The Autonomous Execution Framework classifies hiring interventions into three execution tiers based on reversibility and impact: Tier 1 (fully autonomous) covers low-risk, reversible actions — adjusting outreach timing, logging an alert, updating a Health Score. Tier 2 (one-click confirmation) covers medium-risk actions — pausing an outreach sequence, sending a recruiter check-in, scheduling a brief review. Tier 3 (full manual review) covers high-impact, hard-to-reverse actions — recruiter reassignment, mandate parking, client notification, escalation to sponsor. The framework prevents autonomous systems from taking consequential actions without human awareness while enabling full automation of the high-frequency, low-risk interventions that drain recruiter time.
The Autonomy Boundary Problem
Autonomous execution creates value by removing the latency between signal and action. But not all actions should be autonomous. Recruiter reassignment, mandate parking, and client notification are actions with relationship and reputational consequences that require human judgment. The Autonomous Execution Framework draws a precise boundary between what the system should do automatically, what it should do with one-click confirmation, and what it should never do without a deliberate human decision.
"Autonomous execution is not unlimited AI control. It is a precise definition of which actions are safe to automate and which require human judgment. The framework's value is the boundary — not the automation."
Execution Tier Classification
| Tier | Execution Mode | Example Actions | Reversibility | Human Involvement |
|---|---|---|---|---|
| Tier 1 | Fully Autonomous — executes without notification | Adjust outreach timing; recalculate Health Score; log SLO breach; update pipeline velocity metric | Fully reversible; no external impact | None required — logged for review |
| Tier 2 | One-Click Confirmation — system recommends, human confirms in one action | Pause outreach sequence; send recruiter check-in task; flag mandate for weekly review; adjust stage SLO | Easily reversible; minimal external impact | Single confirmation click; executed within 60 seconds |
| Tier 3 | Full Manual Review — system surfaces, human decides and acts | Recruiter reassignment; mandate parking; client notification; escalation to sponsor; new briefing session | Partially irreversible; relationship or reputational impact | Full decision process; documented with reason |
Frequently Asked Questions
Can the autonomy tier be configured by the organisation?
Yes. Some organisations are comfortable with Tier 2 actions being fully autonomous; others want Tier 1 actions reviewed. The Autonomous Execution Framework defines the default tier for each action type, but every boundary is configurable. High-trust, high-maturity organisations typically push more actions to Tier 1. Organisations new to the system typically start with all Tier 2 actions requiring confirmation.
What prevents the system from taking a wrong autonomous action?
Three safeguards: (1) every Tier 1 action is logged with timestamp, trigger signal, and action detail — fully auditable. (2) Every Tier 1 action is reversible — if the system pauses an outreach sequence incorrectly, it can be resumed in one action. (3) The failure prediction threshold for autonomous action is conservative — the system only executes Tier 1 automatically when the signal is unambiguous, not when it is marginal.
What is an example of a Tier 2 action going wrong and how is it corrected?
Example: the system recommends pausing outreach to a candidate pool at a company where the mandate is healthy, due to a false-positive overload signal. The recruiter sees the recommendation, recognises the error (the signal was from a legacy data point), and declines the confirmation. The system logs the declined recommendation and adjusts its model. This is the design intent of Tier 2: human judgment corrects system errors before they have external impact.