Frameworks Library

    AI Governance · Evolved from Access Control & Permissioning

    Agent Autonomy & Boundaries

    Classification framework for agent autonomy tiers, tool and data scoping, escalation criteria, and kill-switch procedures — ensuring autonomous action is bounded, observable, and reversible.

    Maturity: AdvancedReversibility: LowAI Risk: HighOversight: High
    Steal the kit

    Executive Guidance

    How to hold this framework as a leader

    Autonomy is a product decision with governance, liability, and support consequences. A shipped agent that can take irreversible action on a customer's behalf is a different product than one that suggests an action for the customer to confirm. The board should be able to name the autonomy tier of every customer-facing agent.

    The essential artifacts are an autonomy tier definition, the scope of tools and data an agent may access, the events that require escalation, and a kill-switch procedure with pre-assigned authority. None of these are optional for autonomous action.

    When to Apply

    • Any agent capable of taking action beyond retrieval
    • Customer-facing or regulated workflows involving automation
    • Internal agents with access to sensitive data or systems

    AI-Era Notes

    Autonomy tier is a product decision, not a technical one. It carries governance, liability, and support implications that must be reviewed by legal and executive stakeholders.

    Key Trade-off

    Autonomy accelerates value but expands blast radius; kill-switch authority must be pre-assigned.

    Operating Sequence

    The order in which to install the framework

    1. 01Classify each agent capability by autonomy tier before scoping build.
    2. 02Enumerate the tools and datasets the agent may access, and the ones it may not.
    3. 03Define the events that require escalation to a human, and the response SLA.
    4. 04Assign kill-switch authority to a named role and document the runbook.
    5. 05Review the autonomy classification any time scope, tools, or data change.

    Key Artifacts

    The documents this framework produces

    Autonomy Tier Standard

    Definitions and preconditions for each autonomy tier.

    Owner: Product + Legal

    Tool & Data Scope Register

    The exact access surface for the agent.

    Owner: Product + Security

    Escalation & Kill-Switch Runbook

    Escalation criteria, response SLA, and shutdown authority.

    Owner: Operations + Product

    Operating Checklist

    What "good" looks like when installed

    Classification

    • Autonomy tier is explicitly assigned and reviewed by legal.
    • Reversibility of each agent action is classified.
    • Blast radius of a worst-case failure is documented.

    Boundaries

    • Tool and data access is enumerated and minimized.
    • Escalation events and SLAs are defined and rehearsed.
    • Kill-switch authority is named and reachable 24/7.

    Common Antipatterns

    • Undocumented tool access granted to agents
    • No defined threshold for human escalation
    • Missing kill-switch authority or runbook

    Boardroom Questions

    • What is the autonomy tier of every customer-facing agent we operate?
    • Who can shut down an agent, and how fast?
    • What is the blast radius of a worst-case autonomous action?

    Pairs With

    Model Evaluation FrameworkRisk & Compliance Checklist