# mollick-organizational-theory-agentic-ai-spans-control-2026-02

## Veille

Mollick: Organizational Theory for Agentic AI - spans of control and boundary objects

## Titre Article

Organizational Theory Lessons for Agentic AI

## Date

2026-02

## URL

https://www.linkedin.com/posts/emollick_i-think-agentic-ai-would-work-much-better-activity-7426069089074290688-PG6i

## Keywords

Ethan Mollick, organizational theory, agentic AI, spans of control, boundary objects, coupling, bounded rationality, agent swarms, middle management agents, multi-agent coordination, complex hierarchies, org design

## Authors

Ethan Mollick

## Ton

**Profile**: Applied academic reflection, analytical and provocative register, high conceptual level

**Description**: Ethan Mollick adopts his characteristic tone as a Wharton professor applying established theoretical frameworks to emerging problems. The style is direct, with a touch of self-aware humor ("insert middle management joke here"). The argument proceeds through structured analogy between human organizations and AI agent systems. The post implicitly critiques labs that ignore decades of organizational research. Target audience: AI researchers, agentic system architects, tech managers familiar with organizational theory.

## Pense-betes

- **Central thesis**: Agentic AI would work better if it drew on organizational theory (complex hierarchies, informational limits, spans of control)
- **Current critique**: Agentic systems assume models have unlimited capacity to manage sub-agents - which is clearly false
- **Spans of control for AI**:
- A human tops out at fewer than 10 direct reports
- 100 sub-agents = too many for an orchestrator agent
- **Proposed solution**: "Middle management" agents (with the humor about the term acknowledged)
- **Boundary objects**: Objects passed between groups to convey meaning when a project crosses boundaries (prototype, user story)
- Currently: agents pass plain text + maybe code
- Need: Structured boundary objects that agents of different levels can read/write
- Benefit: Would resolve many coordination failures + reduce token usage
- **Coupling**: Degree of connection between units in an organization
- Too tight: every step requires approval
- Too loose: loss of control
- This tradeoff is well studied in organizational theory
- **Bounded rationality**: Organizational concept that likely also applies to agents
- **Agent swarms**: A "terribly named" term according to Mollick
- The problem will not just be model quality
- It will be **organizational design** choices
- **Critique of the labs**: "I am not sure the labs see this"
- **Call to action**: Need for agent organization experiments led by people who understand real coordination problems

## RésuméDe400mots

Ethan Mollick, a professor at Wharton and an influential AI observer, proposes applying lessons from organizational theory to agentic AI systems. His argument: decades of research on human organizations offer frameworks directly applicable to the challenges of multi-agent coordination.

**The spans of control problem**: Current agentic systems implicitly assume that models can manage an unlimited number of sub-agents, which is clearly false. A human manager tops out at fewer than ten effective direct reports. Mollick estimates that a hundred sub-agents far exceeds the capacity of an orchestrator agent. His provocative solution: create "middle management agents" - an intermediate hierarchy between the main orchestrator and the execution agents.

**Boundary objects**: In organizational theory, boundary objects are artifacts passed between groups (marketing, IT, sales) to convey meaning when a project crosses boundaries - prototypes, user stories, specifications. Currently, AI agents exchange plain text and sometimes code. Mollick advocates for structured boundary objects that agents of different capability levels can read and modify. This approach would resolve many coordination failures while reducing token consumption.

**The coupling problem**: Coupling measures the degree of connection between organizational units. Most agentic systems are either too tightly coupled (every step requires human approval) or too loosely coupled (loss of control and coherence). This tradeoff is well studied in organizational theory, and Mollick bets that many findings apply directly to agent architectures.

**Bounded rationality**: A foundational concept in organizational science, bounded rationality describes how decision-makers operate with incomplete information and finite cognitive capacity. This framework likely applies to AI agents facing large contexts and complex decisions.

**Critique of the labs**: Mollick observes that everyone is rushing toward "agent swarms" (a term he calls "terribly named") without realizing that success will not depend solely on model quality, but on organizational design choices. He doubts that AI labs perceive this dimension and calls for more experimentation led by people who understand real human coordination problems.

## GrapheDeConnaissance

- Ethan Mollick —recommande→ théorie organisationnelle (CONCEPT, 0.98)
- Ethan Mollick —s_oppose_à→ systèmes agentiques actuels (TECHNOLOGIE, 0.95)
- Ethan Mollick —affirme_que→ les labs d'IA ne perçoivent pas la dimension organisationnelle des agents (AFFIRMATION, 0.9)
- spans of control —s_applique_à→ agents orchestrateurs (TECHNOLOGIE, 0.97)
- Ethan Mollick —affirme_que→ un agent orchestrateur ne peut pas gérer un nombre illimité de sous-agents (AFFIRMATION, 0.96)
- boundary objects —améliore→ coordination multi-agents (CONCEPT, 0.94)
- boundary objects —réduit→ consommation de tokens (CONCEPT, 0.88)
- couplage organisationnel —s_applique_à→ architectures d'agents IA (TECHNOLOGIE, 0.92)
- rationalité limitée —s_applique_à→ agents IA (TECHNOLOGIE, 0.85)
- middle management agents —résout→ dépassement du span of control (CONCEPT, 0.9)
- agent swarms —utilise→ design organisationnel (METHODOLOGIE, 0.93)
- Ethan Mollick —travaille_chez→ Wharton (ORGANISATION, 0.97)

---
Canonical: https://www.thekb.eu/en/fiches/mollick-organizational-theory-agentic-ai-spans-control-2026-02/