What counts as a return on AI in marketing?
A return is a benefit attributable to the use of AI after allowing for its costs. The benefit might be lower expenditure, greater capacity or stronger business results.
These outcomes are not interchangeable. Time released creates capacity; it becomes a cash saving only when expenditure actually falls. Extra output has value when it contributes to an objective, rather than simply increasing the volume of material produced.
A financial ROI calculation divides net attributable benefit by investment cost. The difficult part is establishing the benefit and the counterfactual: what would have happened without the change.
How is the time saved by AI measured?
Time savings are measured by comparing equivalent completed work, including preparation, review and correction.
Timing only generation can omit effort transferred to colleagues or later stages. Quality also affects the comparison: faster work that requires repair is a different result from faster work accepted at the same standard.
METR’s early-2025 experiment with 16 experienced open-source developers found they took longer with the tested AI tools despite believing they were faster. It concerned coding, not marketing. The older models and narrow sample make it a warning about self-reported time savings, rather than a forecast of current AI productivity. METR’s study
Which costs belong in an AI marketing ROI calculation?
The relevant costs include both operating expenditure and the work required to introduce and maintain the system.
They can include subscriptions, usage charges, data preparation, integration, staff training and review. The boundary depends on the decision being evaluated. A comparison of two existing tools has a different scope from a transformation programme.
Costs may vary with task volume, model selection and the amount of correction required. A quoted unit price therefore does not establish the cost of an accepted piece of work. Estimates remain conditional on those assumptions until actual use provides evidence.
How can a business distinguish productivity gains from commercial gains?
Productivity concerns the resources used to produce acceptable work; commercial gains concern outcomes such as profitable incremental sales.
A before-and-after comparison may show a change without proving AI caused it. Price, distribution, media spend and seasonality can affect the same result. Controlled comparisons or other suitable evaluation methods can strengthen attribution, with the method chosen for the decision and available data.
Reporting the outcome, measurement method and uncertainty separately makes the claim interpretable. A verified reduction in production time is still useful evidence even when its eventual effect on revenue remains unknown.