Agentic (Pillar: Agentic)
Agentic refers to the use of autonomous or semi-autonomous AI agents that can plan, use tools, and make decisions – often across longer time periods and multi-step processes.
What does Agentic mean in WAF++?
The Agentic pillar addresses the architectural and operational requirements for:
-
Autonomous Agents: Single AI agents with planning and decision-making capabilities
-
Multi-Agent Systems: Cooperating or competing agents with shared goals
-
Agent Orchestration: Coordination across agents (Manager/Worker, Swarm, Mesh)
-
Human-in-the-Loop: Interfaces between human decision-makers and agents
-
Agent Safety: Guardrails, verification, alignment, and fail-safe mechanisms
| Dimension | What is controlled? | WAF-AGN Control |
|---|---|---|
Agent Identity |
Are agents authenticated, addressable, and traceable? |
WAF-AGN-010 |
Tool Use |
Which tools may agents use – and what are the limits? |
WAF-AGN-020 |
Reasoning |
Is decision-making traceable and auditable? |
WAF-AGN-030 |
Memory & State |
Are memories and states managed consistently? |
WAF-AGN-040 |
Safety & Guardrails |
Are fail-safe, alignment, and verification in place? |
WAF-AGN-050 |
Orchestration |
How are many agents coordinated? |
WAF-AGN-060 |
Observability |
Are agent activities visible and measurable? |
WAF-AGN-070 |
Human-in-the-Loop |
When and how does human intervention happen? |
WAF-AGN-080 |
Cost & Performance |
Are agent operations economic and performant? |
WAF-AGN-090 |
Agent Debt |
Are known limitations documented? |
WAF-AGN-100 |
Controls Overview
The Agentic pillar is operationalized through 10 measurable controls (WAF-AGN-010 to WAF-AGN-100).
| Control ID | Title | Severity | Automatable |
|---|---|---|---|
Agent Identity & Authentication |
High |
High |
|
Tool Use Governance |
High |
High |
|
Reasoning Transparency |
Medium |
Medium |
|
Memory & State Management |
High |
High |
|
Safety & Guardrails |
High |
Partial |
|
Multi-Agent Orchestration |
Medium |
Medium |
|
Agent Observability |
Medium |
High |
|
Human-in-the-Loop |
High |
Low–Medium |
|
Cost & Performance Monitoring |
Medium |
High |
|
Agent Debt Register |
Low |
Partial |
Quick Start
New to the Agentic pillar? Recommended reading order:
-
Definition – What is Agentic as a discipline?
-
Scope – What is in scope, what not?
-
Agentic Principles – 5-7 core principles
-
Architectural Patterns – Typical systems
-
Maturity Model – Assess your starting point
-
Agent Design – Architectural questions
-
Controls – The 10 measurable controls
-
Best Practices – Implementation
Agentic Principles
The Agentic pillar is guided by seven core principles:
| 1. Agent Identity First | Every agent must be identifiable, authenticatable, and traceable. |
|---|---|
2. Tool Use Governance |
Not every agent should use every tool. Tool use must be approved, logged, and limited. |
3. Reasoning Transparency |
Decisions must be traceable – not a black box. |
4. Human-in-the-Loop by Design |
Sensitive decisions require human oversight. |
5. Safety & Guardrails |
Fail-safe, alignment, and verification must be in place. |
6. Memory & State Consistency |
Memories and state must be managed consistently. |
7. Agent Debt Transparency |
Limitations, knowledge gaps, and technical debt must be visible. |
Architectural Patterns
Three primary patterns for agentic systems:
| Pattern | Description | When to Use |
|---|---|---|
Single Agent |
One autonomous agent with clear responsibility |
Simple tasks, early experimentation |
Swarm |
Many homogeneous agents with shared goal |
Parallel processing, scalable workloads |
Mesh |
Heterogeneous agents with specialized capabilities |
Complex workflows requiring multiple expertise |
Best Practices
Key implementation considerations:
-
Agent Safety – Safety first
-
Reasoning – Transparency
-
Memory – Consistency
-
Tool Use – Governance
-
Cost – Economy
Related Pillars
The Agentic pillar interacts with all other WAF++ pillars:
| Pillar | Interaction with Agentic |
|---|---|
Operational Excellence |
Observability (OpsEx) enables agent monitoring. Runbooks (OpsEx) guide agent behavior. |
Reliability |
SLOs (Reliability) monitor agent performance. Agent failures trigger SLO burn rate alerts. |
Security |
Guardrails (Agentic) enforce security policies. Security controls are applied via agent tool use. |
Cost |
Agent operations (Agentic) impact cloud costs. Cost controls (Cost) apply to agent compute. |
Architecture |
Agent architectures (Agentic) influence system design. Architecture decisions affect agent capabilities. |
Governance |
Agent Debt Register (Agentic) informs governance prioritization. Controls map to compliance frameworks. |
Sustainability |
Agent efficiency (Agentic) impacts carbon footprint. Model selection affects environmental impact. |
Efficiency |
Agent caching and optimization (Agentic) improve efficiency. Efficiency controls apply to agent operations. |