Two studies landed within a week of each other this month, and if you put them side by side, they don't agree on a single thing.
The first says 82% of enterprises still haven't seen significant revenue gains from AI. Their return on investment has failed to outpace what they've spent on it for two years running — despite bigger budgets, bigger teams, and bigger ambitions every quarter.
The second says 70% of small businesses report AI already increased their revenue this year, with an average return of 3.7x on what they spent.
Same technology. Same models, mostly. Wildly different outcomes. Somebody's math on "the benefits of AI in business" is wrong — and it's worth figuring out whose, because the answer changes what you should be doing this quarter.
What the numbers actually say
On the enterprise side: a fresh HCLTech study found that share of companies whose AI ROI fails to outpace their investment has been stuck at 57% since 2025 — not improving, despite AI production capability climbing every quarter. IBM's own CEO study puts it more bluntly: only around 25% of AI initiatives deliver the ROI they were sold on, and just 16% have scaled past a pilot.
On the small business side, the picture flips. A Thryv-backed survey of small businesses found real, measurable gains that aren't hypothetical:
- 70% say AI increased revenue over the past year
- 3.7x average return on AI tool investment
- 80%+ report productivity gains in the functions where they've deployed it
- 6.4 hours saved per knowledge worker, per week
- 55% say it's already reduced costs
Those are the actual benefits of AI in business, happening right now, in businesses with a fraction of the enterprise budget. Not projected. Not a five-year roadmap. Booked.
Why the giants are losing to the minnows
Here's the part that should stop you: SAP ran the same research and drew the same conclusion enterprises are quietly admitting to themselves. The determining factor between AI success and AI failure isn't the model, the budget, or the vendor. It's leadership, workforce skills, and whether the organisation can actually transform around the tool.
The benefit was never a feature of the software. It was always a function of who'd done the reps.
Think about what that actually means. A Fortune 500 company can license the same frontier model you use. Same context window, same reasoning, same weekend release notes. What it can't buy off the shelf is a workforce that's spent months actually using the thing — testing it against real decisions, building judgment about when to trust it and when not to, wiring it into how work actually gets done instead of bolting it onto a slide deck for the board.
Enterprises run AI through committees, six-month procurement cycles, and a training rollout that reaches maybe a third of the staff who'll touch it. By the time the tool is "approved," the model it was approved for is two versions old. Small businesses skip every one of those steps. The owner tries it on Monday, keeps what works, and is compounding the advantage by Friday.
The one number both studies agree on
Here's the twist. Even the small businesses winning right now aren't done — 70% of them say they need more training to use AI effectively. They're already ahead of the enterprises and they know the gap between "using it" and "being dangerous with it" is still wide open.
That's the actual finding buried in both studies. It's not "AI works for small business and not for enterprise." It's that the benefit compounds with skill, not spend — and right now, skill is the cheapest, least-defended asset in the entire AI conversation. Everyone's racing to buy more tools. Almost nobody is racing to get good at the ones they already have.
You have the structural advantage the data describes: no committee, no six-month rollout, no training program that only reaches a third of your team — because your team is you. The gap between you and a company with 100x your resources isn't the tool anymore. It's the reps.
Close that gap and the "benefits of AI in business" stop being a stat you read about and start being a number on your own P&L.
The tool was never the advantage. The advantage was always going to whoever used it first.
Cheers,
Pete