Why adoption stalls once the early adopters are done
Six months ago your pilot group was thrilled. Now usage has flattened, the early champions have moved to other projects, and your board wants to know why the numbers aren't climbing. This is an AI adoption plateau, and it rarely means the tool failed.
According to IBM's 2026 Institute for Business Value CEO study of more than 2,000 chief executives, only 25% of workers used AI regularly as part of their job in 2026, even though 86% of CEOs believed their people were ready for it. Employees reported a similar split in McKinsey's 2026 State of AI survey: 80% said AI had improved their individual productivity, yet only 37% of respondents attributed any EBIT impact to AI use, a figure McKinsey described as essentially unchanged from 2025.
BCG's surveys of software developers, which the firm treats as a leading indicator for broader white-collar work, found that more than 85% of employees stayed at the task-assistance or delegation stages of adoption, short of the semiautonomous-collaboration stage where BCG says real value creation starts.
In the same BCG global survey of 1,400 C-suite executives, 62% named a shortage of talent and AI skills as their biggest obstacle to getting value from AI.
“AI adoption will not continue as slow, linear growth”
— Ganesh Harinath, Founder and CEO, Fiducia AI · Source: Only 25% of workers are using AI. Here’s how tech leaders are changing that. | IBM
Culture Amp's 2026 benchmark, drawn from 123 organizations and about 112,000 employees, found 85% of employees said their organization encouraged AI exploration, but only 58% said leaders had clearly explained how AI would help the organization reach its goals, a 27-point gap between permission and purpose.
We think most plateaus get blamed on the tool when the real stall sits in leadership clarity or team culture, and swapping vendors rarely moves a flat adoption curve.
Sources: The State of AI: Global Survey 2026 | McKinsey, AI Adoption Puzzle: Why Usage Is Up But Impact Is Not | BCG, Culture Amp’s 2026 AI at Work Benchmark: 85% Are Encouraged to Use AI. 42% Don’t Know Why. | Culture Amp
Editorial exercise: diagnose your plateau in under 30 minutes
Here's a short editorial exercise: score your organization's AI adoption plateau across four categories in under 30 minutes, then name your single biggest blocker before you spend another dollar on training or tools. This is a self-assessment heuristic, not a validated instrument.
Prerequisites: run this alone or with two or three colleagues who see different parts of the organization, such as an L&D lead, a frontline manager, and someone from IT. You don't need survey data yet, just an honest best estimate of what's actually happening.
Inputs: gather whatever you already have before you start scoring.
Pick four AI adoption KPIs to score right now
Score each category from 1 (barely present) to 5 (strong), using your honest best estimate. The table below names the category, a question to ask, and a signal that tells you where the number should land.
| Category | Question to ask (suggestion) | Signal to watch for |
|---|---|---|
| Usage | Who used AI this week versus just once this month? | A handful of names repeat; most people go quiet after week one |
| Skills and training confidence | Can people say what they'd ask AI to do for their own role? | Vague answers, or "I haven't really tried it yet" |
| Barriers: leadership clarity and trust | Have leaders explained how AI fits the team's actual goals? | People shrug or guess when asked why the organization adopted AI |
| Culture: peer visibility and manager behavior | Do people see colleagues or managers using AI openly? | AI use happens quietly, if at all, and managers rarely mention it |
Add the four scores together. As an editorial heuristic, a total above 16 suggests adoption is broad but still shallow and worth probing further, while a total below 10 points to one or two categories dragging the rest down. Look at your lowest single score first.
We'd treat this as a gut check, not a verdict: good enough to point you toward one conversation to have this week, not to approve a training budget.
What a rough score can tell you, and what it can't
Every number in this exercise is self-reported, yours or your colleagues' best guess, not an independently measured behavior. Treat the scores as a hypothesis worth testing, not proof of what's actually happening on your teams.
Strategy confusion as a named barrier isn't new. In McKinsey's 2018 Global Survey of 2,135 respondents, a lack of clear AI strategy was the most commonly cited barrier to adoption, though that survey predates generative and agentic AI tooling and shouldn't be read as a current benchmark.
The pattern it named, confusion about direction, still shows up in the leadership-clarity gap Culture Amp reported in 2026. That's a recurring category, not a figure you should treat as current from the 2018 source itself.
We're skeptical of any adoption score that claims precision after ten minutes of guessing; a rough score earns you a short list of suspects, not a sentencing.
Sources: Adoption of AI advances, but foundational barriers remain | McKinsey, Culture Amp’s 2026 AI at Work Benchmark: 85% Are Encouraged to Use AI. 42% Don’t Know Why. | Culture Amp
Name your single biggest blocker
You're done with the exercise when you can say, in one sentence, which category pulled your total score down the most, and name one concrete example that backs it up.
In our reading of the findings above, two patterns show up often. A high usage score paired with a low culture score often means a handful of enthusiasts are carrying the whole average, not real organizational traction. A high culture score paired with a low leadership-clarity score often means people feel free to experiment but don't know why it matters for their actual job.
Before you settle on one blocker, run the same four scores separately for managers, individual contributors, and executives. Experiences often differ sharply between levels, and a single combined score can hide a gap that matters to one group but not another.
Our bet is that most plateaus labeled a skills gap are really a leadership-clarity gap wearing a skills costume.
Turn your guess into a measured diagnosis
A 30-minute self-score is a starting hypothesis, not a finished diagnosis of your AI adoption plateau. It narrows five possible blockers down to one or two worth investigating properly.
The next move is to validate that hypothesis with real input from the people doing the work, segmented by team and role, before you redesign training or rewrite your AI messaging. Guessing got you this far. Let the next step be checking.
Frequently asked questions
How often should I re-run this self-assessment?
Treat it as a quick check before any budget decision, not a recurring metric. Re-run it whenever usage looks flat again, after a training push, or roughly once a quarter, and compare the pattern of scores rather than any single total.
Who should be in the room for the exercise?
Include people who see different slices of the organization: an L&D lead, a frontline manager, and someone from IT or operations. Three honest perspectives catch blind spots that one person's guess will miss.
What if two categories score equally low?
Pick the one with the clearer concrete example and act on it first. Trying to fix two blockers at once usually dilutes both efforts; segmenting by role or team often reveals that one of the two is really a subgroup problem.
Is a 30-minute score enough to justify a training budget?
No. Treat it as a hypothesis that narrows your attention, then confirm it with a structured employee survey before committing real spend. Self-rated guesses, even careful ones, can miss what's actually happening on the ground.
