AI Adoption in Finance Doubles in Two Years, but Most Companies Fall Short on Oversight
Three-quarters of global companies now use AI in finance functions, up from 30% in 2024, according to a KPMG survey of 1,013 executives across 20 countries. The rapid adoption has pushed companies beyond simple automation into strategic decision-making, but a critical gap has emerged: most lack the systems to verify that their AI is reliable.
Companies report tangible gains. Seventy-one percent said AI improved decision-making speed, 70% saw better decision quality, and 64% achieved higher forecasting accuracy. Finance teams are moving toward what researchers call "agentic AI"-multiple AI functions working together to handle complex tasks like planning and risk evaluation.
The performance split is sharp. Organizations using agentic AI in finance recorded metrics 32 percentage points higher than non-adopters, with forecasting accuracy and ROI each about 40 percentage points higher.
The Trust Problem
Yet only 23% of companies said their AI investments significantly exceeded expectations. The real issue isn't adoption-it's assurance.
Just 42% of surveyed companies have established systems to verify that their AI-based financial processes are trustworthy and stable. Only 29% systematically track AI failures. This matters because companies with AI assurance capabilities-the ability to audit and validate AI decisions-cut errors by 33%, compared with 6% among unprepared firms. They also showed 42% confidence in scaling AI, versus 14% for companies without these systems.
Stronger governance and controls delivered measurable returns. Organizations capable of submitting AI-related audit evidence performed three to six times better than those without such capabilities. Companies tracking AI-related key performance indicators saw ROI improvement rates 10 percentage points higher than firms that didn't monitor metrics.
One-third of companies are now adding "human-in-the-loop" oversight to improve AI trustworthiness, reflecting growing awareness that automation without verification creates risk.
What's Holding Back Progress
Data quality emerged as the single biggest obstacle. Thirty-six percent of respondents cited poor data quality as both the largest constraint on AI use and the most important opportunity for improvement.
Performance gaps between industries also signal data problems. In banking, 71% of respondents reported improvements in forecasting accuracy, compared with 44% in healthcare-a difference researchers attributed largely to variations in data quality and control systems.
What Finance Leaders Need to Do
The path forward requires three parallel efforts. First, fix data quality and integration across systems. Second, build AI governance frameworks and performance measurement systems from the start, not after deployment. Third, hire people with both technical skills and critical thinking-the ability to question AI outputs, not just implement them.
As regulatory scrutiny increases and external auditors demand more evidence, explainability and controllability of AI decisions will become competitive factors. Finance teams that establish assurance systems now will avoid costly rework later.
For finance professionals looking to build these capabilities, resources on AI for Finance and AI learning paths for CFOs can help close the gap between adoption and execution.
Your membership also unlocks:
