Using AI at work: Optimism rises, but so does fatigue

Artificial intelligence is no longer an experimental add-on in the workplace. It is becoming part of daily operations, shaping writing, analysis, customer interactions, scheduling, and decision-making. Yet the latest Henley Business School research shows that, for many employees, adoption is running ahead of confidence. In a survey of 2,900 full-time workers across 29 sectors, 58 percent said they felt optimistic about AI at work, up slightly from 56 percent in 2025. At the same time, 61 percent said they still feel overwhelmed by the pace of change, unchanged from last year..

Henley’s data suggests the dominant mood is not enthusiasm, but caution. Some 28 percent of respondents described “cautious” as their overriding feeling about AI at work, ahead of confidence, and nearly two-thirds—63 percent—said they sometimes choose not to use AI tools even when those tools are available to them. This matters because it points to a widening gap between procurement and practical use. Businesses may be investing in AI platforms, but employees are still making case-by-case decisions about when, and whether, they trust them.

Skill erosion?

The reasons are understandable. Workers identified three main concerns: overdependence, loss of critical skills, and difficulty spotting inaccuracies or bias. Around 42–43 percent worry they will become too reliant on AI; 35 percent fear the erosion of core skills; and 28 percent are concerned about identifying mistakes in AI-generated outputs. These are not abstract anxieties. They go to the heart of what modern work is becoming. If AI assists with drafting, summarising and pattern recognition, where does human judgment begin—and how is it maintained? As Henley’s researchers indicate, the issue is not that workers are rejecting AI outright. Rather, many appear to be using it selectively, cautiously, and without full organisational support.

That lack of support is one of the most striking findings. Six in ten respondents said their employer either does not have AI guidelines or they are unsure whether such guidance exists. While that is an improvement on last year’s 68 percent, it still leaves a majority of workers navigating a consequential technology without clear rules of engagement. AI governance in many organisations appears to be lagging behind AI deployment. This is where risk begins to accumulate, not only around privacy, confidentiality and bias, but also around trust. Workers are being told that AI will improve productivity, yet many are not being given the training or policy framework to feel secure in using it properly.

The generational story is equally revealing. Younger workers are generally more open to AI, but they are also more fearful of its consequences. Overall, 36 percent of workers fear AI could replace their role, but among Gen Z this rises to 44 percent. Yet younger workers are also more likely to trust employers to manage AI-related job changes transparently, and more comfortable with AI directing aspects of their work. This duality—greater fluency alongside greater anxiety—reflects a labour market in transition. Those entering work may be most accustomed to AI systems, but they are also the most exposed to structural changes in entry-level roles.

The Canadian situation

There is an interesting parallel with Canada. Recent Canadian findings suggest many of the same tensions are surfacing there, albeit within a policy environment that is trying to accelerate national AI uptake. An Express Employment Professionals-Harris Poll survey found that 94 percent of Canadian job seekers have concerns about AI in the workplace, with almost half worried about overdependence and creativity loss, while 78 percent fear firms may not need to hire as much because of AI. Another Canadian survey from Abacus Data found that 47 percent of employed Canadians worry AI and automation could force them to change jobs or careers, and that concern rises to 55 percent among younger adults.

At the same time, Canada’s wider AI debate is increasingly focused on whether adoption strategies are moving faster than worker protections. Commentary around the federal government’s new “AI for All” strategy has argued that while growth targets are clear, accountability around displacement, workplace auditing and support for affected workers is less developed. Statistics Canada has also reported that employment overall has continued to grow since generative AI became widespread, but growth has been weaker for younger and less educated workers, precisely the groups most vulnerable to disruption. In other words, both the UK and Canada appear to be confronting the same basic challenge: AI may be scaling faster than institutional readiness

The broader lesson is that successful AI adoption is not just a technology project; it is a workforce project. Employees do not merely need access to tools. They need training, boundaries, confidence, and a credible explanation of how AI will affect their jobs. Without that, what organisations interpret as resistance may simply be a rational response to ambiguity. The recent findings should therefore be read less as a warning against AI than as a warning against careless implementation. If businesses want workers to use AI well, they will need to treat support, governance and communication as seriously as the software itself.