The Future of Work: Hybrid Models, AI and the Changing Office

Context: two competing narratives, both partly true

Two contradictory storylines about the future of work have run in parallel through 2024 and 2025. One says hybrid working has become a durable, settled norm for a large share of the UK's knowledge-work economy. The other says high-profile employers are dragging staff back to the office full-time and hybrid's moment has passed. Both are supported by real evidence, because both are happening at once, in different sectors and companies, rather than resolving into a single national trend.

The data: where hybrid working actually stands

ONS survey data — drawn from its Opinions and Lifestyle Survey and labour market statistics — has consistently found a substantial minority of UK working adults, historically running at roughly a quarter to a third depending on the specific survey wave and which sectors dominate the sample, working in some hybrid arrangement combining office and home or other remote locations. That share rose sharply during and immediately after the pandemic and has since broadly stabilised rather than continuing to climb, suggesting the market found something closer to an equilibrium rather than a one-way ratchet toward ever-more remote work.

Against that backdrop, several large employers moved decisively in the opposite direction. Amazon's mandate for a full five-day office return, taking effect from January 2025, was among the highest-profile reversals of pandemic-era flexibility policy anywhere in the world, and prompted renewed debate in the UK about whether other major employers would follow. Other large firms took more measured positions — increasing minimum in-office days without eliminating hybrid entirely — reflecting a genuinely divergent employer picture rather than a single industry consensus.

What's changing: the productivity evidence is getting more nuanced, not more conclusive

CIPD research and comparable UK workplace studies through 2024-25 continued to find that employees in hybrid arrangements report higher job satisfaction on average, alongside persistent concerns about visibility for career progression relative to office-based colleagues, and genuine difficulty with certain categories of collaborative work. The clearest pattern in the productivity research is task-dependence: individual, focused work measures equal or better when remote in most studies; collaborative, iterative work measures worse; and onboarding new employees into team norms and informal knowledge remains consistently harder without regular in-person contact.

"The mistake most organisations made early on was treating 'remote work' and 'office work' as a single productivity question with a single answer. The better question is which specific tasks, for which specific teams, benefit from which specific arrangement — and that answer varies enormously by function." — a distinction that has become close to consensus across CIPD and comparable workplace research bodies by 2025.

What AI is actually doing to jobs, according to the current evidence

Separately from the hybrid-working debate, AI's labour market impact has become more empirically grounded as more real-world deployment data has accumulated. International Labour Organization research estimates that roughly a quarter of jobs globally have high exposure to generative AI automation, concentrated heavily in clerical and administrative work — data entry, routine correspondence, scheduling, basic document processing — where current large language models can already perform meaningful parts of the task. Roles requiring physical presence, complex professional judgement, or high-trust interpersonal relationships show materially lower exposure in the same research.

Bank of England analysis and mainstream labour economics have converged on describing the likely pattern as task-level automation within roles rather than wholesale elimination of entire occupations — legal document review, first-draft writing, code generation and customer service triage are being automated as discrete tasks inside jobs that also retain tasks AI cannot yet perform well, rather than entire job categories disappearing overnight.

What it means for you

If your role sits heavily in the ILO's high-exposure clerical and administrative category, the realistic planning horizon is proactive skill development toward the parts of your role that involve judgement, relationship management or complex problem-solving that current AI genuinely struggles with — not an assumption that automation is a distant, theoretical risk. If you are negotiating hybrid arrangements with an employer, the CIPD evidence base gives you a genuinely stronger case for retaining flexibility on individually-focused work than on roles built around dense, iterative team collaboration, where employers increasingly have real productivity evidence, not just preference, behind office-attendance requirements. For the compensation side of this conversation, our guide on negotiating a higher salary covers how flexibility itself has become a negotiable term alongside pay in many 2025-26 job offers.

Generational differences add another layer employers are still working through. Younger employees entering the workforce since 2020 have, in several UK workplace surveys, reported valuing flexibility and hybrid arrangements more highly relative to salary than older cohorts who built their careers around full-time office attendance — a mismatch that shows up concretely in graduate recruitment, where employers offering only full-time office roles increasingly report needing to compete harder on other terms, including pay, to match the appeal of hybrid-offering competitors for the same talent pool, a dynamic closely tied to broader shifts already reshaping gig and flexible employment patterns across the UK economy.

What to watch next

Watch whether more large employers follow Amazon's full return-to-office model through 2026, or whether the initial wave proves to be an outlier rather than the start of a broader reversal — UK-specific data on this will lag US trends by some months given differing corporate cultures and tighter UK employment protections around unilateral contract changes. On the AI side, watch ILO and OECD updates to their exposure estimates as more sectors report real deployment data rather than projected exposure, since early estimates based on task descriptions alone have historically proven less reliable than data drawn from actual observed automation in deployed systems.