AI is widely used, and many organizations report savings or revenue gains in particular workflows. But those reports do not establish a universal return on investment: enterprise-level financial impact is less common, and the available figures are largely survey respondents’ reported outcomes—not audited or causal ROI. The clearest current picture is function-specific, with reported benefits most often modest and measured differently across studies.
What current evidence says about AI adoption and financial returns
Adoption is not the same as return. A company can use AI in one or more functions without scaling it, saving money, increasing revenue, or improving its overall financial results.
In Stanford HAI’s 2025 AI Index, 78% of survey respondents said their organization used AI in 2024, up from 55% in 2023. The same report said 71% reported generative AI use in at least one business function in 2024, up from 33% in 2023. These are survey findings about organizational use, not a census or evidence that AI caused a financial gain.
McKinsey’s 2025 global survey found that 88% of respondents reported regular AI use in at least one business function, while approximately one-third said their companies had begun scaling AI programs. The survey also found that 39% of respondents attributed any enterprise-level EBIT impact to AI; most of that group said less than 5% of their organization’s EBIT was attributable to AI use. These figures indicate that reported use is more widespread than reported enterprise-level impact. They do not show what return an organization can expect.
#1 Best Overall
Where organizations report savings and revenue gains
Stanford HAI’s 2025 AI Index reports the share of respondents whose organizations use AI in the relevant function and who reported a financial benefit in that function. The percentages below are the share reporting a gain—not the size of the average gain, an ROI percentage, or a return per dollar invested. The report’s 2024 survey findings indicate that most reported cost savings were below 10%, while the most common reported revenue increase was below 5%.
| Business function | Respondents reporting AI-related cost savings | Respondents reporting AI-related revenue gains |
|---|---|---|
| Service operations | 49% | 57% |
| Supply chain management | 43% | 63% |
| Software engineering | 41% | Not stated for this function in the cited Stanford HAI findings. |
| Marketing and sales | Not stated for this function in the cited Stanford HAI findings. | 71% |
These rates should not be read as a ranking of functions by profitability. The survey reports how often respondents said they saw a type of benefit, not the monetary value of the benefit, the implementation cost, the time required to achieve it, or whether it persisted. A frequent small gain could be less valuable than a less frequent but larger one.
Why survey-reported benefits are not the same as ROI
ROI requires comparing benefits with the full costs of obtaining them. A reported reduction in processing time, for example, is not automatically a cash saving: it may create capacity for other work without lowering payroll or operating expense. Similarly, reported revenue growth does not establish that AI caused the increase or that the gain exceeded the cost of the system and its implementation.
The sources here use different populations, geographies, questions, and definitions of AI and financial impact. Stanford HAI’s function-level figures, McKinsey’s global enterprise EBIT finding, and the separate McKinsey US C-suite results below are useful for understanding reported patterns, but they are not interchangeable measurements. None establishes a directly comparable cross-industry ROI ranking or a universal return figure.
- Reported benefit: A survey respondent says their organization experienced a cost or revenue change. This is useful evidence of perceived outcomes, but it does not independently verify the amount or cause.
- Enterprise financial impact: A respondent attributes some portion of company-wide financial performance, such as EBIT, to AI. This is broader than a single-use-case gain, but remains an attribution reported in a survey.
- Audited or causal return: A verified financial result tied to a defined AI investment, with a credible comparison against what would have happened without it. The cited survey figures do not provide this kind of general cross-industry estimate.
- Expected benefit: A forecast of future revenue, productivity, or employment effects. It should be kept separate from results already achieved.
What a separate US C-suite survey reports about generative AI
McKinsey’s “AI in the workplace: A report for 2025” draws on a US C-suite survey conducted in October–November 2024. In that survey, 19% of respondents reported generative AI revenue growth above 5%, 39% reported growth of 1–5%, and 36% reported no revenue change. Separately, 23% reported any favorable change in costs. These are respondents’ reported perceptions of generative AI’s effects, not audited returns, and the US survey should not be combined with McKinsey’s global survey as if they used the same respondent base.
The same US C-suite survey found that 87% expected revenue growth from generative AI over the following three years. That figure is an expectation, not an achieved result. Forecasts may inform planning, but they do not demonstrate that investments have paid off.
What is associated with stronger reported performance
In McKinsey’s 2025 global survey, high performers were associated with workflow redesign, faster scaling, transformation practices, and objectives that included growth or innovation alongside efficiency. These are associations in survey findings, not proof that any one practice caused better financial results or a guaranteed recipe for ROI.
The practical implication is to evaluate an AI investment as a change to a workflow and operating model, not merely as software adoption. Define the business outcome first, identify the process and people affected, and track whether the measurable result survives the costs and operational changes required to produce it.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →How to measure whether an AI investment is paying off
A useful internal assessment starts with a specific use case and a baseline. The calculation should use the same scope and time period for the benefits and costs; otherwise a small pilot’s benefit may be compared with costs that belong to a different scale of deployment.
Rank #4
- Choose one business outcome. Specify whether the target is lower cost, faster cycle time, higher conversion, increased capacity, fewer errors, or another measurable result. Avoid treating “AI use” itself as the outcome.
- Record the baseline before deployment. Measure the existing workflow over a representative period, including volume and quality. If possible, compare with a similar team or process that has not adopted the system, so broader business changes are less likely to be mistaken for an AI effect.
- Count the full investment. Include implementation and integration work, software or model access, data preparation, security and governance, employee training, human review, ongoing maintenance, and the cost of errors or rework. Separate one-time costs from recurring ones.
- Translate operational change into financial value carefully. Time saved is not cash saved unless it changes spending, staffing, or productive capacity in a way the organization values and can substantiate. Revenue gains should be assessed against a baseline and considered alongside margin, not just sales volume.
- Use a defined calculation and period. A common simple measure is (verified financial benefit − total AI investment cost) ÷ total AI investment cost. State whether the result is a pilot-period figure, an annualized estimate, or a recurring outcome; do not present an estimate as realized cash return.
- Check quality and durability. Track errors, customer outcomes, compliance, employee workload, and whether the gains persist after the initial rollout. A short-term speed improvement can be offset by correction work or other costs.
- Scale only when evidence travels. Re-measure at the next deployment level. A result in one team or workflow does not necessarily transfer to another function, population, or business unit.
This approach does not make a business result causal by itself. It makes the assumptions, comparison, costs, and time horizon visible, so decision-makers can distinguish a plausible contribution from a proven return.
How to interpret newer executive forecasts
A February 2026 NBER working paper, revised in March 2026, is summarized as drawing on nearly 6,000 senior business executives at firms in the United States, United Kingdom, Germany, and Australia. Its reported three-year expectations were average productivity growth of 1.4%, output growth of 0.8%, and employment reduction of 0.7%. These are executive forecasts, not observed effects or realized investment returns, and they should not be compared directly with function-level savings or revenue reports.
What the evidence can—and cannot—tell a business
The evidence supports a measured conclusion: organizations commonly report AI use, and respondents report financial benefits in particular functions, especially revenue gains in marketing and sales and savings in service operations. At the same time, reported enterprise-level EBIT impact is less common, and most respondents who attributed any EBIT impact to AI assigned it less than 5% of EBIT.
Recommended Free Tools
These findings can help identify workflows to investigate, but they cannot answer whether a particular company’s investment has paid off. That requires its own baseline, full cost accounting, outcome measures, and a credible comparison over a stated period. Survey adoption rates, benefit-reporting rates, and executive expectations are not substitutes for that calculation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
