AI adoption has hit 78%, and it is being used as a reason to invest. It is the wrong reason. Adoption measures activity, not return, and Stanford's own AI Index admits the payoff is still unproven. Lead on measured value in your own numbers, not on the crowd.
- AI adoption is now near-universal: Stanford's 2025 AI Index found 78% of organisations used AI in 2024, up from 55% the year before. That number is being used as a reason to invest.
- It is the wrong reason. Adoption measures activity, not return, and the same report is careful to note that the economic payoff of AI is still poorly understood and hotly debated.
- Investment and access are surging (corporate AI investment hit 252.3 billion dollars in 2024; the cost of using a model fell around 280-fold in eighteen months), which makes AI cheaper to buy but does not make it more proven.
- The 78% is social proof, not evidence. It tells you the crowd has moved; it does not tell you the crowd is better off, or that you will be.
- Lead the investment on measured value in your own numbers, run pilots as cheap experiments with a kill switch, and refuse the adoption headline as a target.
Your board asks why you are behind on AI, and the evidence they cite is a number: nearly four in five companies are using it now, so why aren't you doing more? You feel the pull to spend, if only to avoid being in the shrinking minority that has not. It is a reasonable-sounding pressure, and it points you at exactly the wrong target. Adoption is a headcount, not a scoreboard. The number that should drive your investment is whether AI is measurably improving your operation, and that number is far harder to find, and far rarer to see, than the adoption statistic everyone is quoting.
Start with the figure itself, because it is real and it is striking. Stanford's 2025 AI Index, one of the most trusted independent reads on the field, found that 78% of organisations reported using AI in 2024, up from 55% a year earlier, and that the share using generative AI in at least one business function jumped from 33% to 71%. Corporate AI investment reached 252.3 billion dollars, and the cost of running a model at a given level of capability fell roughly 280-fold in eighteen months. Adoption, spending and cheapness are all climbing together. None of that is in dispute.
So why is "everyone's using it" a poor reason to invest?
Because the same report is unusually honest about what those numbers do not prove. Its authors note that while productivity gains show up in study after study, the firm-level economic return, the actual ROI, remains poorly understood, with little agreement on what it even is. Adoption has raced ahead of proof. So the 78% tells you that the crowd has moved. It does not tell you the crowd is better off, and it certainly does not tell you that a project inside your organisation will pay.
An adoption statistic counts activity, not value. That 78% includes every stalled pilot, every unused licence and every "we tried it and quietly stopped." It is the same trap behind why most organisations fail at AI adoption: presence gets mistaken for performance. And the falling cost, impressive as it is, cuts the other way on proof. The cheaper AI gets, the easier it is to buy, and the less the act of buying it demonstrates. A 280-fold drop in cost is a reason to experiment, not a reason to assume value.
Invest in AI on evidence, not on the adoption headline
The AI Strategy Session helps you decide where AI is worth real money in your operation, and how to prove it, in ninety minutes.
Book your Strategy SessionWhat should drive the investment instead?
Evidence you can see in your own operation, gathered cheaply and deliberately, rather than the industry's adoption rate.
- Measure value in your numbers, not the industry's percentage. Pick a metric that matters to you, cycle time, error rate, cost to serve, and prove AI moves it before you scale. The adoption rate of other companies is not a KPI.
- Run pilots as experiments with a kill switch. Most AI initiatives will not pay. Design to find the few that do quickly and cheaply, and to stop the rest without ceremony. The collapse in cost is what makes this affordable.
- Separate capability from advantage. The frontier models have converged, with barely a percentage point now separating the top two, and they are cheap. Owning the same tool as everyone else is not an edge. What you do with it is, as explored in designing the loop the work runs in.
- Refuse "we should be doing more AI" as an objective. Replace it with "where has AI measurably paid, and how do we do more of that." One is fear of missing out; the other is management.
- Fund proof before scale. Put money into measurement and iteration, not just licences, so that when you do scale, you are scaling something you have actually seen work.
78% of organisations use AI. That is social proof, not evidence. It tells you the crowd moved, not that it is better off. Invest on proof in your own numbers, not on the crowd.
What does this change for me as a leader this quarter?
It changes the question you answer to the board. Not "are we adopting fast enough," which the whole market has already answered many times over, but "where has AI measurably paid for us, and how do we concentrate on that." The adoption race is nearly finished; at 78% there is almost no one left to be ahead of. The value race, by the report's own admission, has barely begun, and that is the one actually worth winning.
This is the discipline underneath the end of business as usual: when a technology becomes universal and cheap, being in the crowd is worth nothing, and knowing precisely what it does for you is worth everything. Let the 78% reassure your competitors. Spend your money on the evidence, keep the discipline of proof before scale, and treat every AI pound as an investment that has to show its return, not a subscription to the consensus. These are the same evidence questions a board should be asking of any large bet.
| Source | Finding on AI adoption versus proven value |
|---|---|
| Stanford HAI, 2025 AI Index | 78% of organisations used AI in 2024, up from 55% in 2023, and generative AI use in at least one function rose from 33% to 71%, adoption, not proven return |
| Stanford HAI, 2025 AI Index | Corporate AI investment reached 252.3 billion dollars in 2024 and the cost of using a model fell around 280-fold in eighteen months, spend and access are surging |
| Stanford HAI, 2025 AI Index (authors) | Productivity gains appear across studies, but firm-level ROI and economic benefit remain poorly understood, with little agreement on what the return actually is |
| McKinsey, State of AI (2025) | Most organisations report only modest bottom-line impact from AI so far, with few attributing significant earnings gains to it, value capture lags adoption |
Frequently asked questions
If 78% of organisations use AI, aren't we at risk by not investing more?
Doesn't the falling cost of AI make investing an easy decision?
How should we measure whether our AI investment is working?
About the author
British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.