Developers got AI first. Tools such as GitHub Copilot became the standard, and startups in this field achieved billion-dollar valuations. Now it is becoming clear that finance departments are next. And it is happening sooner than most people think.
American investment firm Battery Ventures decided to test this assumption. Rather than relying on impressions and estimates, it collected data from 129 CFOs and senior finance executives at companies with annual revenues ranging from $50 million to more than $5 billion. The survey ran from December 2025 to February 2026. And what did it reveal? A picture full of contradictions.
There Is Interest, but Results Are Still Lacking
On paper, the numbers look optimistic. 17% of the CFOs surveyed already have AI in production, another 34% are actively testing it, and 28% plan to do so. Only one-fifth are still at the "we're thinking about it" stage.
But beneath the surface lies an uncomfortable reality. Of everyone who has tried AI, only 4% report pilot success rates above 50%. The remaining 96% have simply failed. That number is still far too high. How is that even possible? The interest is there, budgets are being made available, and the technology is improving. The problem lies elsewhere.
Two Obstacles Holding Back the Entire Market
Battery Ventures identified two specific barriers preventing widespread adoption.
The first is model inaccuracy and unclear return on investment. A full 71% of respondents identify AI model inaccuracy as their main concern. One-third cannot justify the investment with numbers. Paradoxically, 95% of respondents are willing to pay a premium for AI, but 44% are concerned about the total cost. CFOs understand the potential, but they need evidence, not promises.
The second obstacle is integration and data quality. This is where one of the survey's most interesting findings emerged. The question was: do you want AI from your existing system vendors, or a new system from AI-native companies? The answer was clear. 77% of CFOs want new AI tools that connect to their existing systems without interfering with them. Only 15% would be willing to replace their entire system with a new one built from the ground up around AI.
But that creates a problem of its own. Corporate financial data is scattered across legacy ERP systems, banking integrations, Excel spreadsheets, and various payment tools. And half of respondents rate the quality of their data as average or below average. Before AI can work, someone has to organize and prepare that data.
When and How AI Will Succeed in Finance
There is, however, a specific reason for optimism. The METR research institute has long tracked how AI is improving at completing tasks independently. Its finding: the length of tasks AI can handle autonomously doubles approximately every seven months. The best models today can complete tasks that would take a human expert several hours. In a year or two, they may be able to handle a full day's work.
This has a direct impact on finance departments. The tasks CFOs most want to automate, such as account reconciliation, invoice processing, variance analysis, or journal entries, are precisely the kinds of structured but complex processes that benefit most from these improvements. Moreover, 57% of CFOs admit that they feel moderate to strong pressure from their boards and investors to do something with AI. And 72% expect their technology budgets to increase over the next two to three years. The pressure is real, and the money will follow.
Once companies overcome these two obstacles, where will finance departments actually deploy AI? The survey shows clear priorities:
- Accounts payable and procurement (52% of respondents) - invoice processing and supplier management
- Financial planning and forecasting (40%) - automation of FP&A processes
- Accounts receivable (35%) - payment tracking and cash flow management
- Financial close and consolidation (27%) - monthly and annual closes
As for how they plan to acquire it, 95% of CFOs plan to buy an off-the-shelf solution rather than develop their own. And 67% believe they need specialized financial AI tools for production use, rather than direct access to general-purpose language models.
65% of CFOs expect to begin using AI or expand its use within the next one to two years. And 92% are willing to redirect part of their payroll spending to AI tools.
