Agentic AI
AIIA's Agentic AI enables autonomous audit assistants that can perform multi-step tasks under human supervision. Unlike simple one-shot suggestions, agentic AI can plan, execute, and iterate on complex audit activities.
What Are AI Agents?
AI agents in AIIA are specialized assistants that can:
- Break complex tasks into steps
- Execute multiple actions in sequence
- Use tools (search, analyze, calculate)
- Produce structured outputs
- Request human approval at decision points
Available Agents
| Agent | Purpose | Key Capabilities |
|---|---|---|
| Audit Companion | General-purpose audit assistant | Answer questions, search history, explain standards |
| Data Agent | Data analysis and querying | Natural language queries, visualization, trend detection |
| Scoping Agent | Engagement scoping assistance | Risk assessment, scope recommendations, methodology suggestions |
| Review Agent | Workpaper review helper | Quality checks, completeness verification, consistency review |
AI Governance for Agents
Agentic AI follows strict governance rules:
- Suggestion-only outputs — agents never modify data directly
- Human-in-the-loop — approval required at each action step
- Full audit trail — every agent action is logged
- RBAC enforcement — agents can only access data the user can access
- Bounded execution — agents have defined scope limits
Using the Audit Companion
The Audit Companion is available as a sidebar chat:
- Click the AI Companion icon (🤖) in the sidebar
- Ask questions in natural language
- The companion can search your audit data, reference standards, and suggest actions
- All conversations are logged
Permissions
| Role | Can Use Agents |
|---|---|
| Auditor | ✅ |
| Manager | ✅ |
| CAE | ✅ |
| QA | ✅ Read-only queries |
| Client | ❌ |