Mandate

Layer: Agent (controllability assessment)
Scale: 0-100
Production Tier: Monitoring-Grade

Purpose

Mandate determines whether human oversight over autonomous behavior is effective in practice, not just documented in policy. It translates control effectiveness into a single operational score used for governance and risk decisions.

How It Works

1Capture oversight telemetry

→

2Aggregate control evidence

→

3Mandate evaluation

→

4Control policy decision

→

5Remediation tracking and re-evaluation

Emits

score (0-100)confidence (0-1)risk bandcontrol gaps

Scoring Dimensions

1. Override Effectiveness

Measures how reliably agent behavior follows approved human overrides.

2. Intervention Latency

Measures whether interventions occur fast enough to prevent irreversible harm.

3. Visibility Depth

Measures how much of the delegation chain is observable to operators.

4. Engagement Quality

Measures operator attentiveness and decision quality under real workload.

5. Escalation Reliability

Measures whether high-risk events are escalated accurately and consistently.

Public note: exact formulas, thresholds, and calibration constants are intentionally withheld.

Input Schema

Field	Type	Required	Description
agent_id	string	yes	Agent under oversight evaluation.
override_events	object[]	yes	Human override actions and outcomes.
latency_events	object[]	yes	Intervention timing evidence.
delegation_graph	object	yes	Delegation path and visibility metadata.
operator_activity	object[]	no	Signals for engagement and alert handling.
escalation_events	object[]	no	Escalation decisions and outcomes.
policy_window	object	yes	Risk-specific intervention windows.

Output Schema

Field	Type	Description
framework	string	mandate
version	string	Scoring specification version.
entity_id	string	Evaluated agent identifier.
score	number	Mandate score from 0 to 100.
risk_band	string	Oversight risk classification.
confidence	number	Confidence in score quality (0 to 1).
control_gaps	string[]	Oversight gaps requiring remediation.
recommended_action	string	Suggested response (monitor, review, restrict).

Score Interpretation

Mandate Tiering Bar

Oversight readiness by score range for control policy decisions.

80-100

Strong Oversight

Interpretation: Oversight is effective and responsive.

Typical action: Standard monitoring.

60-79

Managed Risk

Interpretation: Oversight is usable with moderate weaknesses.

Typical action: Improve controls and monitor closely.

40-59

Material Gaps

Interpretation: Oversight reliability is materially degraded.

Typical action: Restrict high-risk operations.

0-39

Insufficient Control

Interpretation: Oversight is ineffective for current risk profile.

Typical action: Suspend or block until controls improve.

Worked Example

Scenario: an enterprise evaluates three autonomous agents for oversight readiness.

Operational outcome:

1. Agent A remains in standard production monitoring.
2. Agent B enters a remediation plan with tighter alerts.
3. Agent C is restricted until intervention reliability improves.

Illustrative note: values and scores above are example outputs for documentation only.

Use Cases

Human-in-the-Loop Financial Controls

Validate that intervention windows remain effective in fraud, lending, and payment decision flows where delayed action creates outsized loss.

Healthcare and Clinical Workflow Oversight

Assess whether supervisors can observe and intervene in time-sensitive recommendations, especially when downstream outcomes affect patient safety.

Public Sector Operational Governance

Verify that delegated systems in service delivery and enforcement workflows remain controllable and auditable.

Autonomous Vendor Management

Benchmark external agent operators on practical control quality before expanding production scope.