Terms  /  Human-Agent Relationship  /  Agency Erosion
07 · Human-Agent Relationship

Agency Erosion

Gradual loss of human decision-making authority as agents take over more workflow without checkpoints.

Agency Erosion happens slowly. First, the agent makes low-stakes decisions: variable names, formatting, file organization. Then it starts making medium-stakes decisions: library choices, API design patterns, error handling strategies. Eventually, it's making high-stakes decisions: architecture, security configurations, deployment strategies — and the human has become accustomed to accepting agent decisions without review. The human's role has shifted from decision-maker to approver, and eventually to bystander. The erosion is gradual enough that neither party notices until the human realizes they're no longer driving.

Example
A team starts using an AI agent for code reviews. Initially, developers carefully evaluate every suggestion. After three months of mostly-good suggestions, they start clicking "accept all" without reading. After six months, the agent is effectively making all code quality decisions for the team. When a junior developer asks "why do we handle errors this way?", nobody on the team can explain it — the agent decided, and nobody questioned it. The team has lost agency over their own coding standards.

More in Human-Agent Relationship