AI Research · Workplace Strategy
AI Impact Report 2026
Find out how AI gives the average UK worker 10.9 hours a week back — and what the net model means for workforce planning, office space, and your people’s wellbeing.
How AI is changing the working week, and the workplace footprint.
AWA has spent more than 30 years on workplace transformation, shaping how organisations get work done. In that time, AI has reshaped knowledge work more than any technology before it, changing what organisations need from their offices, their teams, and their leaders. We built this AI impact report as part of our ongoing AI transformation and workplace strategy work, for the CHROs, Heads of Real Estate, and operations leaders now planning around it.
Behind the report sits AWA’s task-level model of 412 UK occupations under ONS SOC 2020, each decomposed into 18 task categories and scored against peer-reviewed AI exposure research. The model runs Low, Central and High scenarios across 3, 5 and 10-year horizons with sector-specific adoption discounts, and nets gross contraction against new-role creation on a timeline paced by the capability gap. Workforce stress benchmarks reference Mental Health UK’s 2026 Burnout Report.
What this means for your people
10.9 hrs
per worker, per week, freed from routine cognitive load
In a UK workforce where 91% of adults report high or extreme pressure (Mental Health UK, 2026), the recovered hours are what matter most. In high-burnout sectors the right frame is cognitive relief — the bandwidth to do core work well, and to sustain a career without burning out.
A different question for real estate
54M sq ft
net reduction in UK office demand over five years (88M sq ft gross)
The gross figure counts only the office roles AI removes. Net of the roughly 353,000 new AI-driven roles created over the same period, about 577,000 net roles are removed — and the reduction is back-loaded, deepening from ~28M sq ft at three years to ~90M at ten as creation plateaus. Size the portfolio to the net, phased, before the next lease decision.
Acting on these findings — across people strategy, real estate, and workforce planning — is what our workplace strategy team is for.
Talk to our workplace strategy team →Three ways AI changes the work.
AI does not act on jobs as wholes; it acts on tasks. Across 412 occupations that resolves into three effects — together worth 15.8M FTE of combined workforce capacity, the working-time value of 15.8 million full-time staff.
Effect 01
Substitution
AI takes on whole routine tasks and hands the time back. This is the freed time — and the only one of the three effects that reduces the office footprint.
Effect 02
Amplification
AI makes new, previously unfeasible work possible. This is added output rather than time off the clock, and it keeps people in role — so it does not reduce space.
Effect 03
Cognitive relief
In healthcare, education and the public sector the value is recovery — lifting administrative load where 91% of adults report high or extreme pressure.
Count the flow both ways.
The headline most reports give is gross — the office space freed by the roles AI removes. That is only one flow. The net model counts the other and deducts it: the office demand created by the new AI-driven roles AI brings with it.
88M
sq ft gross reduction, substitution only
54M
sq ft net reduction, after new roles
930k
gross office roles lost over five years
577k
net roles removed, after ~353k created
The net reduction is back-loaded: about 28M sq ft at three years, 54M at five and 90M at ten — the new-role offset fades as displacement keeps rising. These are central-scenario estimates, kept deliberately conservative, and every parameter is adjustable for your own organisation.
See exactly where AI lands across your workforce — and what it means for headcount and office space.
Download the report ↓AI can already do more than firms use.
Job losses do not follow a fixed time-curve. Their pace is set by how fast firms close the gap between what AI can do and what they actually use it for. As that gap closes, more of the freed time turns into real headcount reduction.
55.9%
Theoretical exposure — what AI could do
29.5%
Observed exposure — what firms use today
26.4 pp
The capability gap — the change still ahead
The headroom is large, which is why active redundancies are still modest. Gross displacement is front-loaded — already underway at the hiring margin — so the central and high scenarios, not the low one, are the better planning guide.
Contraction before redundancy.
The dominant near-term channel is quieter than layoffs: posts not created and not refilled in exposed roles — a hiring freeze plus attrition. The entry level shows it first, and it is hard to read in trailing headcount reports.
−29%
Graduate postings in exposed roles (Big Four disclosures)
−12%
22–25 hire rate vs 2022 (ONS Labour Force Survey)
68.5→63.8
Exposed-occupation posting index, 12 weeks (Indeed)
9.6 pt
Early-career unemployment gap (ONS)
Graduate postings across the Big Four: KPMG −29% · Deloitte −18% · EY −11% · PwC −6% (AWA AI Job Market Tracker, week ending 21 June 2026). Hiring-margin contraction front-loads the demand reduction, so lease and fit-out decisions made on trailing headcount will be late.
New AI-driven roles add space.
Every role AI removes flows out of office demand. Flowing back the other way is the demand the new roles create — overwhelmingly office-based knowledge roles that need desks. The net model counts them against the loss, at the same 94.2 sq ft per FTE used for the contraction.
+59k
net AI-attributed roles a year on the tracker’s current signal, and decelerating
~353k
new roles created over five years, central
0.45 → 0.28
creation-to-displacement ratio over the decade
This creation-to-displacement ratio is the single biggest source of uncertainty in the model — the dial that most changes the answer. Creation rises then plateaus — roughly 239k, 353k, then 372k roles at three, five and ten years — so the offset shrinks and the net contraction deepens through the second half of the decade.
Where it lands, by sector.
AI lands unevenly. The heaviest exposure sits in professional, legal and technology roles, at 35–38% practical impact — but it is amplification-led there, so those offices change in use more than in size. The clearest space reductions come from routine processing and administration.
Amplification-led
Legal & professional services
Among the most exposed knowledge work, but the dynamic is amplification rather than substitution. The office changes in use: more room for collaboration, less for routine processing.
Amplification-led
Technology
The most AI-exposed knowledge occupation in live data, at around 72% observed in the UK, with the strongest amplification multiplier in the model. The space effect is in how the office is used.
Substitution-led
Financial services & insurance
High displacement potential in routine processing, and one of the highest-burnout sectors. AI deployment here serves both the space case and the wellbeing case.
Cognitive relief
Healthcare & education
High AI exposure in administrative tasks, but the primary value is relief for staff — cutting burnout and improving retention, freeing professionals for core clinical or teaching work.
Cognitive relief
Public sector
Large administrative workforces with extensive form-filling, case management and reporting. AI can lift productivity and reduce stress-related absence.
Read the report
Your sector, occupation by occupation
The full report maps AI exposure, practical impact and the office-space arc across every UK occupational family — the basis for a role-by-role read of your own workforce.
What CEOs, CHROs, and workplace leaders need to know about AI.
CEO
From adopting tools to redesigning work — the readiness gap, and AI’s duty-of-care case.
CHRO
How many hours your people get back, the capability gap that sets the pace, and how to spend the time well.
Head of Real Estate
Plan to the net, not the gross — about 54M sq ft net over five years (88M gross), phased and back-loaded.
Chief People Officer
Three planning categories — displacement, amplification and relief — and where burnout sits in each.
What the report covers, in full.
Your occupation map
412 UK occupations analysed across 18 AI task categories (ONS SOC 2020). Where AI exposure sits in your sector, occupation by occupation.
The hours-freed breakdown
10.9 hours per worker per week is the headline. The report distributes this across occupation groups — from roles that recover a little to those that recover a full working day or more.
Low, Central and High scenarios
Low, Central and High across 3, 5 and 10-year horizons, with sector-specific adoption discounts — and the net model that subtracts new-role creation from gross contraction. A range to stress-test your own planning against.
The real estate calculation
The gross-to-net office-space model: about 54 million sq ft of net UK reduction over five years (88 million gross), after roughly 353,000 new AI-driven roles. The data behind your next lease decision.
The wellbeing case
91% of UK adults report high or extreme pressure. Why cognitive relief — the report’s named third effect — is the right frame for people leaders.
The data in full
Every occupation, every scenario, every score. The full dataset — AI exposure, practical impact, hours-freed and net office-space figures for all 412 UK roles — behind every headline in the report.
Time saved by AI for different job roles
The twelve UK occupational families where AI frees the most hours per worker per week — gross capacity from task substitution plus capability amplification, from AWA’s 412-occupation model. This is the raw capacity the net office-space model then nets against new-role creation.
Source: AWA AI Impact Report 2026, task-level model of 412 UK SOC 2020 occupations, May 2026 capability calibration, checked against the AWA AI Job Market Tracker (week ending 21 June 2026). Combined value is substitution hours times the capability amplification multiplier for each occupational family; these are gross capacity figures, before the net office-space model.
Save or share this chart
| Occupation group | Hours per worker per week |
|---|---|
| Science, research & engineering | 35.3 |
| Business, media & public service | 26.2 |
| Science, engineering & technology associates | 23.6 |
| Health professionals | 22.1 |
| Business & public service associates | 21.6 |
| Teaching & educational professionals | 21.6 |
| Corporate managers & directors | 21.1 |
| Culture, media & sports occupations | 20.4 |
| Other managers & proprietors | 18.5 |
| Health & social care associates | 17.0 |
| Customer service occupations | 16.6 |
| Secretarial & related occupations | 16.1 |
Get the full analysis.
The full report — methodology, the 412-occupation breakdown, the net office-space model (gross versus net), the capability gap and the hiring margin, sector implications, and the framework for turning these numbers into decisions for your own organisation.
Download the report
Free PDF · no spam · your data is handled per AWA’s privacy policy.
Backed by science and fiercely independent.
AWA is a leading global independent consultancy transforming the world of work. Since 1992, our multi-disciplinary team has used science, research and our depth of experience to change the world of work for the better — evolving how people, places and technology work together, at pace. We call it the DNA of work.
The 412-occupation model behind this report is the next step in that work, built so leaders can act from evidence. Every parameter is adjustable; every figure can be re-calibrated for your own organisation. We help organisations act on it across four services: Workplace Strategy, AI Transformation, Cognitive Wellbeing & Performance and Change Management.
Visit advanced-workplace.com
