The ROI of AI 2025
B · Useful, limited Google Cloud and National Research Group survey of 3,466 senior leaders with sample sizes and question wording disclosed; a vendor survey whose ROI figures are executives' own expectations.
Summary based on the full text·Original: about 30 minutes to read
In short
The second annual Google Cloud report on generative AI returns, based on a survey of 3,466 senior leaders. Its theme is agents: half of companies using generative AI say they already run agents in production, and "agentic AI early adopters" (those putting at least 50% of future AI budget into agents) report returns more often. The report lists five areas where respondents say returns appear, led by productivity, customer experience and marketing, and closes with recommendations and customer quotes.
Key numbers
- 3,466senior leaders surveyed at firms with $10M+ revenue, April to June 2025
- 52%executives using gen AI whose organisations also have AI agents in production
- 88%executives at agentic AI early adopters reporting ROI on at least one gen AI use case
- 74%all executives reporting ROI on at least one gen AI use case
- 39%executives whose organisation has launched more than 10 AI agents
As reported by the source.
English·PDF·opens services.google.com
What it reports
- 52% of executives at gen AI users say their organisation uses AI agents; 39% say it has launched more than 10.
- 74% report ROI on at least one gen AI use case; 88% among agentic early adopters (n=460).
- ROI by area: productivity (39%), customer experience (37%), sales and marketing (33%).
- 78% of executives with C-level sponsorship report ROI on at least one use case (n=1,269).
- Mean of 26% of total IT spend already allocated to AI; 48% are reallocating non-AI budget to fund it.
- Data privacy and security is the top factor when choosing an LLM provider.
How they got it
A 16-minute online survey run by Google Cloud with National Research Group, fieldwork 18 April to 3 June 2025, of upper-level and C-suite leaders at firms with over 100 staff and $10M revenue across about 20 markets (US 1,047). Footnotes give the n and question wording for each figure. Statistics include only firms using gen AI in production.
Limits
- The ROI figures come from the question "In what timeframe do you expect gen AI to deliver ROI", so "seeing ROI" is a self-reported expectation, with no financial measure.
- The sample is limited to firms already using gen AI in production, which inflates adoption and return rates.
- Early adopters are defined by budget share for agents, so their higher ROI may reflect enthusiasm as much as results.
- Published by a vendor that sells agent platforms; case quotes come from Google customers.
Why it matters for companies
Use it as a peer benchmark for spending and adoption, not as proof of return. The figure most worth checking against your own position is the 26% of IT spend already on AI, along with the shift to funding AI by cutting other budgets. Any internal business case should replace "reported ROI" with measured before-and-after figures, which this report does not provide.
Summary based on the full text. Read the original for the full argument and data.Open PDF on services.google.com