The Confidence Collapse
There’s a worrying trend happening in organizations — a growing confidence gap among workers using AI. Not resistance. Not skepticism. Ambiguity about what it all adds up to. And it’s all moving fast.
We all know AI works. From paralegals summarizing depositions, to marketing coordinators drafting social copy, to branch managers prepping for quarterly reviews — everyday, more people are using across every industry.
AI makes tasks manageable. What’s harder to see clearly is what those tasks will look like three months from now, let alone three years. And that ambiguity is making it much harder for many workers to understand what it means — for their roles, their teams, or the direction their organization might be heading.
Amplifying that anxiety, is the speed of evolution. The tools and models that worked last quarter have already been superseded. The use case they pitched as innovative and differentiating, has become table stakes. Disruption isn’t happening in a cadence — it’s compressing, cycling faster, and leaving less room for the kind of clarity that allows workers to build confidence in the tools, process, or strategy in the first place.
We Lost the Window
There used to be a shape to how organizations absorbed new technology. Disruption arrived. Then — critically — a period of relative stability followed. That window was where the real work happened: needs assessments, phased rollouts, training built around how people actually worked, experimentation that led to validated use cases. Implementation followed understanding. Trust formed because there was time for it to form.
That era is over.
AI disruption doesn’t arrive in waves with recovery periods in-between. The cycles are moving quickly and the timelines are compressing. Each new capability triggers fresh experimentation — before the last round has been fully understood. Pilots built around Q1 capabilities are outdated by Q2. Training cohorts finish onboarding and immediately encounter a tool that’s moved.
Meanwhile, the real costs — on the financial side tokenization and licensing, on the human side compounding uncertainty — hit before the value case has been established.
These kinds of adoption strategies were built for a steadier environment. They assumed enough runway existed to absorb another tool, complete another training, run another change management cycle. But that assumption no longer holds.
The statistics around this phenomenon aren’t surprising. It’s what happens when the conditions for trust formation are structurally removed.
Workers Are Using AI They Don’t Trust
A Quinnipiac survey found that over half of Americans use AI to research topics. Nearly a third of employed adults use it for work. And yet, that same survey said a whopping 76% only trusted AI-generated information “some of the time” or “hardly ever.”
Are workers using a tool they don’t fully trust to inform decisions that matter?
Our natural inclination might be to file this under “adoption challenges” or “change management.” It’s not. It’s a judgment problem. A governance problem. A culture problem.
Organizations are measuring whether people are using the tools — not whether they have any framework for knowing when to trust it, the source of its data, or the viability of the solution they used.
Usage is not literacy. Activity is not confidence. A workforce deploying tools it doesn’t trust isn’t ready. It’s complying.
The Generation That Was Supposed to Lead This
The internal narrative was tidy: Gen Z would pull adoption forward. These digital natives were already using tech like AI for everything anyway. All we need to do is point them at it.
The reality? 81% of Gen Z workers believe AI will lead to fewer job opportunities. They booed commencement speakers who talked about AI as the savior of the world.
Compared to the 66% of Boomers who feel the same way, Gen Z — the generation with the most exposure — is the most worried about displacement by it. Familiarity with the tools isn’t a measure of capability, capacity, or willingness.
Fluency doesn’t produce enthusiasm, especially when the stakes are so visible. Gen Z isn’t resisting AI. They’re reading the compression more clearly than the people driving the adoption campaigns.
Leaders Are Wrong About Their Workers — In Both Directions
Adecco surveyed 2,000 C-suite executives across 13 countries. 45% expect AI agents embedded in their workflows within 12 months. Only 30% of workers share that expectation.
Same study, but flipped — showed that 70% of workers feel ready to collaborate with AI agents. Only 39% of leaders believe that.
Both gaps are expensive. Leaders are over-ambitious on timeline — which produces rushed rollouts and workers who feel like something is being done to them. They’re also under-crediting actual workforce readiness — which means the people most equipped to move fast aren’t being asked or trusted to lead.
Miscalibrated in both directions simultaneously. That’s not a training problem. That’s a listening problem.
What the Burnout Numbers Say About All of This
The AI productivity narrative is landing inside organizations that are already stretched.
Glassdoor data shows burnout up 65% year-over-year. 65%!
Employee confidence is at a record low — only 43.8% report a positive six-month outlook. The tech sector posted the largest confidence drop of any industry: down 9.7 points.
The sector most associated with building and deploying AI is where confidence collapsed the hardest.
None of that shows up in a Copilot usage report. But it’s the environment where every rollout lands — a workforce absorbing continuous disruption with no stable ground underneath it, watching to see whether “AI will help you” means them or just the org chart above them.
What Closing the Gap Actually Requires
Adecco’s CEO said it about as clearly as possible, “AI moves at software speed. Organizational trust moves at human speed.”
The stability window — where users had time to establish and adapt — created the conditions for those speeds to reconcile. Without it, the gap expands.
The answer isn’t a better adoption program. Organizations that execute the old playbook with AI may be faster — but they are likely getting the same results, just faster.
And for some organizations, fast may be enough.
The problem is, generating more activity often comes with less trust and a workforce that’s complying without believing. What’s required is a different operating model all together. One built for continuous recalibration rather than sequential implementation. One that treats governance, judgment, and communication as ongoing practices — not one-time rollout components. One that assigns someone to answer the question workers are actually asking: what does this mean for the work I do?
Most organizations haven’t built that. Most haven’t decided who would own it.
Adoption rates are real. So is the structural break that’s making them increasingly meaningless as a measure of readiness.
The window isn’t coming back. The organizations that close this gap will be the ones that stop designing for conditions that no longer exist.
If you’re looking to adapt a more strategic approach to AI, rather than just rolling out licenses to people, StitchDX can help.
Sources: Quinnipiac University Poll (2025), Adecco AI at Work Report (2025), Glassdoor Workforce Confidence Index (Q1 2025), Fast Company / Jon Cooper — “Employee Engagement Was Built for a More Stable Era” (June 2026)