AI at Work

From curiosity to everyday companion

There’s a wide range of AI tools available on the market today — from AI assistants like ChatGPT or Gemini, through to AI agents that can make autonomous decisions and perform actions to complete an objective.

With growing awareness and popularity of these tools, it’s unsurprising that 78% of workers say they are already using AI in some capacity in their role (Fig. 1). In fact, for many users, AI tools have become an everyday companion, as 97% use them weekly, with nearly two-thirds (62%) using them daily.

But despite significant adoption and usage, there’s a bigger divide hidden — the one between managers and the rest of the workforce.

Among managers, AI adoption sits at 79%, while only 54% of non-managers say they use AI in their work (Fig. 2). Time in role also matters: those with six to fifteen years’ experience report the highest adoption(82%), compared with just 68% of those who have been in their role for more than fifteen years.

The sector people work in can also change their relationship with AI entirely. In financial services, 68% of AI users say they engage with it daily (Fig. 3). In education, that figure drops to 49%. Manufacturing sits at 65%, healthcare and pharma at 62%, and retail at 59%. On average, across all sectors, employees are interacting with AI an average of 6.5 days a week, suggesting AI use is becoming as routine as sending an email for some.

When asked about their most common uses of AI in the workplace, respondents noted data analysis, summarizing meetings and writing reports as the top 3 tasks they use AI for most. Somewhat unsurprisingly there are regional differences in how respondents use AI. While all markets surveyed rank data analysis within the top 3, both the UK and the US also include research in the top 3 alongside summarizing meetings. By contrast Germany ranks proofreading and translating languages in the top 3 with data analysis.

How AI enters the workplace

The spread of AI in the workplace is rarely the result of a single, top-down mandate. Instead, adoption comes through multiple entry points, with the majority of respondents (32%) being AI self-starters (Fig 4.), recognizing that they began using AI tools at work without any prompting.

Regional differences matter here too. Interestingly respondents in the US (13%) and Germany (14%) were more likely to have an AI mandate from leadership than those in the UK (9%). While UK respondents were more likely to have had AI tools provided by their IT team (30%) than those in the US (23%) or Germany (23%).

When it comes to driving adoption of AI, encouragement is more common than enforcement, 57% say they’ve been encouraged to use AI in the workplace, and in the UK and US, explicit encouragement reaches 32% (versus 23% in Germany). Among managers, 57% feel encouraged to use AI, compared with just 34% of non-managers (Fig. 5).

The rise of AI agents

If AI tools are now part of the furniture, AI agents are the most recent arrivals, and they’re making an impression. These autonomous systems can carry out multi-step tasks, make decisions, and integrate into workflows in ways that feel fundamentally different from earlier AI tools.

Half of employees (50%) say they are already using AI agents in their role, and 44% report they’re doing so effectively. Looking ahead, 60% expect to use AI agents more effectively within the next year, and 69% say they’re optimistic about their use — an unsurprising figure given that 67% also believe that using AI agents in their day-to-day role will make their job easier!

For most, AI agents are becoming a trusted sidekick in their working day, helping to tackle a variety of tasks.

In light of this, it’s unsurprising that when asked to rank what they felt were the most exciting aspects of AI agents at work, respondents cited time savings, increased productivity, and reduced efforts as the top three most exciting benefits (Fig. 6).

Far from replacing roles outright, this excitement points to AI agents as accelerators, clearing space for higher-value, human-led work.

However, again the benefits of adoption aren’t equal. In fact, respondents feel that those who see the largest impact from AI and AI agents are those in more senior positions. With both middle and senior level managers expected to reap the largest benefits (Fig. 7).

The emotional landscape of AI at work

For all the speculation about resistance to AI and AI agents, the reality is overwhelmingly positive as 88% of respondents associate positive emotions with AI use — 66% cited that it made them feel productive, 59% felt more creative, 55% felt excited and 49% felt empowered.

Negative emotions are rare. Just 13% report them. Stress (5%), confusion (4%), frustration (4%), and disconnection (3%) barely register compared to the positive sentiment.

That optimism extends into how people see the future of work. Half of respondents (50%) believe they’re more likely to be managing AI agents than people in their future careers, and 55% think managing agents would be easier than managing humans. Nearly half (47%) even think it’s likely they will be managed by an AI agent in the future (Fig. 8).

Breaking barriers

For all its momentum, AI adoption still faces significant headwinds. An incredible 94% say there are barriers to wider use in their organization, noting everything from data privacy and security concerns, to a lack of understanding, insufficient training, and even cultural stigma preventing wider adoption (Fig. 9).

Generations, confidence, and the adoption gap

One of the biggest challenges around AI and AI agent adoption appears to be an adoption gap driven by a disparity in AI confidence in the workplace.

While 69% of respondents feel very or extremely confident using AI tools, 23% are only somewhat confident and still have concerns around how to use AI tools safely. A further 8% are not confident at all (Fig. 10).

Those in the US are most confident, with 73% saying they are very confident using AI tools, compared to 69% in the UK and 65% in Germany.

Interestingly, while AI usage is highest among 25-44 year olds (who make up nearly two-thirds of all users), the youngest workers aren’t necessarily the most confident.

Among 18-24 year olds, 76% say they’re very confident using AI tools, but 4% admit to not being confident at all — a higher rate than in some older groups. Confidence peaks among 25-34 year olds (80% very confident) before declining steadily with age. By 55+, fewer than half (46%) say they’re very confident, and 21% say they’re not confident at all.

Managerial experience is again a differentiator: 70% of managers are very confident with AI, compared to just 43% of non-managers (Fig. 11).

Perhaps most telling, 38% of respondents don’t encourage their team to use AI agents because they don’t feel they understand them well enough themselves.

Trust: colleagues first, AI close behind

Given mixed confidence levels, it’s no surprise trust is one of AI’s biggest hurdles in the workplace. Part of this trust challenge stems from the fact that 45% believe using AI tools is still stigmatized as lazy or untrustworthy, while 39% feel they are judged or second-guessed when using AI.

Despite this, the gap between trusting a human and trusting a machine is narrowing. When asked 87% of respondents noted they would generally trust AI to perform a task given to it, compared to 96% that trust a colleague with the same task. However, trust in AI is higher among managers (88%) than non-managers (77%), and it’s not equal across tasks.

Formulaic, repetitive tasks like data analysis inspire 65% trust in AI agents, while more human-sensitive tasks like recruitment decisions drop to 48% (Fig. 12).

Interestingly, many respondents say they trust AI agents more than junior employees or interns, a telling sign of where perceptions of competence are shifting (Fig.13).

Training is the missing link

Training is a persistent weak spot in AI adoption. While 81% say they’ve received some form of training, just a third have had formal, company-provided instruction. Another 32% are entirely self-taught, and 16% learned only through trial and error (Fig.14).

Managers are far more likely to have been trained (82%) than non-managers (62%), and training levels differ sharply by sector. Financial services leads with just 2% reporting no training, while in education it’s 13%, in healthcare and pharma 17%, and in retail 19%.

This highly mixed and inconsistent approach to training, is perhaps why 45% of respondents noted that there is a disconnect between leadership AI enthusiasm and provided AI training within their organization. Emphasizing this point, respondents put training at the top of the list for things companies could do better to help drive further adoption (Fig. 15).

Most worryingly, despite the majority of respondents (97%) recognizing that education in AI safety, and data security is important for employees in the workplace, only 63% have received company advised or mandated training in this area. In fact 22% have had no training in this area at all, and have not sought out to educate themselves on it either. Worse, is 44% of non-managerial staff have received no training in this area, despite them recognizing the importance of it at 93% (Fig.16).

This problem lives on across industries too, where just 52% of those in education have received training on AI safety and data security, compared with 54% in retail, 58% in healthcare and pharmaceuticals, 69% in manufacturing, and 80% in financial services.