Before reemployment · gross worker transitions
—GLOBAL HUMAN LABOR IMPACT LAB / NOV 2026
The labor transition.
How fast is too fast?
Three AI impact scenarios. Ten sectors. Displacement, reemployment and the humanoid wave. Every assumption exposed, every curve reproducible.
01FRONTIER RESEARCH
EXPERIMENTAL FORECAST / V1
EXPERIMENTAL FORECAST / V1
OBSERVED / REMOTE LABOR INDEX41.3%End-to-end project success in a selected freelance benchmark—not 41.3% of jobs. Source ↗
OBSERVED / OCCUPATIONAL EXPOSURE1 in 4Workers globally in occupations with some GenAI exposure. Exposure is not displacement. ILO ↗
OUTLOOK / ROBOTS & AUTONOMOUS SYSTEMS58%of surveyed employers expect robotics and autonomous systems to transform their business by 2030. An expectation, not deployments or job losses. WEF ↗
Before reemployment · gross worker transitions
—Before reemployment · gross worker transitions
—WHICH WORK GOES FIRST?
The exposure sequence.
Earliest threshold crossing in the medium case: modeled forced transitions reach 1.5% of sector employment. Not a prediction of when an industry disappears.
DOES WORK RETURN?
Reentry has a lag.
—potentially re-employed workers in the medium path, with the modeled delay and reintegration ceiling.
Channel shares are Frontier hypotheses, not observed employment statistics. Re-employment can be delayed, lower paid, geographically inaccessible, or unavailable to displaced workers.
THE PHYSICAL AI WAVE
Humanoids don't arrive at software speed.
We model hardware supply, the fraction of robots doing productive work, deployment lag and human-job equivalents separately. Traditional industrial robots and humanoids do not share a single demand curve. Neither is assumed to replace one worker per device.
MODELED HUMANOID UNITS ADDED—
HUMANOID-RELATED ROLE EQUIVALENTS—
ALL ROBOTICS / SHARE OF REDUCTIONS—
ADJUST THE WORLD / THE MODEL BELOW
No hidden assumptions.
Changes recalculate every scenario, sector, return pathway, and export. Blue values are editable hypotheses—not fitted empirical coefficients.
HOW THE MACHINE THINKS / METHODOLOGY
Evidence in. Assumptions out in the open.
The 41.3% RLI project success rate anchors a bounded logistic capability curve. Recent benchmark doubling rates cannot be directly translated into unemployment, so the pace of capability improvement is an independent editable assumption. Faster AI is not equivalent to faster organizational adoption.
Digital job-equivalent reductions are calculated from modeled task exposure, ability, adoption and role-conversion. Robotics reductions use incremental robot shipments, productive use, deployment lag and human-job-equivalence factors; the physical channel is capped by sector task requirements and excludes overlap.
Vacancy and attrition absorption reduce the number of role reductions that become forced labor transitions. Each new displaced cohort gets a separate reemployment curve with a waiting period, sector reintegration ceiling and ramp. The unabsorbed transition gap is not a forecast of measured global unemployment.
Regional wages, demand rebound, demographic change, policy and education adaptation, income shifts, macroeconomic feedback and the distributional impact within countries are not solved by this first version. The three paths are sensitivity narratives, not statistical confidence bands. Twenty-year projections are explicitly speculative.
Sources and default anchors
Download canonical evidence + assumptions ↗MODEL IN ONE LINE
Task potential × adoption × position conversion → role-equivalent reductions; attrition → transitions; cohort reemployment → unreabsorbed gap.RLI, ILO, IFR and employment data are observed inputs. Sector splits, conversion, growth and return rates are Frontier scenario assumptions.