Future of work 2026 → 2035: an interactive timeline of three scenarios

What the timeline shows

The panel reads the same file that drives the Future Demand Radar: twelve productivity markets, three scenarios, and for each a demand trajectory at 2028, 2030, 2032 and 2035 expressed as a multiplier of the 2026 baseline. The multiplier is the median (p50) of 4,000 Monte Carlo draws; the shaded ribbon is the p10–p90 band. Years between the modelled points are interpolated on a log scale and marked as such. The cards are sorted by p50 at the selected year, their colour is the portfolio class the radar assigns in that scenario, and the badge summarises confidence from the band width relative to the median — high under 0.4, medium under 0.65, low above.

Treat every figure as model output. The method, weights, risk penalties and the calibration of the classes are documented on the methodology page and in the radar itself; nothing here is a measurement of the future.

The three scenarios

The scenarios differ only in their growth assumptions; purchase intent, monetisation, differentiation and durability are shared. A — Conservative assumes AI adoption stays gradual and copilots remain features inside existing suites, with attention and burnout demand tracking a hybrid-work plateau. The plateau is visible in official statistics: in 2024 about 22.6 % of employed people in the EU worked from home usually or sometimes, and 33 % of employed Americans did some work at home on days worked, roughly unchanged from the year before A^[6][7].

B — Baseline, the planning case, assumes copilots and delegated agents become routine in knowledge work between 2028 and 2030. Its anchors are survey-based: 82 % of leaders told Microsoft's 2025 Work Trend Index they expect to use digital labour to expand capacity within 12–18 months A^[2], while McKinsey's 2025 State of AI found 23 % of organisations scaling an agentic system and a further 39 % experimenting B^[3]. The EU AI Act's literacy duties, applicable since February 2025, are assumed to sustain skills demand through the decade A^[5].

C — Accelerated takes Gartner's upper bound: a third of enterprise applications embedding agentic AI by 2028 and routine knowledge work increasingly absorbed by agents, so that individual productivity becomes judgement, orchestration and recovery C^[4]. The same source predicts that over 40 % of agentic projects will be cancelled by 2027 — which is why the agents category carries the widest band in every scenario.

How to read the bands and the confidence badge

Bands widen with time because the model adds noise that grows with the square root of the horizon and because expert-judged components receive larger noise when their evidence grade is weaker. In practice almost every category is "high confidence" in 2028, "medium" around 2030–2032 and "low" by 2035. That is a statement about the model's own uncertainty, not about the world; a badge of low confidence means the p10 and p90 are far apart, and for a few categories the p10 is below 1.0 — demand could shrink.

The 2026 column is the index base: every multiplier equals 1.0 and the ranking shown for "what people will need" is borrowed from 2028.

What the model does not know

The forecast uses US search-demand data only, so non-English markets are inferred rather than measured. Labour-market context comes from employer surveys: the World Economic Forum expects 170 million roles created and 92 million displaced by 2030 and 39 % of current skill sets transformed A^[1]; PwC reports a 56 % wage premium for workers with AI skills in 2024 A^[8]. Task-level exposure studies suggest around 80 % of the US workforce could see at least a tenth of their tasks affected by large language models B^[9], and earlier occupation-level estimates of automation susceptibility have been contested precisely because they skipped the task level B^[10]. None of these sources forecasts demand for productivity content or tools; the model uses them to set the direction and plausibility of its expert multipliers, which remain judgement calls. Comparing predicted with observed multipliers is planned but not automated yet; when we do it, changes go through review and nothing is republished automatically.

How to use it

Switch scenarios at a fixed year to see which categories are robust (ranked near the top in A, B and C) and which depend on acceleration. Then scrub the year in your preferred scenario to see when the ranking changes. The "what people will need" list is a reading of each category's build plan and first assumption, ranked by the median multiplier; use it as a prompt for discussion, and go to the radar for scores, risk penalties and alternative weightings before deciding anything.

Frequently asked questions

Are the multipliers predictions of market size?

No. They are relative demand indices from a Monte Carlo model: a p50 of 2.0 in 2032 means the model’s median draw has search-led demand for that category at twice its 2026 level. They are not revenue, users or market size, and they are not measured.

Why do the bands get wider every year?

The model adds log-normal noise that grows with the square root of the horizon, and the expert components receive more noise when their evidence grade is weaker. Wider bands are the model being honest about what it does not know.

What does the confidence badge mean?

It is the band width divided by the median: high below 0.4, medium below 0.65, low above. It reflects the model’s internal uncertainty only; a high badge in 2028 does not mean the real world is predictable.

Why is 2029 or 2033 marked as interpolated?

The forecast is computed for 2028, 2030, 2032 and 2035. Values for the years in between are interpolated on a log scale so the scrubber moves smoothly; they add no information beyond the modelled points.

Where do the evidence ids come from?

Each id (for example AW-05) points to a record in our graded evidence file of 90 sources. The side panel resolves them to their source URLs so you can check the claim yourself.

Sources and evidence

1. Future of Jobs Report 2025 – press release — World Economic Forum (2025), employer survey report (evidence LS-01, LS-02, LS-03) B reports.weforum.org/docs/WEF_Future_of_Jobs_2025_Press_Release_FR.pdf (retrieved 2026-11-09)
2. 2025 Work Trend Index Annual Report: The year the Frontier Firm is born — Microsoft WorkLab (2025), survey report (evidence AW-05) B www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-fro (retrieved 2026-11-09)
3. The state of AI in 2025: Agents, innovation, and transformation (as reported) — McKinsey & Company via ppc.land (2025), survey report, secondary (evidence AW-02) B ppc.land/most-companies-still-pilot-ai-programs-despite-widespread-ado (retrieved 2026-11-09)
4. Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027 — Gartner Newsroom (2025), analyst press release (evidence AW-06, AW-07) C www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts (retrieved 2026-11-09)
5. AI Act | Shaping Europe’s digital future — European Commission (2026), official policy page (evidence RG-01) A digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai (retrieved 2026-11-09)
6. Employed persons working from home by professional status (lfsa_ehomp) — Eurostat (2024), official statistics (evidence LS-09) A ec.europa.eu/eurostat/databrowser/view/lfsa_ehomp/default/table?lang=e (retrieved 2026-11-09)
7. American Time Use Survey – 2024 Results (USDL-25-1060) — US Bureau of Labor Statistics (2025), official statistics (evidence LS-10) A www.bls.gov/news.release/archives/atus_06262025.htm (retrieved 2026-11-09)
8. 2025 Global AI Jobs Barometer – press release — PwC (2025), report (evidence AW-10) B www.pwc.com/gx/en/news-room/press-releases/2025/ai-linked-to-a-fourfol (retrieved 2026-11-09)
9. GPTs are GPTs: Labor market impact potential of LLMs — Eloundou T, Manning S, Mishkin P, Rock D (2024), peer-reviewed study (Science) A doi.org/10.1126/science.adj0998 (retrieved 2026-11-09)
10. The future of employment: How susceptible are jobs to computerisation? — Frey CB, Osborne MA (2017), peer-reviewed study (Technological Forecasting and Social Change) B doi.org/10.1016/j.techfore.2016.08.019 (retrieved 2026-11-09)