KPMG says many companies are still struggling to understand how much artificial intelligence is really costing them.

The consulting firm reported that just 26% of companies have a complete view of their AI-related expenses, according to a survey cited by The Wall Street Journal. The finding suggests that while businesses are rapidly adopting generative AI and other machine learning tools, many have not yet built the financial controls needed to track the technology across their organizations.

The gap matters because AI spending can be spread across several departments and vendors. Companies may pay for cloud computing, software licenses, outside consultants, internal engineering work and data infrastructure without always bundling those outlays into a single budget line. That can make AI look cheaper than it is, especially when projects are still in pilot stages and costs are dispersed among business units.

The survey results also point to a broader management challenge. As executives push to deploy AI more widely, finance teams and technology leaders are being asked to measure returns on investments that are often still experimental. Without clear visibility into spending, it becomes harder to compare projects, prioritize use cases or decide which systems should be scaled up.

KPMG’s findings come at a time when companies are under pressure to justify AI investments. Many firms have announced plans to use the technology to improve productivity, customer service and internal operations, but the economics remain uncertain. Some organizations are finding that implementation costs, data preparation and ongoing oversight are more expensive than expected.

The survey also highlights a tension between rapid adoption and corporate governance. Businesses are eager to avoid falling behind competitors, but they may not yet have the processes in place to monitor AI risks, usage and costs in a consistent way. That can leave executives with only a partial picture of what AI is doing for the business, and what it is consuming in return.

For companies trying to move beyond experimentation, the lack of a full cost view could complicate budgeting and long-term planning. It may also make it harder for boards and shareholders to evaluate whether AI initiatives are delivering value.

The KPMG findings add to a growing body of evidence that corporate AI adoption is advancing faster than the financial and operational frameworks around it. As firms expand their use of the technology, the need for better accounting and oversight is likely to become more urgent.