Finance AI Projects Fail at 90% Rate as CFOs Freeze Hiring

Finance departments are betting big on AI—and losing

Mid-market finance organizations are redirecting resources away from people and toward technology, even as the vast majority of their artificial intelligence experiments fail to produce results. According to Gartner research covering more than 300 finance executives, four out of five CFOs are either freezing or reducing headcount while simultaneously increasing AI spending.

The trade-off isn't paying dividends. Gartner estimates that more than 90% of generative AI proof-of-concept projects in finance departments have failed to generate incremental value, Alok Ajmera, CEO of financial planning software company Prophix, told Global Finance.

Why it matters

The disconnect between AI investment and measurable outcomes reveals a fundamental tension in how finance leaders approach automation. Unlike other business functions that can tolerate probabilistic outputs, finance operations demand absolute accuracy—a requirement that current generative AI tools struggle to meet consistently. This reality is forcing a recalibration of expectations and spending priorities across mid-market companies.

The accuracy problem in finance AI

The core issue stems from a mismatch between how generative AI operates and what finance work requires. "This is not a probabilistic exercise, this is a deterministic exercise," Ajmera explained. "You can't be 99% accurate with your numbers. You have to be 100% accurate."

Finance chiefs remain willing to deploy AI for reporting, commentary, and analytics—tasks where slight imprecision is tolerable. But they resist allowing AI systems to touch journal entries or modify financial data directly, where errors carry regulatory and business consequences.

Skills displacement, not mass unemployment

Despite the hiring freeze, Ajmera rejected predictions of widespread AI-driven job losses. He pointed to software engineering, where agentic coding tools have become the most commercially successful AI application to date, as a counterexample. "We have hired more engineers in 2026 than we did in 2025," he said, despite productivity gains from AI-assisted development.

He expects finance to follow a similar pattern of skill evolution rather than elimination. "I would not be surprised in a couple of years if we start seeing finance operations engineers" whose role centers on managing AI agents within finance teams.

Software consolidation accelerates

Beyond AI, Ajmera described a broader trend of companies unwinding the application sprawl accumulated in recent years. He cited a Midwest manufacturer whose cloud application count ballooned from five or six to 25 or 30 before consolidating roughly nine or 10 tools onto a single platform with Prophix's help.

Looking forward, Ajmera anticipates extended purchasing cycles and intensified scrutiny of technology investments as economic uncertainty persists. "There's a lot of caution in the air," he said.

These details were first reported by Global Finance.