Where could progress suddenly change speed?
EOL | Labor Analytics exists because technological change does not always follow yesterday’s trend. Breakthroughs can become historical discontinuities, accelerating what technologies can do, what they cost, and how they reshape work. Looking only at what today’s conditions have already produced risks missing the turning points that will matter several years from now.
The Scenario Lab watches leading indicators of those possible shifts: technical results, prices, real-world deployment, organizational redesign and scientific discovery. It also examines the technical, economic, institutional, and social-psychological constraints—from trust to perceived risk—that might limit or amplify a transition. These are conditional research assessments, not changes to EOL’s employment forecasts.
Each colored pathway tracks a possible acceleration mechanism. The six columns show where evidence of potential breakthroughs exists, where obstacles still limit them and what consequences remain conditional. Select a stage to examine the evidence and possible turning point.
Choose a pathway to see its possible turning point and where current evidence finds a constraint.
Swipe the full comparison sideways to see all six questions →
Bounded operating evidence: documented activity, not proof of broader adoption.
Pale or dashed continuation indicates downstream effects that have not yet been established by the pathway evidence. Width and color do not measure progress, probability or magnitude. Multiple constraints can bind at once.
One pathway may face several barriers at the same time. For example, agentic office work still faces workflow cost, organizational take-up and reliable end-to-end completion barriers. Cheaper model components could help without proving that the whole workflow is economically viable. These are October 17 interpretive judgments, not changes to the frozen October assessment or EOL forecasts.
Workflow costs and reliable delivery could make more applications investable.
The multiplication sign marks a potential reinforcing interaction: two evidence-supported developments might jointly improve a shared task or its economics. Their interaction is a hypothesis, not a measured multiplication of performance or economic returns.
Agent-enabled fleet scheduling, exception handling and dispatch could improve utilization and reduce supervisor time in material-handling settings.
Critical missing proof: audited, repeatable joint gains beyond separate component improvements.
Transferable manipulation could allow the same machine to perform more productive tasks. Higher utilization could improve payback and encourage larger fleets.
Critical missing proof: task-matched, multi-site nonlinear improvements in actual deployments.
Each record begins with the possible turning point, then shows what evidence supports it, what still restricts progress and what would change our assessment. Detailed source context remains available.
These are selected references supporting the five pathway narratives, not the complete 147-record core EOL evidence inventory. Source IDs and original qualifications are preserved. October-research references may draw from the frozen core or the Lab research crosswalk; a source listed here is not automatically an accepted core-EOL record. Supplemental company evidence and historical comparators are separately labeled and do not silently amend the frozen founding Lab snapshot.
The Lab tests whether advances in technology, unit economics, organizations or science could release constraints and cause progress to speed up nonlinearly. It distinguishes a genuine approaching threshold from publicity or routine incremental progress, and identifies evidence that could falsify a proposed acceleration. It does not score the probability of breakthroughs or change EOL employment forecasts.
Organizational adaptation matters: firms may need to redesign whole workflows, not simply buy tools. Competitive imitation can accelerate diffusion where reliable, quality-adjusted returns are proven, but security, law, integration and accountability can prevent a threshold from translating into adoption.
Reuse the occupation-oriented October 10 evidence stock with its original source IDs and qualifications. Avoid counting the same source again for every linkage.
Look for upstream technical advances, economics, organizational redesign, research automation and real-world constraints that could alter stage timing.
Test omissions, original public dates, claim directness, sources in common, counterevidence and possible alternative explanations.
A pathway can have more than one binding constraint across stages. Distinguish currently binding from secondary or potential next constraints, explain their evidence and what might release them; no numeric scoring.
Seek independent, comparable deployment results: completed-task costs, intervention and review hours, verified scientific outputs, multi-firm adoption and evidence of actual interactions. Record disconfirming results as well.
No Lab scenario, map mark or conjectured threshold automatically changes core EOL stage gauges or the 80 published employment-forecast endpoints.