The future of knowledge work, 2026 to 2035: three scenarios and their signposts

Where we are: adoption without impact, 2026

Every forecast on this page starts from one gap. Stanford's 2026 AI Index reports that 88% of organisations used AI in 2025 and 70% used generative AI in at least one business function, with global private AI investment at US$344.7 billion, up 127.5% in a year.^[1] A McKinsey's November 2025 survey, which we could only read through secondary coverage, gives the same 88% adoption figure, yet finds only 39% of organisations attributing any EBIT impact to AI and about 6% reporting an impact of 5% or more.^[2] B PwC's survey of 4,701 CEOs tells the same story from the top: 56% report efficiency gains in employees' time, 32% saw revenue rise and 34% saw profitability rise, against 46% who had expected it.^[3] A

Individual use has moved faster than organisational returns. Slack's Workforce Index, a vendor survey of 5,156 desk workers, recorded AI use rising from 36% to 60% between November 2024 and April 2025.^[4] A Pew finds 34% of US adults have ever used ChatGPT, 58% of those under 30.^[5] A US labour productivity, meanwhile, grew at a 4.9% annualised rate in the third quarter of 2025 and 1.9% over the year, a strong quarter that cannot yet be attributed to AI with any confidence.^[6] A

The controlled evidence explains why adoption and impact diverge. Generative AI raises output on bounded tasks, most for the least experienced (about 14–15% more issues resolved per hour across 5,172 support agents, over 30% for novices, near zero for veterans),^[7] A but lowers accuracy on tasks just outside the model's competence,^[8] B and human–AI combinations on average under-perform the better of the two alone on decision tasks.^[9] A Returns depend on which tasks are handed over and how they are checked, which is a management variable, not a technology one.

Search interest mirrors the vocabulary shift. Our keyword data shows "future of work" at about 1,000 US searches a month, shrinking 30% a year over three years, while "ai skills" grows 100% a year to 3,600 (DataForSEO, retrieved 2026-11-09). People have stopped asking whether work changes and started asking what to learn.

How to read the three scenarios

The scenarios below come from our Future Demand Radar, a model that scores twelve productivity categories from keyword histories, Google Trends shares, graded evidence and explicit expert assumptions, then runs 4,000 Monte Carlo draws per category to produce a median score with a p10–p90 band. The scenarios differ only in growth assumptions. Two categories anchor this page: AI agents and delegated work and future-of-work skills. The numbers are multipliers of today's US search demand, not jobs or GDP, and the radar's own rule applies: forecasts are model outputs with stated uncertainty, not facts. The interactive 2035 timeline shows the same trajectories year by year.

Scenario	Agents: score (class)	Agents: demand 2028 / 2030 / 2035	Skills: score (class)	Skills: demand 2028 / 2030 / 2035
A, conservative	56.0 (build now)	1.3× / 1.8× / 2.1×	42.5 (monitor)	1.2× / 1.4× / 1.5×
B, baseline	59.7 (build now), band 55.2–64.1	1.7× / 3.0× / 4.4×	45.8 (monitor), band 41.4–50.2	1.5× / 2.2× / 2.8×
C, accelerated	61.7 (build now)	2.2× / 5.0× / 8.8×	48.2 (validate in 12 months)	1.7× / 3.0× / 4.4×

Scenario A: conservative

Copilots stay features inside the suites people already pay for. Agents grow but remain a fraction of classic productivity demand; most delegated work is drafting and summarising, and the cancellation wave that Gartner, an analyst firm, predicts for over 40% of agentic projects by end-2027 arrives roughly as described (an analyst estimate we grade C).^[10] C Attention and burnout demand track the hybrid-work plateau. Under this scenario our model expects agent-related search demand at about 2.1 times today's level by 2035 and skills demand at about 1.5 times. Knowledge work in 2035 looks like 2026 with better autocomplete.

Scenario B: baseline (our planning case)

Copilots and delegated agents become routine in knowledge work between 2028 and 2030, consistent with the 82% of leaders who told Microsoft they expect to deploy digital labour within 12 to 18 months^[11] A and the 62% of organisations at least experimenting with agents.^[2] New demand forms around orchestrating and verifying agents and feeding them personal knowledge, which is why our agent-management playbook treats delegation as a management skill. The EU AI Act's literacy duty, in force since February 2025, sustains skills demand on the compliance side.^[12] A Our baseline scenario expects agent demand at 1.7 times today in 2028, 3 times in 2030 and 4.4 times in 2035 (band 3.1–6.2), and skills demand at 2.8 times by 2035 (band 2.0–4.0). The model assigns a trough risk to 2027–28 if the cancellation wave is deep.

Scenario C: accelerated

Capable agents absorb much routine knowledge work, closer to Gartner's upper-bound claim that 33% of enterprise applications embed agents and 15% of day-to-day work decisions are made autonomously by 2028 (analyst estimate, grade C).^[10] Individual productivity becomes judgement, orchestration and recovery; software-comparison demand shifts to agent platforms; attention becomes scarcer. Our model expects agent demand near 8.8 times today's level by 2035 under this scenario. We do not plan on it, because the organisational-impact data above has not moved in its direction yet.

Jobs: created, displaced, transformed

The two most cited sources measure different things, and conflating them is the commonest error in this genre.

The World Economic Forum's Future of Jobs 2025, an employer survey of more than 1,000 firms extrapolated to the global labour market, projects 170 million jobs created and 92 million displaced by 2030, a net gain of 78 million and churn equal to 22% of today's jobs.^[13] A It also expects 39% of workers' current skills to be transformed or obsolete by 2030^[14] B and estimates that 59 of every 100 workers need reskilling or upskilling by 2030, 11 of whom are unlikely to receive it.^[13] These are employer expectations, not observed counts.

The International Labour Organization measures exposure, not displacement. Its 2025 refined index places about one in four workers worldwide in an occupation with some generative-AI exposure, 34% of employment in high-income countries against 11% in low-income ones; only 3.3% of global employment falls into the highest-exposure tier, and women's employment is roughly twice as exposed as men's (4.7% versus 2.4%).^[15] A Exposure means tasks could be affected, not that jobs disappear; the ILO's own framing is augmentation for most and transformation for few.

Academic work brackets the macro effect. Eloundou and colleagues estimate that around 80% of the US workforce could see at least 10% of their tasks affected by large language models and about 19% could see half or more.^[16] A Acemoglu's modelling of those exposure numbers yields modest aggregate gains, a total-factor-productivity increase he puts at no more than about 0.66% over ten years.^[17] B Autor argues the opposite distributional case: that AI could extend expert-level tasks to more workers and rebuild middle-skill work, if it is used that way.^[18] B Both are working papers; the point is that the range of serious estimates runs from "small and slow" to "large and distributional", not from "nothing" to "the end of work".

Remote and hybrid: the plateau data

Remote work neither collapsed nor kept growing. Eurostat's labour-force survey shows 22.6% of employed people aged 15–64 in the EU usually or sometimes working from home in 2024 (8.9% usually, 13.7% sometimes), with Germany at 23.8%, France at 33.9% and Spain at 15.4%.^[19] A In the United States, 33% of employed people did some work at home on days they worked in 2024, against 35% in 2023; the split by education is stark, 50% of those with a bachelor's degree or higher versus 18% with a high-school diploma.^[20] A For knowledge work the plateau is high and stable, which means the attention and coordination problems of hybrid teams are a durable feature of the next decade rather than a pandemic artefact.

Signposts to watch each quarter

A scenario is only useful if you can tell which one you are in. These are the indicators we check every quarter, with the reading that would move us.

Signpost	Source	Favours A if	Favours C if
Share of organisations scaling agents	McKinsey State of AI (currently 23%)	Flat or down for two surveys	Above 35%
Share reporting any EBIT impact from AI	McKinsey (currently 39%)	Stays under 45%	Above 55%
Agentic project cancellations	Gartner follow-ups, vendor churn	Over 40% realised	Well under 25%
ILO highest-exposure tier	ILO index updates (currently 3.3%)	Unchanged	Rising with observed displacement
US and EU home-working shares	ATUS, Eurostat (33%, 22.6%)	Falling	Rising
"ai agents" and "ai skills" search volume vs our predicted multipliers	Our quarterly keyword re-pull	Below the p10 path	Above the p90 path
AI Act high-risk dates	European Commission (2 Dec 2027)	Delayed again	Held, with enforcement

What to do now, by career stage

Early career (0–5 years). The controlled studies show the largest gains for novices on bounded tasks.^[7] Use that to produce more, but keep a share of each task unassisted so you build the judgement that later lets you check the agent. Learn to write briefs and acceptance tests; see AI skills for 2030.

Mid-career (5–20 years). Your advantage is knowing which side of the frontier a task sits on. Convert that into a delegation practice: brief, test, log. Build the coordination and trust skills covered in human skills in an AI economy, which the labour-market data rewards and which exposure indices barely touch.

Senior and managerial. The gap between 88% adoption and 39% impact is yours to close. The levers are task selection, verification and accountability, not model choice. Budget time for checking as a real cost, and treat the 2027 AI Act dates as a planning horizon for documentation.

Prediction log

We publish our predictions so they can be marked. The current entry, from the 2026-11-09 run of radar method 1.1 (scores updated 2026-11-10 to method 1.2; the demand multipliers below did not change), predicts 2028 US search-demand multipliers under scenario B of 1.73 for AI agents, 1.48 for future-of-work skills and 1.58 for personal AI operating systems. The observed values stay empty until we re-measure demand; that comparison is done by hand for now, and any change to the model goes through review, never an automatic publication. When a prediction misses, this page says so in this section rather than quietly updating the table above.

What to do next

Open the Future Demand Radar to see all twelve categories under the three scenarios, including the bands this page summarises, and the 2035 timeline for the year-by-year path. If you want the system-level response rather than the forecast, the OS Builder turns thirteen answers about your work into a plan with an AI-delegation section and a four-week checklist. The quarterly signpost readings and any prediction misses go out in the newsletter below.

Frequently asked questions

Is the future of work about AI replacing jobs or changing them?

The measured evidence points to change for many and displacement for few. The ILO places 25% of global employment in occupations with some generative-AI exposure but only 3.3% in the highest tier, and the WEF's employer survey projects more jobs created than displaced by 2030 (170 million versus 92 million) alongside 39% of skills transformed.^{[13] [15]} A Exposure estimates describe tasks that could be affected, not jobs that end.

Which scenario do you expect?

We plan on scenario B, the baseline: delegated agents routine in knowledge work by 2028–2030, with agent-related search demand about 3 times today's level by 2030 and 4.4 times by 2035 (band 3.1–6.2). It is a model output with stated uncertainty, and the quarterly signposts above are how we find out whether it is wrong.

Has remote work declined?

It has plateaued. Eurostat records 22.6% of EU employees usually or sometimes working from home in 2024, and the US American Time Use Survey records 33% of employed people doing some work at home on days worked, against 35% a year earlier.^{[19] [20]} A The share is far higher for degree-holders (50%) than for those with a high-school diploma (18%).

Why do the forecast numbers measure search demand rather than jobs?

Because search demand is something we can measure monthly, predict and be scored against, and because the radar's purpose is to decide what guidance, tools and products to build. Jobs and productivity forecasts are cited here from the WEF, ILO, Stanford and academic sources with their grades, not produced by us.

How often is this page updated?

Quarterly, when we re-pull the keyword data, read the new McKinsey, ILO and statistical releases, fill in the prediction log and re-check the signposts. The updated date at the top of the page reflects the last pass.

Sources and evidence

1. Artificial Intelligence Index Report 2026, Chapter 4: Economy — Stanford HAI (2026), report A hai.stanford.edu/assets/files/ai_index_report_2026_chapter_4_economy.p (retrieved 2026-11-09)
2. The state of AI in 2025: Agents, innovation, and transformation (as reported by ppc.land and idm.net.au) — McKinsey & Company (QuantumBlack) (2025), report B ppc.land/most-companies-still-pilot-ai-programs-despite-widespread-ado (retrieved 2026-11-09)
3. 28th Annual Global CEO Survey (press release) — PwC (2025), report B www.pwc.com/gx/en/news-room/press-releases/2025/pwc-2025-global-ceo-su (retrieved 2026-11-09)
4. Workforce Index spring 2025: The New AI Advantage — Slack Workforce Lab (2025), report A slack.com/blog/news/the-new-ai-advantage (retrieved 2026-11-09)
5. 34% of U.S. adults have used ChatGPT, about double the share in 2023 — Pew Research Center (2025), report A www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-c (retrieved 2026-11-09)
6. Productivity and Costs, Third Quarter 2025, Preliminary — US Bureau of Labor Statistics (2026), dataset A www.bls.gov/news.release/archives/prod2_01082026.htm (retrieved 2026-11-09)
7. Generative AI at Work — Brynjolfsson, Erik; Li, Danielle; Raymond, Lindsey (2025), observational A doi.org/10.1093/qje/qjae044 (retrieved 2026-11-09)
8. Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of Artificial Intelligence on Knowledge Worker Productivity and Quality — Dell'Acqua, Fabrizio; McFowland III, Edward; Mollick, Ethan R.; Lifshitz-Assaf, Hila; Kellogg, Katherine; Rajendran, Saran; Krayer, Lisa; Candelon, François (2023), RCT B doi.org/10.2139/ssrn.4573321 (retrieved 2026-11-09)
9. When combinations of humans and AI are useful: A systematic review and meta-analysis — Vaccaro, Michelle; Almaatouq, Abdullah; Malone, Thomas (2024), meta-analysis A doi.org/10.1038/s41562-024-02024-1 (retrieved 2026-11-09)
10. Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027 (press release; analyst estimate, headline verified via search only) — Gartner (2025), report C www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts (retrieved 2026-11-09)
11. 2025 Work Trend Index Annual Report: The year the Frontier Firm is born — Microsoft WorkLab (2025), report B www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-fro (retrieved 2026-11-09)
12. AI Act | Shaping Europe's digital future (Regulation (EU) 2024/1689, application timeline incl. 2026 Omnibus amendment) — European Commission (2026), report A digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai (retrieved 2026-11-09)
13. Future of Jobs Report 2025 – press release (French edition) — World Economic Forum (2025), report B reports.weforum.org/docs/WEF_Future_of_Jobs_2025_Press_Release_FR.pdf (retrieved 2026-11-09)
14. Jobs market at a crossroads: which are the fastest growing and declining jobs? (summarising WEF Future of Jobs 2025) — Euronews (2025), report B www.euronews.com/business/2025/02/01/jobs-market-at-a-crossroads-which (retrieved 2026-11-09)
15. Generative AI and Jobs: A Refined Global Index of Occupational Exposure (ILO Working Paper 140) — International Labour Organization (2025), report A www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-o (retrieved 2026-11-09)
16. GPTs are GPTs: Labor market impact potential of LLMs — Eloundou, Tyna; Manning, Sam; Mishkin, Pamela; Rock, Daniel (2024), observational A doi.org/10.1126/science.adj0998 (retrieved 2026-11-09)
17. The Simple Macroeconomics of AI (NBER Working Paper 32487) — Acemoglu, Daron (2024), report B doi.org/10.3386/w32487 (retrieved 2026-11-09)
18. Applying AI to Rebuild Middle Class Jobs (NBER Working Paper 32140) — Autor, David (2024), report B doi.org/10.3386/w32140 (retrieved 2026-11-09)
19. Employed persons working from home by professional status (lfsa_ehomp) — Eurostat (2024), dataset A ec.europa.eu/eurostat/databrowser/view/lfsa_ehomp/default/table?lang=e (retrieved 2026-11-09)
20. American Time Use Survey – 2024 Results (USDL-25-1060) — US Bureau of Labor Statistics (2025), dataset A www.bls.gov/news.release/archives/atus_06262025.htm (retrieved 2026-11-09)