What is a pattern?

Components answer "what." Patterns answer "when and how."

A RAD component is a building block — a Disclosure Alert, an Approval Gate, an Audit Trail. A RAD pattern is the choreography: which components appear at which moment, in what order, with what conditions, and in response to which system events.

Span the full user journey

Patterns cover complete flows — not single moments. Each one maps a user journey from trigger to resolution.

Encode governance requirements

Each pattern specifies when consent is required, what must be logged, and what cannot be auto-approved.

Reference specific components

Every pattern names which RAD components it requires, with links to the component documentation.

Pattern Families

Pattern Family · 01

Agentic Execution Flow

How RAD components combine to cover the full arc of an autonomous agent run — from pre-execution disclosure through active execution, approval gates, and post-run audit.

Pattern Family · 02

Error & Recovery

How RAD error and recovery components combine to handle the distinct failure modes of AI systems — from model errors and agent pauses to confidence failures and scope violations.

Pattern Family · 03

Multi-Agent Orchestration

The execution arc for a network of agents. Governing principle: trust doesn't transfer automatically. Consent to run the orchestrator is not consent to run every subagent it may spawn.

Pattern Family · 04

Multi-Agent Governance

The failure and accountability layer for multi-agent runs — handling network-level failures, conflicting outputs, and aggregate audit rollup across the full agent network.

How to use patterns

Start with the pattern. Reach for the components.

Patterns define the choreography. Components are the instruments. Pick the pattern that matches your scenario, then use the required components to execute it. The pattern tells you when each component should appear and why.

01

Identify your scenario — what is the agent doing? What could go wrong? Who needs to authorize what?

02

Choose the pattern family that matches your failure mode or execution arc. Read the governing principle first.

03

Implement the required components in the order the pattern specifies. Follow the governance requirements — they are not optional.

All Sub-Patterns

Agentic Execution Flow

01Pre-Execution Disclosure 02Active Execution State 03Approval Gates 04Post-Run Audit

Error & Recovery

01Model Error Recovery 02Agent Pause & Override 03Confidence Failure Escalation 04Scope Escalation Recovery 05Bias & Statistical Risk Alerting

Multi-Agent Orchestration

01Pre-Orchestration Disclosure 02Spawn-Time Authority Gates 03Parallel Execution State 04Handoff & Context Visibility

Multi-Agent Governance

01Partial Failure Isolation 02Conflicting Output Resolution 03Aggregate Audit Rollup