Agent Governance Scorecard

Scope of Applicability

This scorecard applies to collaborative agentic systems where humans and AI agents share responsibility for value-laden decisions—including recommendations, prioritization, risk assessment, and action selection.

It is domain-agnostic: applicable to software delivery, enterprise operations, regulated environments, and mission-oriented decision support (e.g., C2-style workflows).

The standard is evidence-first: observable behavior and enforceable controls—not intent or roadmaps.

Roadmap items, planned features, aspirational designs, future commitments, or policy statements do not constitute evidence. If a criterion cannot be met without architectural redesign, it must be marked "No".

Why This Exists

AI agents are moving from experimental tools to production systems. But most platforms lack the architectural foundations for enterprise governance:

No central oversight — Agents operate without accountability
No audit trails — "The AI did it" is not an acceptable answer
No permission boundaries — Agents inherit full user privileges
No human intervention points — Autonomous systems with no guardrails

The result: AI agents that can't be audited, can't be controlled, and can't be trusted in regulated environments.

This scorecard provides objective criteria for evaluating any AI agent platform's governance readiness.

The Six Dimensions

1 Control Towers

"Organizations must establish control towers for AI — treating agents as organizational resources that need management and accountability."

Control towers provide centralized authority over agent operations — including the authority to intervene, constrain, or halt agent execution. Visibility alone is insufficient; control towers must have operational power.

Criterion	Description
Central orchestration authority	Single point of coordination for agent activities with power to direct, constrain, or halt
Agent registry and accountability	All agents registered, identified, and trackable to responsible parties
Dependency-aware execution	System understands agent interdependencies and can manage cascading effects
Real-time execution oversight	Live visibility into what agents are doing with ability to intervene

2 Decision Integrity

"Complete visibility into agent actions and decisions."

When AI agents make or influence decisions — including recommendations, prioritizations, and risk assessments — the reasoning must be preserved and traceable. The key question: "Why was this recommendation made, and who accepted it?"

Criterion	Description
Decision reasoning preserved	Why the agent made each decision is recorded (not just what)
Reasoning survives agent handoffs	Context transfers when work moves between agents; lineage is not lost
Alternatives explicitly recorded	What options were considered, not just what was chosen
Confidence explicitly represented	Agents express uncertainty; humans can calibrate trust accordingly

3 Observability

"Traceability over blind trust."

Every agent action must be visible, recorded, and auditable. Human and AI actions should flow through the same audit infrastructure—no separate systems, no gaps in lineage.

Criterion	Description
Complete action coverage	Every agent action is logged
Unified human + agent audit trail	Same audit system for human and AI actions
Tamper-evident records	Audit logs cannot be modified without detection

4 Governance Enforcement

"Governance isn't bureaucracy. Governance is scaffolding."

Governance must be enforced at runtime, not merely documented in policy. Controls must be architectural — agents cannot bypass them regardless of prompt engineering, configuration changes, or emergent behavior.

Criterion	Description
Runtime governance enforcement	Rules enforced during execution, not just at design time or deployment
Non-bypassable controls	Agents cannot circumvent governance mechanisms through any means (architectural, not policy-based)
Pre-execution blocking	Unauthorized actions prevented before they occur, not logged after the fact

5 Human-in-the-Loop (Calibrated Trust)

"Calibrated trust means knowing when to trust AI and when to intervene."

Humans must be able to intervene at the right moments—before harm occurs, not after. This requires systems designed for calibrated trust, not blind automation or post-hoc review.

Criterion	Description
Confidence-based escalation	Low-confidence decisions automatically escalate
Pre-harm intervention	Humans can intervene before damage occurs
Blocking human approval	Critical actions require explicit human authorization

6 System Evolution & Drift

"Derived from principles for operating agentic systems safely at scale."

Agent behavior changes over time — through retraining, prompt updates, model swaps, or emergent drift. In governed systems, evolution must be auditable, changes must be attributable, and rollback must be operationally real (not theoretical).

Criterion	Description
Scoped learning boundaries	Agent learning is bounded and controlled; cannot self-modify beyond defined limits
Auditable behavioral change	Changes in agent behavior are logged with attribution (who, what, when, why)
Reversible evolution / rollback	Agent behavior can be reverted to previous states within operational timeframes
Drift detection over time	System actively detects when agent behavior deviates from baseline

Download the Scorecard

Use these tools to evaluate any AI agent platform — including ours.

Governance Scorecard

The complete evaluation framework with all 6 dimensions and 20 criteria

Download PDF

Assessment Worksheet

Self-assessment template with evidence fields for each criterion

Download PDF

GitHub Repository

Full documentation, contributing guidelines, and community assessments

View on GitHub

How to Use

For Evaluating Vendors

Download the worksheet
For each criterion, assess: Yes, Partial, or No
Require evidence — demos, architecture docs, or live system access
Remember: Roadmaps don't count. Only current capabilities.

For Self-Assessment

Be honest — the scorecard is only useful if accurate
Document your evidence in the worksheet
Identify gaps and prioritize architectural improvements
Reassess periodically as your platform evolves

For RFPs and Procurement

Include scorecard criteria in your evaluation matrix. Use these criteria to ensure both AI actions and human approvals are attributable within the same audit trail.

Sample RFP language is available in the GitHub repository.

About This Scorecard

The Agent Governance Scorecard is developed and maintained by Equilateral AI, a governed AI agent orchestration platform.

We built this scorecard because we needed objective criteria to evaluate our own platform — and found nothing comprehensive existed. We're releasing it publicly because the industry needs shared standards for AI agent governance.

This is not a marketing document. Use it to evaluate us. Use it to evaluate our competitors. Use it to evaluate your own internal platforms. The goal is better governance across the industry.

Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0).

Is Your AI Agent Platform Governable?