Three years after ChatGPT launched and changed how everyone talks about AI, here is where most finance teams actually stand: experimenting a lot, winning a little.
A new report from the April 2026 issue of the Journal of Accountancy — the AICPA’s official publication — puts a number on the disconnect. Close to 60% of finance teams are either piloting or fully implementing AI projects. But only 7% of CFOs say they are seeing a strong impact from that investment.
That is a brutal gap. And it is not a technology problem. The tools are capable. The gap is almost always about how organizations actually deploy them.
The Anecdote That Explains Everything
The Journal of Accountancy piece includes a case study that finance leaders will recognise immediately. A team used ChatGPT to consolidate chart-of-accounts data across 50 business entities. It worked — correctly handling the first ten or so — then unexpectedly changed its approach without explanation.
That is not a failure of AI capability. That is a failure of implementation discipline. The team had not established consistent prompting, validation checkpoints, or human review gates. They handed a powerful tool to someone who had not been trained to use it reliably, then were surprised when the output drifted.
This pattern repeats across finance departments, operations teams, HR, and customer service. Organizations rush to get AI into workflows because leadership wants to show AI progress. But no one asks the harder question: what does good AI output look like here, and how will we know when we have it?
Why 7% Sounds Low But Is Actually Accurate
The 7% figure comes from a Gartner report cited in the AICPA article. It might seem pessimistically low, but it tracks with what practitioners are reporting on the ground.
The gap between piloting AI and getting real ROI from it comes down to a few consistent failure modes:
The wrong starting point. Most teams start with whatever AI tool is available — ChatGPT, Microsoft Copilot, a point solution from their ERP vendor. The problem is they rarely start with a clear definition of the outcome they are trying to achieve. They experiment with the tool and hope an ROI story emerges. It rarely does.
No one owns the implementation. Running an AI pilot is often someone’s side project. It gets squeezed between regular responsibilities, handed off across teams, or quietly deprioritised when the initial enthusiasm fades. Strong AI ROI requires someone who owns the outcome and has the authority to change the workflow around the tool.
Data quality problems get exposed, not solved. AI tools are very good at surfacing problems in your data. What they cannot do is fix those problems for you. Teams that hit messy data mid-implementation often stall the entire project rather than addressing the underlying data hygiene issue. The AI becomes the villain in a story where the real problem was always the data. Companies with strong data literacy consistently outperform those without it — and that gap widens when AI enters the picture.
Skills do not keep up with tools. Finance teams can access AI tools immediately. But understanding how to write reliable prompts, validate AI output, build review processes, and avoid overconfidence in generated results — those skills take time to develop. Without that foundation, the tools get used inconsistently and the results vary wildly.
What This Means for Business
The AICPA article frames this primarily as a training and implementation challenge for finance professionals. That framing is right. But the implications go wider.
If 60% of finance teams are running AI experiments and only 7% are seeing genuine returns, CFOs and business leaders need to ask honest questions about their AI strategy:
Are we measuring outcomes or activity? Counting the number of AI tools deployed is not a measure of progress. The only measure that matters is whether workflows are faster, cheaper, more accurate, or more scalable than they were before.
Do our people understand the tools they are using? Not deeply — you do not need a data science PhD to use AI effectively in finance. But teams need enough literacy to know when AI output looks wrong, how to prompt for consistent results, and when human review is non-negotiable. That baseline is not being built at most organisations.
Are we building on clean data? AI tools amplify what is already there. If your chart of accounts, master data, or reporting structures are inconsistent, AI will not fix them — it will make the inconsistency more visible and more expensive.
Is anyone accountable for the outcome? Pilots without owners become shelfware. Every AI initiative needs a defined owner, a defined outcome, and a defined timeline for making a decision about whether to scale or stop.
The Real Opportunity in the Gap
The 60%/7% stat is actually good news for organizations willing to invest seriously. The tools work. The problem is implementation and capability — and those are solvable problems.
Teams that close the gap between AI adoption and AI ROI share a few common traits. They pair tool deployment with deliberate skills investment. They build data governance before they build AI workflows. They treat AI implementation like any other change management challenge — not a technology project, but a people and process project that happens to use technology. Before committing to AI deployment, it’s worth checking whether your organisation actually has the prerequisites in place — most don’t realise they’re missing one until they hit the wall mid-project.
For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.
The 7% does not have to be your ceiling. But closing the gap requires more than downloading another AI tool and seeing what happens.
If your finance team — or any team — is sitting in that 93% who have experimented with AI without strong results, the answer is almost never a better tool. It is a clearer process, cleaner data, and people who actually understand how to use what they have been given.
That is the work. And it is entirely within reach.