Framework · AI Strategy

AI Decision Rights ArchitectureAIDRA

Who decides what when no human is watching.

A structured extension of standard decision rights models that adds autonomous actors to the authority map — across delegation scope, autonomy class, and escalation protocol. It does not replace your existing decision rights framework — a TDDA map, a RACI model, or an architecture governance charter. It extends it with a parallel dimension: the Agent Authority Map.

Open the Agent Authority Map Read the full article

The problem

An air traffic control tower governs every piloted aircraft in its airspace. Then a drone enters.

The drone never calls the tower. The tower has no authority over it — not because the drone is malicious, but because the airspace governance framework was designed for piloted aircraft.

Your decision rights architecture was designed for humans. Then an agent enters. It decides things — and no cell in your decision rights matrix has its name on it.

Three questions every enterprise must answer

01 · Delegation scopeWhich decisions can be delegated to an agent at all?
02 · Autonomy classWhat autonomy level does the agent hold for the decisions it is authorized to make?
03 · Escalation protocolWhen human intent and agent decision diverge, who resolves the conflict?

Dimension 1

Delegation Scope

Whether a decision class is eligible for AI delegation at all. It is not about capability — an agent may be technically able to make a non-delegable decision. The question is whether the organisation has authorised it to.

Delegable: The decision class can be authorized for agent execution under defined constraints.; API routing, log analysis, schema selection within approved boundaries, standard incident triage
Conditional: The decision class can be delegated only when specific safeguards are in place.; Purchase order approval (with threshold), customer response (with review), code generation (with human gate)
Non-Delegable: The decision class must always be resolved by a human actor.; Architecture principle changes, vendor selection, regulatory attestation, budget allocation, team structure

Dimension 2

Autonomy Class

Where delegation is permitted, how much freedom the agent has. The autonomy level is not a property of the agent — it is a property of the decision class under defined conditions.

Class	Definition	Governance requirement
A1 — Execute	Agent decides and acts without human notification.	Audit logging only. Agent decision record required for all actions.
A2 — Act-and-Report	Agent decides and acts, but must log the decision for human review within a defined window.	Agent Decision Log (ADL) entry created per action. Human review window defined (e.g., 24 hours).
A3 — Propose-and-Confirm	Agent proposes. Human confirms. Agent acts only after confirmation.	ADL entry created. Human confirmation recorded. Agent cannot proceed without confirmation.
A4 — Recommend	Agent provides recommendation. Human reviews and decides.	Agent recommendation logged. Human decision recorded separately.

An Agent Decision Log (ADL) entry is a structured record that captures each delegated decision — what was decided, by which agent, under what boundary, and who owns the outcome.

Dimension 3

Escalation Protocol

What happens when a decision cannot be resolved at the current authority level.

Pattern 1

Agent-to-Human Escalation

The agent reaches its autonomy boundary and transfers the decision to a human. This requires a named escalation target. If no escalation target is named, the agent cannot escalate — it will either make the decision outside its authority or stall.

Pattern 2

Human-to-Agent Override

A human reviews an agent's decision and overrides it. Every override generates an ADL entry that documents the rationale and, where appropriate, triggers a boundary review.

Pattern 3

Agent-to-Agent Conflict Resolution

Two agents produce incompatible outcomes. Either a priority rule gives one decision domain authority by design (security prevails over deployment), or both agents escalate to a named human authority and neither acts until the human resolves it.

How the three dimensions work together

%% caption: How an agent decision passes delegation scope, then autonomy class, then conflict escalation
flowchart TD
    D[Agent Encounters Decision] --> S{Delegation Scope}
    S -- "Non-Delegable" --> H[Human Decides\nDecision not delegated]
    S -- "Delegable / Conditional" --> A{Autonomy Class}
    A -- "A1: Execute" --> E1[Agent acts freely\nAudit log only]
    A -- "A2: Act + Report" --> E2[Agent acts\nADL created\nHuman reviews within window]
    A -- "A3: Propose + Confirm" --> E3[Agent proposes\nHuman confirms\nADL recorded]
    A -- "A4: Recommend" --> E4[Agent recommends\nHuman decides separately]
    E1 --> C{Conflict Detected?}
    E2 --> C
    E3 --> C
    E4 --> C
    C -- "No" --> G[Decision complete\nADL archived]
    C -- "Yes" --> P{Escalation Protocol}
    P -- "Priority Rule" --> PR[Higher-authority domain prevails\nSecurity > Deployment]
    P -- "Escalate to Human" --> EH[Both agents transfer to\nnamed human authority]
    PR --> G
    EH --> G

Interactive

Build your Agent Authority Map

Take one decision an agent makes — or will make — in your organisation. Classify it on each dimension and the map shows what AIDRA requires and where the gaps are. Add every decision class, then take the matrix to your next architecture review.

1 · Decision class

2 · Delegation scope

DelegableThe decision class can be authorized for agent execution under defined constraints.ConditionalThe decision class can be delegated only when specific safeguards are in place.Non-DelegableThe decision class must always be resolved by a human actor.

3 · Autonomy class

A1 — ExecuteAgent decides and acts without human notification.A2 — Act-and-ReportAgent decides and acts, but must log the decision for human review within a defined window.A3 — Propose-and-ConfirmAgent proposes. Human confirms. Agent acts only after confirmation.A4 — RecommendAgent provides recommendation. Human reviews and decides.

4 · Escalation target5 · Review cadence

What AIDRA says about this row

Start with delegation scope: has the organisation authorised an agent to make this decision at all?

Your Agent Authority Map

Nothing mapped yet. Add your first decision class above, or load the worked example from the article to see a complete map. Your map stays in this browser only.

Worked example

The AIDRA matrix

Decision class	Delegation scope	Autonomy class	Escalation target	Review cadence
Incident triage (low severity)	Delegable	A1 — Execute	—	Quarterly
Incident triage (high severity)	Conditional	A3 — Propose-and-Confirm	On-call Engineering Lead	Monthly
Purchase order (< defined threshold)	Delegable	A1 — Execute	—	Quarterly
Purchase order (defined threshold to ceiling)	Delegable	A2 — Act-and-Report	Head of Procurement	Weekly
Purchase order (> ceiling)	Non-Delegable	—	Head of Procurement	Per instance
Schema selection (internal)	Delegable	A2 — Act-and-Report	Domain Architect	Monthly
Integration pattern change	Conditional	A3 — Propose-and-Confirm	Integration Architect	Per instance
Vendor evaluation scoring	Non-Delegable	A4 — Recommend	Architecture Board	Per instance

Where the cell reads Non-Delegable or A4 — Recommend, the decision belongs to a human. The matrix makes it visible.

In practice

A SaaS company running three production AI agents — incident triage, deployment risk assessment, and customer routing.

Before: The incident triage agent believed to run at A2 (Act-and-Report) was actually running at A1 (Execute). The deployment agent's A3 proposals were being executed by default, because no escalation target was named.
After: AIDRA assigned the incident triage agent's executive escalation a Conditional delegation scope with a confirmed escalation target, and gave the deployment agent a named escalation owner and a defined confirmation window.

The agents did not change. The decision rights architecture did.

Implementation · 30–60 days

No new technology required

AIDRA is a structured extension of the decision rights model you already have.

Phase 1 · Weeks 1–2
Map the Agent Decision Inventory
Identify every decision class that agents are currently making — or will soon make. Document what is happening; do not design the target state yet.
DeliverableAgent Decision Inventory — every decision class, current autonomy level, and escalation path (if any)
Phase 2 · Weeks 3–4
Build the Agent Authority Map
Assign delegation scope, autonomy class, escalation target and review cadence to every decision class. Where agent behavior falls outside the assigned rights, record a Decision Rights Violation — not a technical bug.
DeliverableAgent Authority Map — the full AIDRA matrix for every agent decision class
Phase 3 · Weeks 5–8
Operationalize and Govern
Add a decision rights check to every new agent deployment, and compare documented autonomy classes against observed behavior at each governance review.
DeliverableNo agent enters production without a completed decision rights map

Go deeper

AI Decision Rights: Who Decides What When No Human Is Watching?The full framework, the failure modes and the 30–60 day implementation guide.AI-Native Teams Don't Need New Titles. They Need Named Owners.The three roles that own each AIDRA dimension inside a team.AI Governance Just Became a Job Title. Nobody Defined What It Owns.Using AIDRA to write the job description for an AI governance role.

Agents already in production?

Map their decision rights before an incident does it for you

Bring the map you built above — or the agent you could not classify — to a free 30-minute diagnostic.

Schedule a Free Diagnostic AI Strategy practice