Workers think AI could do half their job by 2029. Is HR ready?

BCG’s latest AI at Work research suggests the workplace AI debate has moved into a more urgent phase. Based on a survey of more than 11,000 leaders, managers

and frontline employees across 14 countries, the firm’s fourth annual report finds that 61% of workers believe AI agents will be able to do at least half of their job within three years.

Yet half of companies still have no clear rules for how humans and AI should work together. Some 88% of workers say they will need major upskilling in the next five years, yet only 36% feel properly trained today. That training gap has not moved in a year.

AI is now part of the work experience for many employees but the research shows there is a question mark over whether organisations are building the skills, expectations, safeguards and trust needed to use it well.

What does BCG’s AI at Work research show?

BCG’s latest AI at Work research shows that AI use is becoming mainstream across the workforce, including among frontline workers. Employees are reporting productivity gains, greater enjoyment and more confidence in some areas, while also experiencing fear, mental strain and uncertainty about what AI means for their jobs.

The report’s headline finding is that 61% of workers think AI agents could do at least half their job by 2029. AI agents are systems that can act with a degree of autonomy to complete tasks, make recommendations, interact with other systems or support decision-making. For HR, that raises practical questions about job design, accountability, capability and performance expectations.

The research also finds that AI has already taken over simpler tasks in many areas of work. That may sound like progress but it changes what employees are judged on. When routine work is automated or accelerated the remaining human contribution can become more complex. Judgement, creativity, escalation, verification, empathy, collaboration and ethical decision-making become more visible.

This is why the skills question is central. If 88% of workers expect to need major upskilling in the next five years, while only 36% feel properly trained now, the issue is capability in the present and not the future.

AI adoption is racing ahead of workplace readiness

AI adoption is no longer confined to pilots, digital teams or knowledge workers experimenting at the edges of their job. BCG’s findings suggest it is entering everyday work at speed, with frontline adoption rising sharply.

The firm says 74% of frontline workers are now regular AI users, up 23 percentage points in a single year. This is at a time when much of the organisational conversation about AI still centres on office-based workers, productivity tools and corporate functions. The new data suggests AI is already reaching the shop floor, service counter, warehouse, field site and customer-facing environment.

Organisations have, however, not caught up. BCG finds that only 28% of frontline employees see real alignment between what their leaders say about AI and what actually happens on the ground.

This figure points to a credibility gap. Leaders may talk about transformation, efficiency or empowerment, while employees experience unclear expectations, uneven access to tools, weak training or mixed messages from managers.

The risk for HR is that AI adoption becomes highly individualised. Employees work out their own ways to use tools and teams develop informal norms. Some workers gain a productivity advantage while others feel exposed, anxious or left behind. This is not a sustainable operating model for AI-enabled work.

The productivity prize is real but guidance is missing

BCG’s research points to significant productivity gains. Some 42% of workers say they are saving at least a full working day per week thanks to AI. This is a substantial opportunity for organisations under pressure to improve performance, reduce waste and increase responsiveness.

Yet the finding that 66% of workers get no guidance from their employer on what to do with that time should concern HR leaders. Time saved through AI does not automatically become value. It can be absorbed by more tasks, higher expectations, increased pace or fragmented work. It can also create pressure if employees feel they must conceal efficiency gains for fear of being given more work or having their role reduced. Finally, it can widen differences between teams if some managers reinvest time in learning, service or innovation while others simply expect faster output.

This is where HR has a clear role. Organisations need to decide what productivity gain means in human terms. If AI saves time, should employees use that time for customer service, learning, problem solving, innovation, relationship building or recovery from work intensity? Should performance measures change? Should job descriptions be updated? Should teams have explicit conversations about how AI-enabled capacity will be used?

Without this productivity becomes a vague promise rather than a managed outcome.

AI is increasing enjoyment and strain at the same time

According to the research 67% of workers say AI has increased their enjoyment at work. At the same time 41% say AI has increased their mental strain.

This shows that AI can make work more satisfying by removing repetitive tasks, speeding up information retrieval, supporting creativity or helping employees feel more capable but it can also create extra cognitive load. Workers may have to check outputs, learn new systems, manage uncertainty, handle faster work cycles or worry about whether their skills remain valuable.

This is the human reality of AI adoption. The emotional response is mixed because the work itself is changing.

What should HR leaders do now?

HR leaders need to treat AI adoption as a workforce transformation issue. The BCG findings suggest that many organisations are still approaching AI as a technology deployment, productivity project or individual learning challenge. This is too narrow for the scale of change now emerging.

The first task is to set clearer rules for human-AI working. Employees need to know where AI use is expected, where it is permitted, where it is discouraged and where human judgement must remain visible. This includes guidance on verification, confidentiality, bias, escalation, decision rights and accountability.

The second task is to update skills strategies. Prompting is only one part of AI capability. Workers also need to understand how to check AI outputs, challenge recommendations, interpret data, communicate uncertainty and decide when a human decision is needed. For managers the capability requirement is even broader. They need to help teams make sense of AI, redesign work, discuss fear honestly and prevent productivity gains from turning into unsustainable pace.

The third task is to make work redesign explicit. If AI agents can do a growing share of tasks, jobs need to be examined at task level. HR should be asking which tasks are being automated, which are being augmented, which new tasks are emerging and how performance expectations should change.

The fourth task is to listen more closely to frontline workers. BCG’s finding that only 28% of frontline employees see alignment between leadership messages and ground-level reality should be treated as a warning. Frontline workers often experience technology in the most operationally unforgiving conditions. Their insight is essential to designing tools and workflows that people can trust and use. For a deeper look at what this means in frontline environments, read our analysis of Starbucks, Josh Bersin’s frontline workforce research and why AI must be designed around the real conditions of work. (LINK HERE)

The fifth task is to measure the human side of AI. Productivity data alone will not show whether AI is improving work. HR should also track confidence, perceived skill readiness, mental strain, trust in leadership, manager support, work intensity and employee understanding of AI rules.

The most important shift for HR is to move the AI conversation closer to work. Broad statements about productivity, transformation or future skills are no longer enough. Employees need practical clarity on how AI changes their job, what skills they need, how their contribution will be valued and what support they can expect.

This is where HR leadership matters. The organisations that handle this well will treat AI as a redesign of work and capability, supported by trust and clear rules. The organisations that leave employees to work it out alone may still see adoption rise, but adoption without readiness is a weak foundation for lasting value.