How AI Works in LOQI
LOQI uses AI to accelerate audit work — analyzing documents, generating audit programs, drafting reports, and summarizing interviews. This page explains how the AI works and what controls are in place.
The role of AI
AI in LOQI is an assistant, not a decision-maker. Every AI output is:
- Presented as a draft for your review
- Editable — you can modify, regenerate, or discard any output
- Cited — claims reference specific passages in your source documents
- Transparent — AI-generated content is clearly marked
You make all final decisions. The AI handles the time-consuming parts (reading hundreds of pages, drafting initial text), and you apply your professional judgment.
What AI can do
| Feature | What the AI does |
|---|---|
| Risk Analysis | Reads uploaded documents and identifies risk items with categorization and rationale |
| Audit Program | Generates objectives, procedures, and testing attributes from your risk analysis |
| Report Sections | Drafts report content based on your findings, test results, and documents |
| Interview Summaries | Summarizes interview transcripts into key points and action items |
| Exception Extraction | Reads evidence documents and identifies potential test exceptions |
| Procedure Conclusions | Drafts a conclusion for a procedure based on test results |
| PBC Item Suggestions | Suggests what documents to request from management |
How the AI grounds its output
LOQI uses a technique called Retrieval-Augmented Generation (RAG):
- Your documents are split into chunks and indexed
- When the AI generates content, it first retrieves relevant chunks from your documents
- The AI generates text that references those specific chunks
- Citations link back to the exact passages used
This means the AI's output is grounded in your actual documents, not generic knowledge.
Data privacy
- AI processes your data only when you trigger a generation
- Prompts and responses are not stored by the AI provider after processing
- Your data is not used to train AI models
- All processing occurs within your deployment region
What's next
- Reviewing AI Output — Best practices for reviewing generated content
- Citations — How to verify AI citations