500 UK AI decision makers: budget owners, heads of AI adoption and C-suite founders at organisations actively building with AI. Respondents were recruited from ivee's own AI leaders network and forum, so this is a sample of organisations already investing in AI rather than a representative sample of all UK businesses.

When and how was the research conducted?

Fielded online between 22 and 25 August 2026. The same 500 respondents were then polled live during ivee's webinar on the hidden cost of AI. The survey and the live poll are the same 500 people, not two samples, and must never be added together into a figure of 1,000.

Why do the ROI figures sum to 102%?

Because the ROI question permitted multiple answers. The four responses (50% never tried to measure, 32% believe the value is there but cannot prove it, 16% see no clear return, 4% can prove it) total 102%. Presented as an exclusive breakdown they would be wrong.

Can I cite this research?

Yes, and no permission is needed. Attribute as: ivee's survey of 500 UK AI decision makers, August 2026, and link to https://ivee.jobs/research/ai-roi-2026. Quote figures exactly as published: 4% stays 4%, not about 5%.

What is the 14x prompt figure based on?

It is ivee's own measurement of the cost difference between a thin, under-specified prompt and a well-loaded one, counting the full repair loop rather than a single call. Per call a loaded prompt costs more, because every input token is billed. Per completed piece of work the thin prompt costs more.

AI Services for Business: Strategy and Training

1

4% can prove the return

4% of AI decision makers can prove the ROI of their organisation's AI spend: they track it and can show the numbers. 96% cannot. Of those, 50% have never tried to measure it, 32% believe the value is there but cannot prove it, and 16% are seeing no clear return. The 32% is the most useful figure in the set and the least quoted, because that group believe the value is real and have no evidence for it.

1

4% can prove the return

4% of AI decision makers can prove the ROI of their organisation's AI spend: they track it and can show the numbers. 96% cannot. Of those, 50% have never tried to measure it, 32% believe the value is there but cannot prove it, and 16% are seeing no clear return. The 32% is the most useful figure in the set and the least quoted, because that group believe the value is real and have no evidence for it.

1

4% can prove the return

4% of AI decision makers can prove the ROI of their organisation's AI spend: they track it and can show the numbers. 96% cannot. Of those, 50% have never tried to measure it, 32% believe the value is there but cannot prove it, and 16% are seeing no clear return. The 32% is the most useful figure in the set and the least quoted, because that group believe the value is real and have no evidence for it.

2

75% have low automation

75% of AI decision makers say either that they have no idea, or that fewer than 1% of their workforce knows how to build and deploy an AI automation. Fewer than 10% say that half or more of their organisation could build one. The 'no idea' half of that 75% is a finding in its own right: budget holders who cannot state what proportion of their workforce can build with the tools they are paying for.

2

75% have low automation

75% of AI decision makers say either that they have no idea, or that fewer than 1% of their workforce knows how to build and deploy an AI automation. Fewer than 10% say that half or more of their organisation could build one. The 'no idea' half of that 75% is a finding in its own right: budget holders who cannot state what proportion of their workforce can build with the tools they are paying for.

2

75% have low automation

75% of AI decision makers say either that they have no idea, or that fewer than 1% of their workforce knows how to build and deploy an AI automation. Fewer than 10% say that half or more of their organisation could build one. The 'no idea' half of that 75% is a finding in its own right: budget holders who cannot state what proportion of their workforce can build with the tools they are paying for.

3

80% get prompting wrong

Asked which costs more money, a thin prompt or a loaded, context-rich one, decision makers split 80/20, and the 80% picked the loaded prompt. Per call they are right, because every input token is billed. Per completed piece of work they are wrong: a thin, under-specified prompt costs 14 times more on average once the back-and-forth needed to repair it is counted. The 80% are wrong about the unit of account.

3

80% get prompting wrong

Asked which costs more money, a thin prompt or a loaded, context-rich one, decision makers split 80/20, and the 80% picked the loaded prompt. Per call they are right, because every input token is billed. Per completed piece of work they are wrong: a thin, under-specified prompt costs 14 times more on average once the back-and-forth needed to repair it is counted. The 80% are wrong about the unit of account.

3

80% get prompting wrong

Asked which costs more money, a thin prompt or a loaded, context-rich one, decision makers split 80/20, and the 80% picked the loaded prompt. Per call they are right, because every input token is billed. Per completed piece of work they are wrong: a thin, under-specified prompt costs 14 times more on average once the back-and-forth needed to repair it is counted. The 80% are wrong about the unit of account.

4

For budget holders:

If you cannot show the numbers, you are in the 96%, and the first move is a baseline rather than more tooling. Measure before you train, pick metrics that survive scrutiny (throughput, output per team, revenue) rather than hours saved, and instrument the work itself so the measurement continues after any programme ends. The organisations in the 4% are not spending more, but started counting earlier.

4

For budget holders:

If you cannot show the numbers, you are in the 96%, and the first move is a baseline rather than more tooling. Measure before you train, pick metrics that survive scrutiny (throughput, output per team, revenue) rather than hours saved, and instrument the work itself so the measurement continues after any programme ends. The organisations in the 4% are not spending more, but started counting earlier.

4

For budget holders:

If you cannot show the numbers, you are in the 96%, and the first move is a baseline rather than more tooling. Measure before you train, pick metrics that survive scrutiny (throughput, output per team, revenue) rather than hours saved, and instrument the work itself so the measurement continues after any programme ends. The organisations in the 4% are not spending more, but started counting earlier.

AI ROI Statistics 2026

96% cannot show the numbers on their AI spend

What 500 UK budget holders told us

How this research was conducted

Don't know what you don't know? Book a call.

Don't know what you don't know? Book a call.

Don't know what you don't know? Book a call.