The Business Case Most Companies Used Is Already Wrong
In July 2026, SAP and Oxford Economics published the results of a survey covering 2,600 director and C-suite executives across 13 countries and a straightforward range of company sizes. The finding that should be on every CIO's desk right now is this: enterprises report that AI is helping them surface business insights and improve customer interactions. It is not delivering the cost savings or time efficiencies that most teams used to justify the investment in the first place.
That is not a technology failure. The models are working. The capabilities are real. The problem is that the business case built in 2023 and 2024 to get AI approved, which was almost universally framed around operational efficiency, headcount reduction, and time savings, is describing a different outcome than the one showing up in actual results. Companies went to their CFOs with one promise and came back with a different kind of value. In 2026, that gap has become a budget defense problem.
The Numbers Behind the Mismatch
The SAP survey is not an outlier. It is the most recent signal in a dataset that has been building for most of 2026.
Fewer than one in three corporate decision-makers in a separate Gartner survey could identify specific financial outcomes attributable to their AI investments. Seventy-three percent of enterprise AI deployments fail to achieve projected ROI, according to McKinsey. A Harris Poll commissioned by Dataiku found that 98 percent of tech leaders are facing increasing board pressure to demonstrate ROI, while 71 percent of CIOs believe their AI budget will face cuts or a freeze if targets are not met. Sixty-one percent of senior business leaders report feeling more pressure to prove AI returns now than they did a year ago.
Bain's survey of enterprise AI deployments produced a six-word summary that captures the situation more precisely than most research: "The technology worked. The value didn't arrive."
That sentence is the entire problem. The technology is doing what it was designed to do. It is the business case surrounding it that has not kept up with what is actually being delivered.
Why Saved Hours Do Not Show Up in the P&L
The dominant ROI metric for enterprise AI in 2023 and 2024 was productivity: hours saved per employee per week. Procurement teams used it to build the original business cases, and finance departments largely accepted it. The logic seemed sound. If AI makes each employee meaningfully faster, the aggregate efficiency gain translates to cost reduction or output growth.
It mostly has not worked out that way. Workday's 2026 research on AI value measurement found that a significant portion of the hours employees nominally save with AI tools gets consumed correcting AI errors, rewriting low-quality outputs, and verifying results before they can be used. Speed at the task level does not necessarily convert to value at the business level, and boards have learned to ask about the conversion, not just the speed.
The more fundamental issue is that productivity gains are not a P&L line item unless they convert into something that is. Faster work translates to measurable financial value only when it results in fewer employees doing the same work, more output being produced with the same headcount, or revenue being generated that would not have existed otherwise. In most enterprise AI deployments, none of those three conversions is happening cleanly or in a measurable way. The time savings are real, diffuse, and invisible to the CFO.
What is showing up instead, as the SAP survey confirms, is a different and genuinely valuable kind of output: better data, cleaner insights, and improved customer engagement. Those outcomes matter. But they were not what was in the business case, and they require a different measurement framework to defend at the board level.
The CFO Is Now In the Room
What changed between 2024 and 2026 is not the technology. It is the accountability structure around it. Through 2023 and into 2024, AI investments were largely approved on competitive grounds. The argument was that falling behind early adopters would create a productivity gap that would compound over time. Finance departments accepted that framing and treated AI spend more like an insurance premium than a standard capital allocation.
That dynamic has reversed. CFOs are now asking what the investment has actually returned, and the threshold for continuation has risen. Forrester Research found that enterprises are deferring 25 percent of planned 2026 AI spend to 2027 as financial scrutiny increases. Gartner's data shows that achieving enterprise-wide cost optimization is the number one priority for finance chiefs right now. The era when AI spend could be approved on strategic grounds without a rigorous measurement plan is over.
Sixty-five percent of CEOs report misalignment with their CFO on AI's long-term value. That misalignment is almost always a measurement problem, not a disagreement about the technology. One side is measuring capability and adoption. The other side is measuring financial outcomes. They are looking at the same deployment and arriving at different conclusions because they are using different frameworks.
What a Defensible Business Case Looks Like Now
The fix is not better AI tools. It is a recalibrated measurement framework that accounts for what AI is actually delivering rather than what the original business case assumed it would deliver.
That starts with identifying where value is genuinely accruing. If AI is surfacing business insights and improving customer interactions, those outcomes need to be converted into financial terms that finance teams recognize. Better insights that lead to faster commercial decisions have a measurable dollar value if the decision pipeline is tracked. Improved customer interactions that reduce churn have a measurable value if baseline churn rates are established before deployment. The value is real, but it requires infrastructure to connect the AI output to a financial outcome.
The second piece is separating AI outputs from AI outcomes. Most companies measure the former, which is what the AI produces, and present it as evidence of the latter, which is what the business did differently as a result. Boards have gotten sophisticated enough to notice the gap, and a presentation of AI usage statistics is no longer a substitute for a clear articulation of how those outputs changed a financial result.
How KAIDATA Approaches This
The measurement problem is exactly where we start with clients who are heading into CFO reviews or budget renewal cycles. The question we ask first is whether the organization can trace a line between an AI output and a financial outcome, and for most companies, that line either does not exist or has never been formally established.
We build the reporting infrastructure that makes the connection legible. That means identifying what AI is actually changing in the business, defining the financial metric that corresponds to that change, establishing a baseline before AI is attributed credit, and building the monitoring layer that tracks the outcome rather than just the activity. That work is less exciting than deploying a new model, but it is the only thing that survives a CFO review with the budget intact.
The SAP survey is a data point, but it is also a warning. If your AI investment is delivering insights and engagement value while your business case still promises cost savings, the gap between those two things is a budget risk. Closing it requires a measurement framework, not a better pitch deck.
