AI Writing Assistant
The AI Writing Assistant helps auditors create workpaper content faster and more consistently. It can draft narratives, generate conclusions, suggest test procedures, and enhance the clarity of existing text — all as suggestions requiring explicit human approval.
Accessing the AI Assistant
The AI Writing Assistant is available within any workpaper:
- Open a workpaper in edit mode
- Click the AI Assistant button (✨ icon) in the toolbar
- Choose an action from the menu
Available Actions
Draft Narrative
Generates a complete workpaper narrative based on:
- The engagement objective and scope
- Linked risks and controls
- Test procedures and their results
- Evidence file summaries
How to use:
- Click AI Draft Narrative
- Review the generated text in the preview panel
- Edit as needed
- Click Apply to insert into the workpaper
Generate Conclusion
Produces a workpaper conclusion based on:
- Test procedure results
- Exceptions found
- Severity of findings (if any)
The conclusion follows a standard format:
- Objective restated
- Summary of work performed
- Results summary
- Overall conclusion (satisfactory / needs improvement / unsatisfactory)
Suggest Test Procedures
Recommends test procedures based on:
- The engagement's risk and control framework
- Industry-standard testing approaches
- Historical test procedures from past engagements (via RAG)
Clarity Check
Reviews existing workpaper text for:
- Objectivity and professional tone
- Ambiguous language
- Missing cross-references
- Consistency with IPPF standards
Summarize Evidence
For attached evidence files (PDFs, spreadsheets), the AI can:
- Extract key data points
- Summarize document contents
- Identify relevant sections for the workpaper
AI Governance
All AI-generated content follows AIIA's strict AI governance rules:
- Suggestion only — AI output appears in a preview panel, never directly in the workpaper
- Human approval required — the auditor must click Apply to insert content
- Audit logged — every AI interaction is recorded with:
- The prompt sent
- The response received
- Whether it was applied or rejected
- The user who made the decision
- Citations provided — AI references source documents and data
- Confidence score — displayed to help gauge reliability
Provider Configuration
The AI Assistant uses the model configured in Administration → AI Models. The active model determines:
- Response quality and speed
- Token limits (how much text can be processed)
- Whether processing is local (Ollama) or cloud-based
For air-gapped deployments, configure Ollama as the provider for fully offline AI assistance.
Best Practices
| Do | Don't |
|---|---|
| ✅ Review every AI suggestion before applying | ❌ Blindly apply AI-generated text |
| ✅ Edit AI text to match your audit findings | ❌ Use AI output as the final word |
| ✅ Check citations and confidence scores | ❌ Ignore low-confidence suggestions |
| ✅ Use AI for drafting, then refine manually | ❌ Skip the review step |
Permissions
| Role | Can Use AI Assistant |
|---|---|
| Auditor | ✅ |
| Manager | ✅ |
| CAE | ✅ |
| QA | ❌ |
| Client | ❌ |
| Viewer | ❌ |
The ai:suggest permission is required. This is included by default for Auditor, Manager, and CAE roles.