AI ROI: why did two years of AI tools not move your revenue?
Home › Articles › Hire AI › AI ROI
Updated 2026-11-01
AI ROI stayed flat because AI tools were pointed at tasks inside departments, while revenue leaks in the handoffs between them. 56% of 4,454 CEOs report no significant financial benefit from AI (PwC, 19 January 2026). The Thirteen Seams by Jon Weberg maps those handoffs, which shows how hiring an AI employee differs from buying an AI tool.
- 56% of 4,454 CEOs report no significant financial benefit from AI to date; 12% report both cost and revenue benefits (PwC, 19 January 2026).
- 37% of 1,719 respondents attribute at least some EBIT impact to AI use (McKinsey, The state of AI in 2026, 25 August 2026).
- Executives report a 1.8% labour productivity gain from AI in 2025; the gain implied by their revenue and employment is 0.6% (NBER working paper 34984, March 2026).
- Nearly three quarters of AI high performers redesigned workflows around AI, against about one quarter of other companies (McKinsey, 25 August 2026).
- More than half of generative AI budgets go to sales and marketing tools, while the largest returns were found in back office work (MIT NANDA, reported by Fortune, 18 August 2025).
You bought the tools, your team opens them every day, and this December the revenue line still moved on its own schedule, didn't it, so before you renew a single seat, what did your AI subscriptions cost over the last 24 months, and which line of revenue moved because of them?
Why did AI tools not move your revenue?
AI tools sped up tasks inside departments; revenue leaks in the handoffs between them. That is why 56% of CEOs in PwC's January 2026 survey report no significant financial benefit to date.
The writer drafts faster, the setter summarises calls faster, the closer preps faster, and every one of those gains stays inside a team whose number was already counted, while the lead nobody called back and the booked call nobody confirmed sit between two teams, where a faster task never reaches, because a faster task is not a closed handoff, is it?
How many companies get a return from AI?
A minority of companies report a measurable financial return from AI. The three largest recent surveys agree on the direction.
| Source | Date | Sample | Finding |
|---|---|---|---|
| PwC Global CEO Survey | 19 January 2026 | 4,454 CEOs, 95 countries | 56% no significant financial benefit; 33% cost or revenue; 12% both |
| McKinsey, The state of AI in 2026 | 25 August 2026 | 1,719 respondents | 37% attribute some EBIT impact to AI; about 6% are high performers |
| NBER working paper 34984 | March 2026 | Nearly 750 executives, mostly CFOs | 1.8% productivity gain reported for 2025; 0.6% implied |
The numbers are flat, not negative, which tells you the tools did their tasks and the money sat somewhere they were never aimed, so you are not behind, you are in the majority, and the question becomes what the 12% did that you can copy.
Why do most AI projects fail to pay back?
Most fail on fit to the work, not on the model. MIT NANDA found a 95% failure rate for enterprise AI solutions, with vendor purchases succeeding about 67% of the time (Fortune, 18 August 2025). Gartner (25 June 2025) predicts over 40% of agentic AI projects will be cancelled by the end of 2027, citing rising costs and unclear business value (as reported by New Electronics).
If you shelved an agent pilot this year, Gartner's forecast already counts it, and "unclear business value" is the polite name for a project that never had a number it answered for, so which of your pilots was given a handoff and a measure before it was switched on?
What do the companies that got a return do differently?
They changed how work moves between teams, not which tool each team uses. Nearly three quarters of AI high performers fundamentally redesigned workflows, against about one quarter of other companies (McKinsey, 25 August 2026). CEOs with strong AI foundations are 3 times more likely to report meaningful financial returns, and those with both cost and revenue gains 2 to 3 times more likely to have embedded AI extensively (PwC, 19 January 2026).
Your team did not use AI wrong; nobody redrew the route the customer takes between them, so if the winners redesigned how work crosses from one team to the next, which one join in yours would you redesign first?
Where does revenue leak that AI tools never touch?
Revenue leaks in a revenue seam: a handoff between two teams that nobody owns. The two that pay most often are the list nobody mails and the booked call nobody confirms.
Jon Weberg found John Crestani's return in 43,297 contacts nobody was mailing, which paid $10,000 to $30,000 a week for fifteen straight weeks, and found AI Acquisition's on a 22 closer floor whose show rate moved from 19% to 43% in about a month, $1,322,312 collected that month, inside a $70,000,000 a year business still running on the CRM he built, and neither needed a new tool, only an owner on the join, which is the line drawn in the six differences between an AI tool and an AI employee; how many contacts in your CRM have gone ninety days unmailed, and out of every 100 calls your team booked in November, how many happened?
How do you audit your own AI subscriptions in 15 minutes?
You audit your AI subscriptions in 5 steps and about 15 minutes, with one sheet.
- List every AI subscription and its monthly cost.
- Write the task each one does.
- Mark whether it owns a handoff between two teams, yes or no.
- Mark whether it has a written job and a number it is measured on, yes or no.
- Total the monthly cost of every subscription with two noes.
| Subscription | Monthly cost | Task it does | Handoff it owns (yes or no) | Written job and measure (yes or no) | Keep or cancel |
|---|---|---|---|---|---|
You would rather fix the aim than buy another tool, wouldn't you, and the total at the bottom of that sheet is the budget already in your hands for the one account worth keeping.
What changes when AI is given a job on a handoff instead of a task?
It gets an owner, a measure and a Friday read. A tool waits to be opened; an AI employee runs one written handoff, logs each action and passes anything below 70% confidence to a named person.
Picture January: you cancel the subscriptions that never owned a handoff, keep the account you already pay for, start with the first job to give AI, then point the second at the handoff that cost you most this year, choosing which AI employee goes first by dollars, so next December's total has a line with a number beside it. Your next step takes three lines:
- Run the fifteen minute audit above.
- Give one AI account one written job on your costliest handoff.
- To have every handoff ranked in dollars and owned, Book My AI Seam Architect Install.
What else do people ask about AI ROI?
What is the AI productivity paradox?
It is the gap between heavy AI adoption and small measured gains. Executives in an NBER survey (March 2026) reported a 1.8% productivity gain from AI in 2025, while their own revenue and employment implied 0.6%.
Are CFOs right that AI is paying off?
Some are. McKinsey (25 August 2026) found 37% of respondents attribute at least some EBIT impact to AI. The rest report time saved that has not reached the P&L.
Is the problem the model or my prompts?
Usually neither. AI misses what your business never wrote down and works on tasks no one tied to a handoff. Context and a job fix both.
Should I cancel my AI subscriptions?
Cancel the ones that own no handoff and have no measure. Keep one strong account and give it a written job.
How long before AI shows a return on a handoff?
The first job, Email Triage, is judged by the next morning. A revenue handoff shows its rate on the Friday log within the first two weeks.
Jon Weberg is the founder of Profitalize and the originator of The Thirteen Seams, a revenue diagnostic that finds the money a company loses in the handoffs between its teams. About Jon Weberg
