── ── Mental model
Principal–Agent Problem
One party (the principal) delegates to another (the agent) whose interests differ and whose actions can't be fully observed — producing agency cost: monitoring spend + agent bonding spend + residual loss. Formalized by Jensen & Meckling (1976). Structure produces the behavior, not character — so the fix is structural.
Run Principal–Agent Problem on a real problem
Bring something you're actually deciding — free, in the browser.
How it works
Step 1 — Identify structure Principal / Agent / What principal wants / What agent would do absent intervention / What principal cannot observe.
Step 2 — Diagnose misalignment 1-3 dominant types: effort · risk · time horizon · info asymmetry · adverse selection · moral hazard · multitasking · hidden self-dealing.
Step 3 — Estimate agency cost Monitoring cost + bonding cost + residual loss = total. Order-of-magnitude is enough.
When to use it
- someone asks why an employee, executive, contractor, board member, or fund manager isn't acting in the org's interest
- a compensation or incentive structure is being designed
- outsourcing or partnership terms are being negotiated
- someone says 'agency cost,' 'moral hazard,' 'skin in the game,' or 'incentive misalignment
When not to use it
When the decision is routine and reversible, applying a formal method costs more than it returns.
Worked example
Delegating to an Autonomous AI Agent, 2024–2026
The 2024–2026 wave of "agentic AI" — LLM-based systems given tools, memory, and the authority to take multi-step actions (write and merge code, send emails, move money, file tickets, operate a browser) — is a textbook principal–agent relationship wearing new clothes. You (the principal) delegate a task to a software agent whose objective function, information set, and effective incentives may all diverge from yours, and whose intermediate reasoning you cannot fully observe. Jensen and Meckling's 1976 definition applies almost verbatim: a relationship in which one…
Install this skill (free, MIT)
npx skills add deciqAI/knowledge-skillsUseful? Star the repo — stars help other builders find it.
FAQ
What is the principal–agent problem in simple terms?
Whenever one party (the principal) delegates work to another (the agent) whose interests differ and whose actions can't be fully observed, the agent will sometimes act in their own interest at the principal's expense. The gap is structural — it comes from incentives and information, not bad character.
Where does the principal–agent problem show up in startups?
Everywhere delegation exists: salespeople optimizing commission over customer fit, agencies billing hours over outcomes, executives building empires over shareholder value, and investors pushing risk profiles founders don't share. Each is an incentive-and-observability gap, and each has a structural fix.
How do you reduce agency costs?
Three levers: align incentives so the agent wins when the principal wins (equity, outcome-based pay), improve observability (metrics, reporting, audits), and accept some residual loss — perfect alignment costs more than it saves. The fix is always structural; exhorting agents to behave better is not a mechanism.
Related mental models
Whatever the other party does, each player is individually better off defecting — so both defect, and both end up worse than mutual cooperation.
Regression to the mean is the statistical regularity that any noisy measurement producing an extreme value tends to be followed on retest by a less-extreme value…
Most organizations operate well below their potential resource access.
势 (shì) is the propensity of a situation to evolve in a direction — a wave you read and ride, not power you exert.