AI Adoption Is Increasing, but Results Remain Uneven

New research paints a more cautious picture of AI adoption in the workplace, which contrasts sharply with the early hype around the technology. Investment is accelerating, employers are redesigning roles, and hiring expectations are changing. Yet the evidence also suggests that AI adoption has not yet delivered consistent organisation-wide productivity gains. For employers, the central issue is no longer whether AI adoption will affect the workforce, but how it should be implemented, governed and integrated into job design.

Productivity Gains Remain Uneven

Gallup’s State of the Global Workplace 2026 report found that while AI adoption is improving productivity for many individual workers, those gains are not yet translating into broad organisational transformation. Among employees in organisations that have implemented AI, 65% said it had a positive impact on their personal productivity. However, only 12% strongly agreed that AI had transformed how work gets done across their organisation.

This distinction matters. AI adoption may help employees draft documents, summarise information, analyse data or automate routine tasks, but those improvements do not automatically reshape workflows, reporting lines, decision-making or workforce structures. Gallup’s findings suggest that the real productivity challenge is not simply giving employees access to AI tools, but ensuring that managers actively support AI adoption and that AI is properly embedded into day-to-day processes.

Wider economic evidence points in the same direction. Survey evidence reported by CEPR found that 89% of firms reported no impact from AI on productivity over the previous three years, although firms expected larger effects in future. This reinforces the view that, while AI adoption is widespread, measurable productivity gains remain concentrated rather than universal.

Workforce Strategy Is Changing, but Not Simply Through Job Cuts

AI adoption continues to raise concerns about job displacement, particularly in sectors where routine cognitive work can be automated. However, as we highlighted in a recent article, research suggests many organisations are shifting away from viewing AI adoption mainly as a substitute for employees.

This is supported by new research conducted by EY-Parthenon, which found that while most business leaders expect AI adoption to reshape workforce strategy, relatively few now expect it simply to reduce hiring (20%, compared to 46% in 2024). Instead, employers are focusing on redesigning roles (44%), reskilling existing employees (42%) and building AI-related capability.

That is an important shift. It suggests that AI adoption is not only replacing tasks, but changing how roles are structured. Employers may need fewer employees for some activities, but greater capability in areas such as oversight, prompt design, data interpretation, governance, compliance and customer-facing judgement. The employment risk is therefore not limited to redundancy. It also includes role drift, skills gaps, uneven workloads and potential disputes where job expectations change faster than consultation, training or contractual documentation.

Investment Is Rising, but Implementation Is Often Weak

Organisational investment in AI adoption remains high. New research by Orgvue found that 92% of organisations had invested in AI, with 83% planning to increase spending. However, the survey also found that 78% of organisations had encountered difficulties with AI projects, with 35% reporting outright failures and 43% saying projects had stalled at the pilot stage.

The reasons are revealing. The survey found that 57% of organisations adopted AI primarily because competitors were doing so, suggesting that some investment is being driven by competitive pressure rather than clear workforce planning. Many employers also reported concerns about choosing the wrong tools, insufficient expertise, structural barriers and employees using AI without proper oversight.

This is highly relevant to workforce strategy. AI projects are less likely to succeed where they are treated purely as technology purchases. Employers need to understand which roles are affected, which tasks can be augmented or automated, what controls are needed, and how work will be redesigned. Without that analysis, organisations risk investing heavily without achieving meaningful operational change.

Hiring for AI Capability Creates New Risks

AI adoption is also reshaping recruitment. A new survey by TestGorilla has found that 53% of hiring managers now place greater value on practical AI capability as opposed to specialist technical knowledge. Nevertheless, 59% of organisations reported making a poor AI-related hire in the previous year, highlighting a clear gap between the skills employers are seeking and the methods they use to assess them.

This points to a practical problem. Many employers can identify confidence with AI tools, but not necessarily competence. A candidate may speak convincingly about AI, but lack the ability to apply it safely, accurately or productively in a specific role. TestGorilla’s findings suggest that hiring processes need to move beyond interview answers and tool familiarity towards practical assessments, work samples and role-specific testing.

For employers, this also has legal and operational implications. Poor AI hires may contribute to failed projects, weak governance, data risks, bias, inaccurate outputs or over-reliance on unverified systems. As AI becomes embedded into work, employers will need to assess not only technical fluency, but judgement, accountability and understanding of organisational risk.

UK Employers Face a Practical AI Adoption Gap

Research also suggests that many businesses are still at an early stage of AI adoption. The British Chambers of Commerce reported that AI adoption among SMEs has risen sharply, from 25% in 2024 to 54% in 2026. However, most firms are still using generic AI tools rather than deeply integrated systems, and more than nine in ten reported no impact on headcount over the previous 12 months.

That supports the wider pattern: AI is being used, but often in a relatively limited way. Many employers are experimenting with tools before fully redesigning work. This may reduce immediate job loss concerns, but it also means productivity gains may remain modest until organisations invest in training, process redesign and governance.

Governance, Trust and Workforce Impact

The risks of AI adoption are not confined to productivity. Employers must also consider transparency, employee trust, data protection, bias, monitoring and the potential use of AI in performance management or workforce surveillance. Recent commentary has warned that AI may widen workplace inequalities where it empowers some employees while increasing control and monitoring over others.

This creates a wider employment issue. Where AI is used to allocate work, monitor output, assess performance, screen candidates or inform disciplinary decisions, employers will need clear policies, human oversight and defensible decision-making processes. Poorly governed AI systems could increase legal risk rather than reduce it.

What Employers Should Take From the Evidence

The evidence points to a clear conclusion: AI in the workplace is moving fast, but successful implementation is not automatic. Productivity gains are most likely where AI is integrated into workflows, supported by managers, linked to workforce design and accompanied by proper training.

Employers should therefore avoid treating AI as a quick route to headcount reduction. The stronger approach is to identify affected tasks, assess skills gaps, redesign roles, train managers, update policies, and ensure that AI adoption is subject to appropriate governance. In practical terms, AI should be viewed less as a standalone technology project and more as a workforce transformation programme.

Organisations that take that approach are more likely to convert AI investment into measurable value. Those that rush implementation, rely on vague assumptions, or hire for AI capability without proper assessment may find that the promised productivity gains remain out of reach.