Start your day with intelligence. Get The OODA Daily Pulse.
Informing your decisions with actionable intelligence
Start your day with intelligence. Get The OODA Daily Pulse.
Informing your decisions with actionable intelligence
Artificial intelligence is changing more than who performs the work. Increasingly, autonomous systems are redistributing decision authority, organizational power, and opportunities for economic value. Leaders must assess those changes alongside employment displacement.
For executives, the immediate challenge is to distinguish what AI can perform from what organizations can reliably automate – and what automation actually changes for workers. Task exposure, job losses, workforce restructuring, and declining human authority are related developments, but they require different evidence and different responses.
As machines perform more physical and cognitive work, who retains the ability to decide, govern, and benefit? We take up these questions in an afternoon session at OODAcon 2026 titled John Henry’s Last Stand: Understanding Human Labor Displacement This session will provide context of use as we all seek to frame what is in store for us. The presenter is long term friend of OODA and the OODA Network David Bray (CEO, LeadDoAdapt Ventures)
Some initial context on this topic: From task automation to autonomous enterprises, the emerging contest concerns who performs work, who exercises authority, and who captures the value machines create.
John Henry’s contest with the steam drill endures because it makes technological displacement personal. A worker confronts a machine that threatens the economic value of his skill. Strength, pride, and endurance become the terms of competition. The tragedy is that winning the immediate contest cannot settle who controls the next version of the workplace.
The contemporary contest is spreading across cognitive work and, through robotics, returning to physical work. Software can assist with analysis, documentation, coding, and coordination. Agentic architectures connect those activities into processes that can execute with less continuous human direction. The question becomes how much of an occupation, an organization, or an economic relationship must change when a machine can perform a growing share of its constituent work.
The OODA Loop research assembled for this project traces that changing question across 2025 and 2026. It begins with human talent and enterprise task exposure, expands into early-adoption industries and the infrastructure supporting AI diffusion, and moves toward autonomous agency, organizational redesign, and the preservation of human decision-making authority.
Our central finding is that displacement is a process with several distinct outcomes:
Understanding those differences is the starting point for strategic action.
OODA Loop Research and OODA Network discussions reflect a progression from task automation to enterprise transformation, infrastructure constraints, autonomous governance, and human agency. The research points toward five connected developments: the decomposition of occupations into tasks; the integration of those tasks into increasingly autonomous workflows; the redesign of enterprises around those workflows; physical and financial constraints that shape the pace of adoption; and a struggle over the distribution of authority and economic returns.
What evolved is a framework for examining labor displacement without assuming that increasing machine capability produces a uniform or inevitable employment outcome:
The research began with tasks. Enterprise Task Audits to Identify “LLM-Ready” Functions Across Business Units and What Enterprise IT Executives and Entrepreneurs Need to Know About LLM Displacement Now establish the practical starting point: identify activities that large language models (LLMs) can support or perform, then evaluate their implications for workflows, staffing, and competitive advantage. Automating selected activities does not, by itself, establish that an entire occupation will disappear.
The next stage concerned organizational redesign. The Rise of AI-Native Companies, The May 2026 OODA Network Meeting: The Enterprise AI Inflection Point and the Collapse of Legacy Operating Assumptions, and Designing and Orchestrating an Agentic AI Operating System: Emergent Behaviors and Enterprise Value Creation shift attention toward how firms coordinate work. If autonomous agents execute connected workflows, organizations may change supervisory roles, management structures, staffing requirements, and the relationship between workforce size and output.
Early-adoption sectors provide comparative evidence. OODA Loop research connected The Healthcare AI Acceleration Story, The Cybersecurity AI Acceleration Story, and The Financial Sector AI Acceleration Story through Future Scenarios: Lessons Learned from AI Early Adoption Industry Sectors. These sectors offer settings for examining how automation interacts with professional judgment, regulation, security, and institutional accountability.
Capability does not automatically create productivity. The Hidden Cost of AI Acceleration: The AI Productivity J-Curve and Why Capability Arrives Before Results highlights the complementary investments required to turn technical advances into operational gains. The external METR developer productivity study reinforces the need for measurement: experienced developers working in familiar repositories took longer when using the early-2025 AI tools tested.
Physical constraints complicate displacement narratives. Constraints to Growth Across Exponential Technologies: Electric Power, Skilled Trade Labor, and Strategic Infrastructure Mobilization at Scale and The Atoms Meets Algorithms Crisis: Why AI Infrastructure Can Be a Big Blind Spot introduce electricity, construction, industrial capacity, and skilled labor into the analysis. The insight? Cognitive automation could coexist with persistent shortages in the workers and infrastructure needed to deploy it.
Autonomy makes authority a central labor question. From Model Risk to Action Risk: The Governance Gap That Will Define Enterprise AI Winners and The August 2026 OODA Network Meeting: Autonomous Agents, Escalating Capability, and Unsettled Authority connect machine execution with permissions, accountability, and delegated decision-making. A worker may remain employed while losing discretion over how work is performed; another may gain authority by directing and governing automated systems.
Human agency provides the strategic endpoint. Why and How Human Agency and Decentralized Approaches to AI Are Transforming Problem-Solving at the Edge and The Power of the Edge: Decentralized Interoperability as the Foundation for Resilience offer approaches for preserving local judgment and operational adaptability. OODA analysis treats judgment, trust, creativity, coordination, and accountability as capabilities to examine and strengthen, without assuming they are permanently immune to automation.
The resulting research thesis is that labor displacement is also a question of institutional design and ownership: how organizations deploy autonomous systems, who controls their actions, and how the resulting economic gains are distributed.
The future is therefore conditional.
Capabilities matter, but deployment economics, institutional choices, demand growth, worker participation, and the creation of new tasks help determine whether capability produces broad augmentation, concentrated gains, or sustained displacement.
The strongest implication is that human labor displacement cannot be understood through job-loss projections alone. We need to track employment, compensation, work quality, learning opportunities, organizational authority, ownership, and dependence on external platforms.
These measures can move in different directions.
These scenarios are planning tools, not forecasts or assigned probabilities. They can coexist across sectors and regions.
Broad augmentation. AI raises the capacity of existing workers, and growing demand absorbs much of the gain. Employment remains resilient as tasks and skills change. Indicators include improved service quality, wage growth, internal mobility, and continued investment in junior talent.
Polarized employment. Productivity grows while benefits concentrate among owners and scarce specialists. Routine cognitive work and some entry-level pathways contract. Indicators include weaker hiring, widening compensation differences, and declining progression despite improving enterprise output.
Accelerated displacement. Reliable workflow autonomy combines with strong cost pressure and weak demand growth. Organizations reduce labor faster than new work, and institutions can absorb displaced people. Indicators include sustained reductions in employment and labor hours following verified deployments, persistent reemployment difficulty, and wage pressure across exposed roles.
Distributed enterprise expansion. Accessible AI lowers the cost of creating specialized businesses and providing services. More people gain productive capacity, although income may remain uneven. Indicators include viable small-enterprise formation, wider market participation, and manageable dependence on dominant platforms.
Tracking these futures requires separating:
For enterprise leaders, premature workforce reductions can remove the knowledge needed to make automation reliable. Delayed adaptation can leave organizations operating with cost structures and processes that competitors have already redesigned. Both risks require evidence about actual work rather than assumptions derived from model demonstrations.
For workers, the exposure is broader than redundancy. It includes wage pressure, reduced hiring, intensified performance expectations, and the erosion of professional development.
For policymakers and investors, the issue connects technological competitiveness to economic participation: increasing productive capacity does not by itself determine who can access that capacity or share in its returns.
This chronology follows OODA Loop Research’s broad arcs and the evolution of OODA Network meeting discussions throughout 2025 and 2026:
The research begins with people. OODA Loop Research: From Human Risk to Talent Superpower Strategy and The Human Factor Remains the Future of Cybersecurity establish the OODA Loop research agenda’s initial orientation: human behavior and expertise remain consequential even as technical systems become more capable.
Applied to labor displacement, that framing changes the unit of analysis. An employee is more than a collection of outputs that a machine might reproduce. Workers also carry operational context, relationships, tacit knowledge, and the ability to recognize when a formally correct answer will fail in practice. Those contributions must be identified before an organization can assess what it is actually replacing.
The August 2025 research on Enterprise IT Convergence and Nancy Morgan’s AI4 2025 Read-in broadens the issue into work culture and organizational futures. Automation enters an existing institution with established responsibilities and expectations. Its effects depend on how that institution reorganizes around it.
Enterprise Task Audits to Identify LLM-Ready Functions Across Business Units Paired with What Enterprise IT Executives and Entrepreneurs Need to Know About LLM Displacement Now supplies the first practical Enterprise IT discipline: assess activities before making claims about entire occupations.
A task audit should examine inputs, outputs, exception rates, verification requirements, dependencies, and consequences of failure. A draft, a recommendation, and an authorized transaction are different kinds of work. Producing each faster does not imply that the surrounding job can safely disappear.
Consider an illustrative analyst workflow. AI may accelerate document review and preliminary synthesis. The analyst still needs to establish whether sources are credible, reconcile conflicting evidence, understand the decision-maker’s priorities, and accept responsibility for a recommendation. Whether employment declines depends on how much time the full process saves, whether demand for analysis expands, and how management reallocates capacity.
The distinction also exposes a subtler transition risk. Organizations can automate the routine work through which junior employees previously acquired expertise. If they retain senior judgment but eliminate the path for developing it, they consume a capability they are no longer replenishing. Protecting apprenticeship becomes a design requirement for sustainable automation.
The late-2025 research shifts attention from individual firms to the systems enabling technological adoption. The New Strategic Economy: Geoeconomics and the Architecture of Influence and From Cloud to Country connect finance, infrastructure, compute access, and national competitiveness.
For labor, the implication is uneven diffusion. The ability to automate depends on access to capital, usable data, infrastructure, and implementation expertise. Organizations with similar workforce profiles can consequently follow very different trajectories. National exposure to AI does not establish that firms in that country can deploy it economically at scale.
Defense industrial base and strategic investment analysis add a further dimension. Decisions about which industries receive capital influence where new work emerges, what skills are valued, and which regions can participate in technological expansion. Labor outcomes become part of industrial strategy.
Energy Acceleration: When Power Becomes the New Strategic Bottleneck makes the physical foundation explicit. AI deployment relies on electricity, facilities, equipment, and people who can build and maintain them. A digital system that substitutes for some office tasks can simultaneously create demand for work far from the office.
This coexistence is central to the displacement story. Aggregate labor shortages and occupational displacement can occur together. The person losing a clerical role may lack the training, location, resources, or inclination needed to enter electrical construction. New demand is an opportunity, but it is not an automatic transition mechanism.
The project’s Healthcare AI Acceleration Story, Cybersecurity AI Acceleration Story, and Financial Sector AI Acceleration Story provide sector lenses for the labor question. Their underlying early-adoption experiences reach back into 2021–2025, even as this synthesis brings them into the 2026 research progression.
Healthcare foregrounds the relationship between automation, operational fragility, and consequential decisions. Reducing documentation or administrative effort can free capacity, but the resulting workforce changes depend on care demand, implementation quality, and how responsibilities are redistributed. Time saved is a resource that an institution must decide how to use.
Cybersecurity supplies an adversarial complication. Automation can accelerate defensive work while adversaries also automate their activity. A useful planning hypothesis is that some productivity gains will be absorbed by a rising volume and complexity of threats. Faster execution can therefore coexist with continued demand for experienced practitioners.
Finance brings governance and economic coordination into view. The public introduction to the financial-sector template describes AI integration across fraud detection, compliance, markets, and customer interaction. The labor implication is that interconnected automation changes who monitors activity, handles exceptions, and remains accountable when a system acts.
Future Scenarios: Lessons Learned from AI Early Adoption Industry Sectors ties these sector experiences together. For this project, the lesson is that an automation decision changes the surrounding workflow and institution. A workforce assessment must include those effects, rather than stopping at the technical performance of an isolated tool.
The February 2026 OODA Network Meeting on Living in Beta provides an organizational response to uncertainty: experimentation and adaptation. Applied to workforce planning, this means building reversible deployments that generate evidence about quality, cost, and human work before making structural commitments.
The March research on The Great Decomposition and From Retrieval to Governance moves from impressive outputs toward operational architecture. Different workloads require different combinations of models, data, controls, and infrastructure. Their economics cannot be reduced to the cost of generating an answer.
The Hidden Cost of AI Acceleration: The AI Productivity J-Curve supplies the connecting concept. Integration, process redesign, training, review, and organizational learning can precede measurable returns. The cost of completed, acceptable work is the relevant comparison with human labor.
External evidence makes that caution concrete. METR’s July 2025 randomized study found that 16 experienced open-source developers took 19 percent longer on assigned tasks when AI tools were allowed. That was a bounded result for particular tools, developers, and repositories, not a general verdict on AI-assisted coding. In its February 2026 update, METR reported that selection and measurement problems made its follow-up estimates unreliable; it considered greater acceleration plausible but could not confidently quantify it. Together, the studies argue for measuring actual deployment conditions and updating conclusions as those conditions change.
The infrastructure sources introduce another check on extrapolation. Constraints to Growth Across Exponential Technologies and The Atoms Meets Algorithms Crisis frame electricity, skilled trades, and industrial execution as constraints. Cognitive automation does not dissolve the practical requirements of building the systems that support it.
Physical AI may eventually automate more of those activities. Yet robotics demonstrations should be distinguished from dependable operation across variable workplaces. Installation, maintenance, safety, integration, and liability all affect adoption. The physical wave deserves its own evidence base rather than a timeline borrowed from software.
The May 2026 OODA Network Meeting, The Rise of AI-Native Companies, and the June 2026 Exponential Organizations Discussion mark a shift in the research architecture. The enterprise itself becomes the object of redesign.
Retrofitting AI into existing jobs and building an organization around AI are different strategic choices. A new firm may organize around outcomes, shared data, and automated processes from its inception. An established enterprise must also resolve inherited dependencies, incentives, and responsibilities. Neither structure guarantees success, but the difference matters for labor demand.
Designing and Orchestrating an Agentic AI Operating System links that redesign to coordination, governance, observability, and emergent behavior. The displacement question now includes work that connects other work: routing requests, tracking progress, consolidating reports, and managing handoffs.
Some coordination functions may be automated; others may become more valuable as autonomous systems multiply. The outcome depends on whether the organization can define reliable execution boundaries and retain the expertise to handle exceptions. Reducing supervisory layers without that capacity can turn apparent efficiency into accumulated operational risk.
A plausible form of organizational displacement is reduced hiring rather than immediate layoffs. An enterprise could expand output with a relatively stable workforce while declining to replace departures or recruit junior staff. That possibility makes hiring, internal mobility, career progression, and work intensity essential indicators alongside headcount.
The governance research makes a consequential distinction in From Model Risk to Action Risk. A system that produces advice and a system permitted to execute an action create different responsibilities.
When an agent can change a record, commit resources, or trigger a workflow, labor transformation includes delegated authority. Who sets the objective? Who defines the permissions? Who can inspect the execution, challenge its assumptions, and stop it? The Agentic AI Survey and Trust Is the Throttle place those questions within the capability-and-control framework.
The August 2026 OODA Network Meeting brings unsettled authority into the foreground. The September Meeting extends this discussion into economics, governance, and offensive cyber.
For workers, a formal oversight role can conceal practical loss of agency. A reviewer may remain accountable while lacking enough time or information to scrutinize machine decisions. Meaningful human control requires a manageable review burden, access to relevant evidence, authority to intervene, and competence maintained through practice.
Why and How Human Agency and Decentralized Approaches to AI Are Transforming Problem-Solving at the Edge and The Power of the Edge introduce an alternative design emphasis: place useful capability and decision authority closer to operational problems.
The analytical implication is conditional. Local AI could help a small organization or specialist undertake work previously requiring a larger institution. That advantage depends on access, interoperability, support, and economic independence. Distributing interfaces while concentrating control over data and infrastructure may leave the underlying dependence intact.
The related Cognitive Infrastructure Policy research extends the question beyond employment. The platforms through which people acquire knowledge and exercise judgment can influence their ability to contest decisions. Human agency therefore includes the capacity to understand and challenge the systems shaping work.
About the Author
Daniel Pereira is the Director of Research at OODA. He is a strategic foresight practitioner, research program strategist, and operations-focused leader with 20+ years of experience managing complex, multidisciplinary research initiatives across academia, industry, and government-adjacent environments.
Informing your decisions with actionable intelligence
Informing your decisions with actionable intelligence