Framework
Tiered Autonomy
A framework for calibrating how much decision-making authority to grant AI agents at each layer of an enterprise operation.
The problem
Enterprise teams deploying AI agents face a recurring question: how much should the system decide on its own? Grant too little autonomy and you have an expensive autocomplete. Grant too much and you have an audit liability. Neither extreme compounds.
The framework
Tiered Autonomy maps agent decision authority across three tiers — Tier 1 (recommend, human decides), Tier 2 (decide, human reviews), Tier 3 (decide and act, human audits) — and prescribes which tier is appropriate based on reversibility, stakes, and the maturity of the model in context.
When to use it
Use Tiered Autonomy when scoping an agentic deployment for the first time, when a deployed agent is exhibiting unexpected behaviour, or when preparing an AI governance brief for a board or audit committee. The framework surfaces the implicit autonomy decisions that most teams make informally — and makes them legible to non-technical stakeholders.
What success looks like
A team that has applied Tiered Autonomy can answer three questions without ambiguity: What decisions does the agent make on its own today? What escalation path exists when it encounters a boundary condition? How does the organisation move a class of decisions from Tier 2 to Tier 3 as confidence grows?