AI Adoption Is Rising, but Financial Returns Are Lagging

AI can help an employee finish a task faster without changing the company’s costs, revenue or earnings. That distinction helps explain the gap in McKinsey’s 2026 survey: 80% of respondents said AI improved their personal productivity, while 37% attributed at least some EBIT impact to AI. About 6% met McKinsey’s definition of high performers: respondents who attributed at least 5% of EBIT to AI and described its impact as significant. These are reported attributions, not evidence that any one practice causes financial returns. 68

Why productivity gains may not show up in earnings

A faster individual task is only one link in a business process. If approvals, handoffs, systems and decision rights remain unchanged, the time saved may not reduce costs or increase revenue. McKinsey’s 2026 survey associates stronger AI performance with workflow redesign, while its earlier research found workflow redesign had the strongest relationship with EBIT impact among the factors it examined. That relationship is suggestive, not proof of causation. 514

The distinction matters for measurement, too. Tool access, usage and time saved can show that AI is being used; on their own, they do not show whether the business is earning more or spending less. The reported gap between personal productivity and EBIT impact makes it important to track both. 68

A practical path from AI pilots to business value

Start with a specific business outcome

Choose a process where an accountable owner can define the problem and the result to test. That might mean reducing the cost of a service process or improving conversion—but the target should fit the business, not simply count AI activity. Establish a baseline before changing the process so results can be compared against it.

Keep the initial scope focused enough to evaluate. A defined process and outcome make it easier to see where AI helps, where people still need to intervene, and whether the change is worth expanding.

Make data quality and ownership part of the work

A DMEXCO 2026 commerce session described supplier product data arriving in different formats, with inconsistent labels and missing information. Without an agreed standard, those problems can carry through to product feeds and require manual correction. 25

Set clear responsibility for the data, the business result and decisions about reviewing AI output. In commerce, consistent and complete product information—including accurate prices where relevant—also matters as discovery spans AI assistants, social media and marketplaces. 273334

Redesign the workflow, not just the task

Look at the full sequence of work: what enters the process, what AI handles, where approval is needed, and who acts on the result. Remove or change unnecessary handoffs where appropriate, but keep people accountable for judgment, exceptions and decisions that carry business risk.

This is a redesign of how work gets done, not just a new tool added to an existing process. McKinsey’s research links workflow redesign with stronger reported financial impact; its 2026 survey does not establish that redesign alone produces those returns. 514

Measure the result in business terms

Track whether the change affects revenue, costs, margin or EBIT—not only adoption, output volume or time saved. Include the costs of implementation and human review in the assessment, and compare results with the baseline. This is a practical way to test whether a productivity improvement reaches the company’s financial results.

The takeaway

The survey does not offer a guaranteed formula for AI returns. It does show why adoption and productivity are incomplete measures of success: respondents reported far more improvement in personal productivity than positive EBIT impact. Businesses can respond by pairing focused use cases with reliable data, clear ownership, redesigned workflows and financial measurement—while keeping people responsible for leadership and judgment. 6814