Coase, Williamson, and 88 Years of Theory That Explain the AI Bill
Two economists explained why firms exist and how they organize. Their theories remain fully relevant in 2026, now applied to networks where humans and AI agents operate side-by-side.
90-Second Summary
In 1937, a young British economist named Ronald Coase explained why firms exist. His economic answer: because coordinating economic activity inside an organized structure costs less than coordinating every single transaction on the open spot market. In 1985, American economist Oliver Williamson extended the argument, demonstrating that markets and hierarchies represent the two ends of a complex governance spectrum with various hybrid structures in between. Both received the Nobel Prize in Economics. In 2026, the typical enterprise has AI agents operating inside its hierarchy alongside humans. This represents a fourth economic governance structure that neither predicted, yet their theoretical frameworks fully accommodate it. The foundational economics have been established for 88 years; the continuous financial measurement of this new hybrid structure is the critical operational work that remains to be done.
There is a recurring equation troubling corporate boards in 2026. The public narrative around AI promises individual performance increases of 30% to 50%. Enterprises respond by deploying AI tools across departments, putting an agent on every desk, and integrating generative platforms into every major workflow. Individual task metrics improve and individual processes run faster. Yet, the consolidated operating margin fails to improve. In several cases, it has declined.
Where are the gains? Why is this performance not manifesting in the operating margins presented by the CFO to the board? The explanation is not new. It was written in 1937 in a short essay that changed the economics of organization forever, and extended in 1985 in a book that became one of the most cited in social sciences. It applies directly to the hybrid workforce of 2026 with critical but non-revolutionary adjustments.
1937: The Question Nobody Was Asking
Ronald Coase was 26 years old when he published "The Nature of the Firm" in 1937. His essay raised a question that seemed naive to the economists of his era: if the market is highly efficient, and if prices organize resource allocation better than any central planner, why do firms exist? Why isn't every transaction negotiated transaction-by-transaction, spot-by-spot, on the open market?
Coase's answer was elegant: coordinating transactions via the open market is expensive. Finding a reliable vendor, negotiating a price, drafting a contract, monitoring execution, and managing disputes when deliverables fall short all consume time, attention, and capital. Coase termed this collective overhead transaction costs.
When coordinating an activity via the open market costs more than coordinating it within an organized structure (a firm), the activity is brought in-house. When it costs less via the market, it is outsourced. A firm grows until the cost of coordinating the next transaction internally equals the cost of coordinating it externally. This represents the economic boundary of the firm.
Coase was awarded the Nobel Prize in Economics in 1991, 54 years after publishing his essay. The delay was not due to a lack of importance, but because his 1937 argument required decades to be fully understood by the academic and corporate worlds. When it was, it became the foundation of modern institutional economics.
1985: Williamson Expands the Foundation
Oliver Williamson built upon Coase's theory by asking a practical question: transaction costs exist, but how much do they cost, under what operational conditions, and how should an enterprise choose between different organizational structures? In 1985, he published "The Economic Institutions of Capitalism", which became one of the most influential texts in modern social sciences.
Williamson's contribution contains three key insights that apply directly to 2026. First, markets and hierarchies are not binary choices; they represent the ends of a spectrum with hybrid structures in the middle. Relational long-term agreements, joint ventures, stable vendor networks, and franchises are all intermediate points between "pure market transactions" and "pure hierarchical firms."
Second, three variables determine which organizational structure is optimal: asset specificity (the level of investment dedicated to a specific relationship), environmental uncertainty (the level of volatility in the market context), and transaction frequency (how often the transaction repeats). When asset specificity and frequency are high, hierarchies win. When both are low, markets win. In the middle, hybrid structures are optimal.
Third, the primary focus of analysis must be the governance structure—not the product or service itself. How you govern the economic relationship between the parties is what determines transaction costs. Williamson called this approach New Institutional Economics and was awarded the Nobel Prize in Economics in 2009.
| Governance Structure | Optimal Conditions | Dominant Cost Driver | Classical Corporate Example |
|---|---|---|---|
| Market | Low asset specificity, low transaction frequency, low uncertainty | Search, negotiation, and spot contracting costs | Commodity procurement, standardized hardware acquisition |
| Hybrid (Relational) | Medium asset specificity, medium frequency, medium uncertainty | Monitoring compliance + risk of vendor opportunism | Long-term supply agreements, franchise networks, joint ventures |
| Hierarchy (Firm) | High asset specificity, high frequency, high uncertainty | Internal management overhead + loss of market incentives | In-house core operations, product development, corporate finance |
88 Years: The Framework that Sustained Every Technological Wave
The first wave of practical corporate application for the Coase-Williamson framework was the outsourcing movement of the 1990s. Enterprises systematically analyzed which business activities had low asset specificity and could be migrated to the market. Standard IT support, basic bookkeeping, and operational HR were outsourced; core proprietary capabilities were kept in-house.
The second wave was the offshoring trend of the 2000s. The same transaction cost economics were applied to global geography. Where workflows could be executed remotely without sacrificing asset specificity, they migrated to lower-cost regions. The economic logic remained unchanged; only the geographical boundaries of the firm shifted.
The third wave was the platformization of the 2010s. Networks like Airbnb, Uber, and corporate marketplaces proved that algorithmically-driven hybrid networks could outperform traditional hierarchies in specific domains. Transactions became micro, dynamic, and automated. Williamson's framework predicted this intermediate structure; advanced web technology made it economically viable in scale.
Over 88 years, this economic foundation has sustained every major wave of technological disruption. Each shift redefined the optimal boundaries of the enterprise, but the underlying transaction cost logic remained absolute.
2026: The Fourth Structure
Today, we are witnessing a new organizational phenomenon: AI agents operating within the firm with economic behaviors that do not fit into any of Williamson's three traditional categories. They are not external vendors (they are not market transactions). They are not salaried personnel with employment agreements (they lack career motives, individual incentives, or opportunistic agency). They are not relational partners (they do not negotiate terms).
They represent a fourth organizational structure: agents operating within the hierarchy. They exhibit very high asset specificity (trained on proprietary data and custom prompts), zero interest-based opportunism (they do not deviate for personal gain), yet high output uncertainty (due to hallucinations, model drift, and rare edge-case failures). Crucially, their transaction frequency is potentially infinite (with near-zero marginal cost per inference).
Neither Coase nor Williamson lived to see this structure, but their theories accommodate it. The key question for 2026 is an update of 1937: now that we have a fourth economic structure operating within the firm, where do the optimal boundaries of the enterprise lie, and what is the cash cost of governing this hybrid coordination?
| Governance Structure | Operational Composition | Dominant Cost Driver | Financial Measurement Method |
|---|---|---|---|
| Market | Spot transactions with external vendors | Vendor search, negotiation, and compliance auditing | Direct invoice costs + procurement overhead |
| Relational Hybrid | Long-term agreements with specialized third parties | SLA monitoring, performance reviews, and contract risks | Contract management payroll + recurrent audit costs |
| Classical Hierarchy | Salaried human personnel in organized structures | Intra-firm coordination (meetings, alignments, escalations) | Direct payroll + human calendar time allocated to meetings |
| Agentic Hierarchy | LLMs, agentic workflows, and automated workflows in production | Model inference costs + human remediation and review | Token spend + payroll spent on calibration and validation (A2H) |
The fourth structure introduces a new layer of coordination economics. It does not replace the other three; it intersects them. In a typical 500 FTE growth-stage enterprise, critical business decisions now traverse all four structures simultaneously, which is why your corporate coordination costs have become so difficult to trace.
The Boundaries of the Firm Shift as Agentic Coordination Costs Decline
This fourth structure carries a profound operational implication. As the unit cost of coordinating via agents falls, activities that historically required classical human hierarchies are migrating to agentic structures. Simultaneously, activities that were outsourced to the market are returning in-house, now operated by proprietary agents.
The optimal boundary of the firm is being redrawn in real time. Companies that can measure the exact cost of coordinating across all four structures will secure a durable structural advantage. The enterprises that fail to measure will discover within a few years that their competitors have engineered a fundamentally lower coordination cost structure, enabling superior operating margins.
Economic theory explains why these boundaries shift. Continuous financial measurement of your coordination edges is the operational step required to manage them. The 4 interfaces of human-agent coordination in cash terms outlines how to implement this analysis.
The Measurement Challenge
Quantifying transaction costs has always been a challenge in institutional economics; Coase noted in 1937 that transaction costs are difficult to isolate directly. Prior to AI, this limitation was acceptable because the most expensive structure—the human hierarchy—could be reasonably monitored through salaries and formal organizational structures.
In the era of AI, the agentic hierarchy introduces a rapidly growing share of operational costs that bypass traditional payroll tracking. Time spent on prompt calibration is distributed across brief intervals during daily tasks. The human alignment overhead required to handle agent outputs lacks a clear line item in your ERP. The engineering hours spent remediating agents that fail to communicate are recorded as standard software development payroll.
This is the invisible overhead that highlights why human-agent coordination costs are the critical missing link in AI governance. Coase's theory explains why these costs are strategic; Williamson's taxonomy provides the framework to separate them into distinct, measurable structures.
Regulatory Compliance vs. Economic Governance
It is vital to distinguish economic coordination governance from regulatory compliance. The EU AI Act and national data privacy frameworks regulate rights, algorithmic safety, and model transparency. These are necessary, enforceable, and have clear legal deadlines. However, they do not measure coordination economics.
These are parallel layers. An enterprise can achieve 100% regulatory compliance while continuing to lose millions in unmonitored coordination friction. What the EU AI Act means for your operations represents one challenge; governing the cost of coordinating across four economic structures is another. Addressing the former without the latter leaves the enterprise legally protected but operationally inefficient.
Applying the Theory: Four Practical Movements
Implementing these theoretical concepts does not require purchasing a complex software suite. Four practical steps are sufficient to bring Coase and Williamson into your next QBR:
- Map Your Four Governance Structures. Identify which structure executes each of your five most critical business workflows: Market (external SaaS vendors), Relational Hybrid (specialized third-party contractors), Classical Hierarchy (internal human teams), or Agentic Hierarchy (generative workflows and automated agents).
- Estimate Coordination Costs. Calculate the coordination cost for each structure using loaded payroll rates and token invoices. An initial estimate with a 15-25% margin of error is sufficient to drive meaningful executive discussions.
- Evaluate Your Firm Boundaries. Analyze if your current boundaries are economically optimal: Are there activities in your classical hierarchy that are candidates for agentic transition? Are there outsourced activities that should return in-house via agents?
- Review Quarterly. Integrate the distribution of your coordination costs by structure into your standard forecasting, alongside ARR and gross margins. Boundaries shift quarterly; the companies that measure them are the ones that manage them.
Frequently Asked Questions
Why are Coase's theories relevant to AI governance in 2026?
Because he answered the fundamental question that has returned to the executive agenda: why do firms exist instead of everyone contracting individually on the open market? Coase demonstrated that coordinating transactions on the open market has real friction costs. Because AI radically alters these coordination costs, it shifts the optimal boundaries of the firm.
What did Williamson add to Coase's foundation?
Williamson proved that markets and hierarchies are not binary choices; they are the ends of a spectrum containing relational contracts, joint ventures, and stable networks. He showed that the optimal structure is determined by asset specificity, environmental uncertainty, and transaction frequency. In 2026, this taxonomy gains a fourth structure: agents operating within the hierarchy.
Why does their theory support a fourth structure today?
Because corporate AI agents are not external market vendors, nor are they human employees under employment contracts. They operate inside the firm with characteristics of both: they exhibit high asset specificity (trained on proprietary data) and zero opportunism, yet high output uncertainty (hallucinations, drift) and near-infinite transaction frequency. They require a distinct, fourth economic classification.
Is this the same as transaction cost economics applied to outsourcing?
No. Traditional outsourcing uses transaction economics to make simple make-or-buy decisions. The application in 2026 is different: corporate agents operate internally but exhibit economic behaviors that mix internal and external characteristics. The challenge is not choosing between making or buying, but governing a hybrid structure that did not exist when traditional outsourcing frameworks were designed.
How does an enterprise begin applying this economic framework?
By analyzing five critical workflows from the last quarter: What share of the total execution time was spent on coordination versus actual execution? Was the coordination human-to-human, or did it involve AI agents? This analysis establishes the baseline for your current firm boundaries, enabling you to systematically evaluate where migration to agentic structures makes financial sense.
The Bottom Line
Ronald Coase was 26 years old in 1937 when he wrote the essay that still explains why firms exist. Oliver Williamson was 53 in 1985 when he consolidated the framework that showed how to classify their economic governance. Both received the Nobel Prize in different decades. Their economics remain entirely sound; they simply require the addition of a fourth structure to accommodate the reality of 2026.
A robust theory does not lose value when the world changes; it gains new relevance. In 2026, the new variable is the AI agent operating within the firm alongside human teams. The coordination tax of this hybrid workforce is being paid today, distributed across small intervals in every business decision. The enterprise that measures this tax captures the real margins promised by AI; the company that does not will discover at a future board meeting that their operating margins failed to match the individual productivity gains they expected.
Coase and Williamson explained the economics. The continuous financial measurement of this fourth structure is the operational work that remains. It is up to you to decide whether you will begin measuring now, or wait for the next regulatory or competitive wave to demand an answer.