transformation. The report combines survey findings with perspectives from finance leaders and AI experts, including contributors from GitHub, Meltwater, Docusign, Foxit, Joe & The Juice, and CFO Alliance.

The findings point to a clear tension between adoption and trust. Sixty-eight percent of finance teams already use AI in their daily work, while another 28% are piloting or considering it. Yet only 39% of CFOs say they are comfortable allowing AI to act independently without human review. That caution tracks with how finance controls have traditionally operated, with outputs that affect payments and reporting reviewed before they are acted on.

“AI has moved incredibly quickly from something finance leaders were experimenting with to something many teams now use every day. But adoption alone doesn’t mean transformation,” said Daniel de Sousa, CEO of Rillion. “The interesting part of this research is the gap between how ready finance looks on paper and what’s happening underneath. Trust, skills and manual work are still holding many organizations back.”

Automation still relies on manual work

The gap is especially visible in repetitive areas such as invoice processing, a core accounts payable task that has long been a target for automation because of its volume and repetition. Nearly nine in ten CFOs report shortcomings with their current invoice capture and data extraction solutions. Almost half, or 45%, say human review is still required after invoices have been processed. That review step persists for practical reasons: extraction errors in invoicing can result in duplicate or late payments.

“Finance teams have been automating processes for years, but too much automation still depends on people stepping in when something changes,” said de Sousa. “The next step isn’t simply adding more AI. It’s building processes that can handle the reality of how businesses operate, without manual work creeping back in.”

Finance may have an AI skills blind spot

The report also highlights a disconnect between the skills finance leaders believe they need today and those they expect to matter in the future.

Only 21% cite a lack of internal AI expertise as a major barrier to adoption today. At the same time, 60% say understanding AI tools will become one of the most important skills in finance. That expectation is becoming harder to defer as AI features are built into the mainstream finance and ERP software teams already use, turning AI familiarity from a specialist topic into part of everyday finance work.

“We shouldn’t think about AI competence as learning how to use a particular tool,” said de Sousa. “The bigger shift is learning how to work differently: knowing what to automate, how to challenge the output and where human judgment still matters. Those capabilities take time to build, which is why finance leaders need to start investing in them now.”

Those using AI most expect the biggest transformation

Finance teams already using AI widely are more than three times as likely, at 51% versus 16%, to expect AI to transform most finance processes as those at the beginning of their AI journey.

“The people with the most hands-on experience with AI are also the ones who see the biggest change coming,” said de Sousa. “That should be a wake-up call for organizations still waiting on the sidelines. The scale of the transformation may be hardest to see when you’re at the beginning of the journey.”