Adoption is near universal. Value capture is rare. The gap between those two facts is the most consequential thing happening in business technology right now — and it's the reason Think Modern exists.
Almost every organization is now using AI somewhere. A small minority are getting measurable financial return. The reporting on why is contradictory enough that a reasonable executive could read three credible sources and come away with three different conclusions.
Cutting through that is the point of this page. Across every source we track, the difference between the two groups comes down to the same handful of things: data quality, unclear ownership, no measured baseline, and a pilot chosen for how it demoed rather than what it moved.
That's a solvable problem. It's also not a technology problem — which is exactly why buying more technology hasn't solved it.
We track the published research on AI adoption continuously. Below is what it says, and our reading of what each finding means in practice.
PwC's research found most chief executives unable to point to a measurable return. The barriers they name have shifted from technical to organizational — the tools work, the adoption doesn't.
Our reading
Return is invisible when nobody measured the baseline. Usually the value exists but was never captured in a number anyone can defend.
Read the analysis →Gartner attributes the forecast to unclear ROI and weak risk controls — governance failures rather than capability limits.
Our reading
Projects get cancelled because success was never defined. We've catalogued the five patterns these follow, and each announces itself early.
Read the five patterns →Survey after survey puts data quality and availability at the top of the obstacle list — ahead of budget, talent, and technology.
Our reading
Most AI programs fail in the plumbing. Assessing data readiness takes a week and routinely saves a quarter. It's the first thing we check.
How we sequence it →BCG's ten-twenty-seventy rule: ten percent algorithms, twenty percent technology and data, seventy percent people and process.
Our reading
Budgets are usually allocated in reverse proportion. A company spending seventy percent on technology has decided, without noticing, to solve ten percent of the problem.
Why sequence beats spend →The companies building these systems publish their own adoption research, and it paints a considerably more optimistic picture than the independent surveys. We track both, because the gap between them is more informative than either number alone.
From a survey of 3,466 senior leaders across 24 countries: 52% had deployed AI agents in production, and among those reporting productivity gains, 39% said productivity had at least doubled.
Source: Google Cloud, ROI of AI Study.
Why this differs from the independent data →From a survey of over 500 technical leaders. The obstacles named: integration challenges (46%), data quality requirements (42%), and change management (39%) — the same list the independent research produces.
Source: Anthropic, 2026 State of AI Agents Report.
Read our analysis →Analyzing actual usage rather than opinion, roughly a 4 percentage point higher success rate among experienced users — holding after controls for task, model, language, and country. The report reads it as learning-by-doing.
Source: Anthropic Economic Index, March 2026.
Why this is the finding that matters →OpenAI's enterprise research reports daily time savings at that scale. Meanwhile Meta, per Bloomberg reporting, in February 2026 became the first major technology company to tie employee performance reviews to AI usage.
Sources: OpenAI, State of Enterprise AI; Bloomberg.
What time saved does and doesn't mean →You'll encounter a claim that ninety-five percent of enterprise AI pilots fail, repeated widely — including by firms selling AI consulting. We give it less weight than most.
It originated in a preliminary working paper, self-described as such, produced by a project with a commercial interest in the solution it recommended. Several outlets reported a sample size larger than the paper itself claimed. The concern underneath is real; plenty of pilots do stall. But the specific number is shakier than its ubiquity suggests.
We mention this partly because it matters and partly because it's a useful test of anyone advising you. A firm that repeats a convenient statistic without checking it will apply the same standard to your business case.
Read: the labs say it's working, independent research says it isn't →
The Modern Readiness Assessment answers that in thirty minutes, with a scorecard you keep either way.