Why your rollout checklist won't move adoption
A board member asks you to prove AI adoption is working. You pull up licenses activated and trainings completed. None of it answers the question your AI adoption framework is supposed to answer: are people actually changing how they work?
Most rollout plans still read like a deployment checklist: buy licenses, run a training session, publish a dashboard. An AI adoption framework built that way treats adoption as a logistics problem, not a people problem.
Employees who aren't using AI rarely say the tool is hard to find. They give reasons that have nothing to do with deployment speed: ethical doubts, habit, doubts the tool even fits their job. Fix deployment and those reasons stay untouched.
The fix is sequencing, not more tooling: start an AI adoption framework with the barriers your workforce actually reports, then design interventions around them, not the other way round.
What employees say is holding them back
Gallup's February 2026 study of 23,717 U.S. employed adults found that 43% of employees who never use AI describe themselves as ethically opposed to using it at work.
That number should reframe how you read your own non-user data. We'd treat that as a judgment call, not a training gap waiting to be closed.
The same survey reports that a preference to keep working the way they already do shows up as a common reason among both non-users and employees who use AI only occasionally.
- Ethical opposition to using AI at all
- Doubts that AI is relevant to their specific job
- Preference for the way work already gets done
Sources: AI in the Workplace: What Separates Adopters and Holdouts
Why manager behavior changes the numbers
A separate Microsoft-led study of 1,800 workers globally found that when managers actively modeled AI use, employees reported markedly higher perceived value, more critical thinking and more trust in the tools than employees whose managers stayed hands-off.
That's a self-reported association from one study, not proof that manager modeling causes the lift, but it points at a lever deployment checklists skip entirely: the manager, not the tool.
We'd rather put a manager-enablement track on the framework from day one than wait for a usage dip to justify it: by the time usage drops, the habit gap has already set in.
Ganesh Harinath, founder and CEO of Fiducia AI, made a related point on the forum floor at IBM Think 2026, talking about what actually shifts behavior day to day:
“The key is to make AI the standard operating layer inside the tools employees already use”
— Ganesh Harinath, Founder and CEO of Fiducia AI · Source: Only 25% of workers are using AI. Here’s how tech leaders are changing that. | IBM
Sources: 2026 Work Trend Index report: Agents, human agency, and opportunity
Three pieces every AI adoption framework needs beyond deployment steps
Once you know which barriers are real for your workforce, the AI adoption framework needs structure to act on them. Three pieces cover ground deployment checklists miss.
| Component | What it addresses | Example signal to track |
|---|---|---|
| Barrier inventory | Ethics, usefulness and habit, by group | Share of non-users citing ethical opposition |
| Manager enablement | Modeling, quality standards, room to experiment | Managers observed using AI in team-facing work |
| Governance anchor | Risk and trustworthiness built in from the start | Framework controls mapped to a recognized standard |
Most programs treat governance as a compliance appendix bolted on after launch. We think that's backwards: trust and risk questions belong in the framework from the first pilot, not the retrofit.
Sources: AI in the Workplace: What Separates Adopters and Holdouts, 2026 Work Trend Index report: Agents, human agency, and opportunity, AI Risk Management Framework | NIST
Where NIST's AI RMF fits into governance
NIST's AI Risk Management Framework is a voluntary standard, not a compliance mandate, meant to help organizations build trustworthiness considerations into how they design, develop, use and evaluate AI systems.
NIST ties the framework to an ongoing revision connected to a White House AI Action Plan, so treat today's version as a living reference for your AI adoption framework, not a finished checklist to copy once.
Some leaders wait for a framework to be finalized before writing any policy. Our view: start mapping your own risk categories against NIST's structure now; a living standard still beats no structure at all.
Sources: AI Risk Management Framework | NIST
Measure your barriers before you choose interventions
Treat all of this as a map you build before you commit budget, not a one-time audit. Gallup's data shows non-users and occasional users report different dominant barriers, so the fix that matters differs by group, not by a single rollout plan.
AI Adoption Score's AI Adoption Barriers Survey is built for exactly this question: which of ethics, usefulness or habit are actually present in your teams, surveyed directly rather than assumed company-wide.
Reframe the question from 'how many people activated the tool' to 'which barrier is actually in the way here.' That's the version of the framework your board actually needs to see.
Sources: AI in the Workplace: What Separates Adopters and Holdouts
Frequently asked questions
How often should we re-survey employees on AI adoption barriers?
Re-survey on a recurring cadence rather than once at launch. Gallup's 2026 survey found non-users and occasional users report different dominant barriers, so the right intervention differs by group, and a single snapshot risks assuming one barrier profile fits everyone.
Who should own the AI adoption framework: IT, L&D or the AI team?
As editorial guidance: put a Head of AI or equivalent in charge of the whole framework, with L&D owning training content and managers owning day-to-day modeling. Splitting ownership by silo is how barrier-specific pieces get dropped.
Does a manager-enablement track replace formal AI training?
No, it complements it. Formal training builds skills, while manager modeling shapes whether people feel safe applying those skills day to day; the Microsoft-led study links the two to different, reported outcomes.
Is NIST's AI RMF required for a private-sector AI adoption framework?
No, NIST describes the AI Risk Management Framework as voluntary, not a compliance mandate. You can still use its structure as a reference point for your own governance component, even without a legal requirement to adopt it.
Sources: AI in the Workplace: What Separates Adopters and Holdouts, 2026 Work Trend Index report: Agents, human agency, and opportunity, AI Risk Management Framework | NIST
