AI ROI in 2026: What McKinsey's Survey Shows

The gap is the main AI ROI story in 2026. Half said AI helps them make better decisions, but those individual gains have not yet become broad financial returns for most organizations.

AI is producing useful work faster than it is producing company-wide profit. The positive EBIT share was essentially unchanged from 2025, despite wider use and more scaled deployments.

“These individual gains have yet to translate into broad financial impact for organizations.”McKinsey, The state of AI in 2026: On the road to ROI

The quote captures the tension. Productivity can rise without a matching gain in company earnings. A faster task is only one link in a longer workflow, and the value can disappear if the rest of that workflow stays the same.

What counts as a high-performing AI organization?

McKinsey defines an AI high performer in two parts. The organization attributes at least 5% of EBIT to AI use and describes the value from AI as significant.

The small group that met that standard was unchanged from 2025.

Source: McKinsey, *The state of AI in 2026: On the road to ROI, based on 1,719 survey participants.*

The most important movement is not the share of winners. It is what the winners changed. Nearly three-quarters of high performers said they fundamentally redesigned workflows because of AI, up from 55% last year. Only one-quarter of other respondents said the same.

Where the gains show up

Financial gains follow two different branches. Cost reductions appear most often in supply chain management, service operations and manufacturing. Revenue gains appear most often in marketing and sales, product and service development, and software engineering.

Source: McKinsey, *The state of AI in 2026: On the road to ROI.*

This split matters for measurement. A support team should not be judged by the same revenue test as a sales team. Likewise, an AI sales tool should not earn credit for time saved if the commercial goal was more revenue.

Our example (illustrative): One support team saves 80 hours a month. Each hour costs 50 units, and the monthly tool and usage cost is 1,000 units. The same team can produce 2 very different financial outcomes.

Our method is: ROI = (gain minus cost) / cost.

Outcome A: time is saved, but EBIT does not move.

• Time equivalent: 80 hours × 50 units = 4,000 units (our calculation).
• The hours are not redeployed and no cost is removed, so the EBIT gain is 0 units.
• ROI: (0 minus 1,000) / 1,000 = negative 100% (our calculation).

Outcome B: the workflow changes and a cost reaches EBIT.

• The team redesigns the workflow and removes a 3,000-unit contractor cost.
• The gain reaching EBIT is 3,000 units.
• ROI: (3,000 minus 1,000) / 1,000 = 200% (our calculation).

Our reading: Time saved counts as a return only when it removes a cost or produces revenue. Otherwise, it is faster work, not EBIT.

This is the gap behind McKinsey's statement that individual gains have yet to translate into broad financial impact.

What the 6% do differently

High performers redesign work instead of placing an AI tool on top of the old process. Their lead over other organizations is visible in workflow design, leadership and measurement.

High performers were also twice as likely to say senior leaders showed real commitment to AI initiatives. They were twice as likely to have defined processes for measuring impact.

The pattern suggests a practical sequence. Choose one business outcome, redesign the workflow around it, name the accountable leader and define the measure before scaling. Tool adoption alone does not complete that sequence.

Costs can limit use

Per-token costs have fallen, but token consumption and generation have risen faster.

About 20% of respondents said AI-related operating costs, including token costs, had constrained AI use. Yet 60% expected their organizations to increase AI investment over the next year.

Our reading: The budget signal is expansion with tighter control, not retreat. Leaders are likely to ask which workflows deserve more capacity and which pilots should stop. The right denominator is the business result, not the number of prompts or tokens.

How marketing teams should read the numbers

Our reading: The survey does not prove that marketing AI produces company-wide ROI. It does show that marketing and sales is the function where respondents most often see revenue gains from AI.

A small marketing team can use that finding as a measurement guide. Track the chain from the buyer question to the answer, the cited page and the commercial action. Do not treat a faster draft as the final return.

That is also why AI visibility and business outcomes need separate measures.

For a practical workflow, see how to optimize content for AI search from question selection through rechecking the same cohort.

What the survey can and cannot tell you

The online survey ran from May 4 to June 8, 2026. It included 1,719 participants in 97 nations. The data were weighted by each respondent's nation contribution to global GDP.

The results describe what respondents said about their organizations. They do not calculate a universal dollar return for every AI project. Function-level answers also use different respondent bases because each question was asked of people whose organizations used AI in that function.

The strongest conclusion is narrower and more useful. Individual gains are widespread, company-wide EBIT impact is not, and the small group seeing significant financial value changes workflows and measurement together.

Xtrusio's free check shows when ChatGPT and Gemini mention you, which competitors appear instead and what to work on next.

Get your free AI visibility check

See where AI answers mention your brand and which competitors appear instead.

Sources

1. McKinsey & Company, The state of AI in 2026: On the road to ROI, August 2026.