What AI Adoption Numbers Do Not Tell You

You will have seen the headline. A large majority of organisations now report using AI. Stanford’s AI Index put the figure at 78% of organisations in 2024, up from 55% the year before.

That number is accurate and it is one of the most misread statistics in business at the moment, because of what the question actually asked.

What the question measures

These surveys ask whether an organisation uses AI in at least one business function. That is a low bar and deliberately so, because the researchers are tracking diffusion rather than value.

A company where the marketing team drafts with a chat tool answers yes. So does a company that rebuilt its entire lead handling. Both appear as one adopter, and the gap between them is the whole story.

So the figure tells you the technology is present nearly everywhere. It says nothing about whether anything changed as a result, and reading it as evidence of value is how businesses end up feeling behind for no reason.

Why the gap between adoption and result is so wide

Three patterns account for most of it, and none are technical.

The use case was chosen to demonstrate rather than to run. Impressive cases tend to be rare events. Something happening twice a quarter builds no habit, so within a year it is quietly abandoned.

Nothing was connected. A tool producing good output that somebody then moves by hand has relocated work rather than removed it.

Nobody owned it afterwards. Processes change. Without somebody noticing when a system has drifted out of alignment, it degrades silently.

The number that would be useful

If you want to know where a business genuinely stands, the survey question to ask is not whether they use AI.

It is: name one process that runs differently than it did a year ago, and say what number it moved.

Most companies answering yes to the adoption question cannot answer that one. Not because they are pretending, but because nobody established a baseline, so there is nothing to compare against. That is also why so much investment gets described afterwards as producing no measurable return: the measurement was never set up.

What the same research does say clearly

The Stanford index is more useful on capability than on adoption. It found that AI both speeds up task completion and improves the quality of output, with the largest gains among people who were previously less skilled at the task.

That last clause is the one worth acting on. The effect is not primarily making your best person faster. It raises the floor, which is precisely the constraint in a business where one experienced individual is the bottleneck on everything and where hiring their equal takes a year and a half.

Read that way, the research points somewhere specific: the highest return sits where capability is currently limited by one person’s availability, not where an already efficient process could run slightly faster.

Being behind is not the risk

The adoption figure creates a sense of urgency that pushes businesses toward the wrong decision, which is to adopt something quickly so as not to be in the minority.

The businesses producing results are rarely the earliest. They are the ones that picked a process running every day, wrote down how it actually works, agreed a number before building, connected it to the systems they already use, and put a name against it.

That sequence takes a few weeks and it is available at any point. Arriving at it two years after everybody else costs almost nothing. Adopting quickly without it costs the budget and the internal credibility for a second attempt, which is considerably more expensive.

How to read any statistic in this field

• Ask what population was surveyed. Figures from large enterprises describe a different world from one where the owner does the quoting.

• Ask what the question was. Using, piloting, planning and seeing measurable value are four different questions with wildly different answers.

• Ask who paid for it. A vendor survey and an academic index are both useful and they are not the same kind of evidence.

• Ask when it was collected. Data gathered a year before publication describes a year ago.

Applied honestly, that filter removes most of what circulates and leaves a small number of figures worth acting on.

Your market is a different population

One more reason the headline figure misleads a business of five to fifty million: it is not describing you.

Surveys reporting very high adoption typically sample larger organisations, where a single team experimenting is enough for the whole company to answer yes. Firm level data covering the broad population of businesses, including the small ones, produces considerably lower numbers.

Both are accurate. They measure different populations, and confusing them is how an owner concludes they are the last business in their sector to move when in fact most of their direct competitors have done nothing at all.

If you want a realistic sense of where your market actually is, look at what your competitors visibly do rather than at a global percentage. Response times, whether enquiries get answered in the evening, how long a quote takes. Those tell you where the bar is in your sector, which is the only bar you are being compared against.

The number worth tracking internally

Rather than benchmarking against a survey, track one figure of your own.

How many processes in your business run without a person moving information between systems. It starts at zero in most companies. Every one you add is measurable, and after four or five the effect on capacity is obvious without any statistic.

That number is more useful than any published figure, because it describes your business rather than an average of thousands of others that share nothing with it.

Sources

• Stanford Institute for Human-Centered Artificial Intelligence, Artificial Intelligence Index Report 2025, reporting that 78% of organisations used AI in 2024, up from 55% the year before.

• Stanford Institute for Human-Centered Artificial Intelligence, Artificial Intelligence Index Report 2024, 15 April 2024.

AI Optimize starts from a process that runs every day and agrees the number before the build, so you can answer the question the surveys do not ask. That work sits under Workflow Automation.