Arkananta AI Audit Assistant
- Automation
- 2026
The chatbot answers audit questions and points to the document behind each answer. It can also find and send the file an auditor needs. The knowledge base lives in Google Drive and is searched through a RAG pipeline.
The problem.
An audit answer is only useful when someone can check its source. The system also needed to accept new files without rebuilding the workflow each time the knowledge base grew.
How we framed it.
Google Drive is the source of truth. The n8n workflow retrieves relevant documents, gives that context to the model, and returns the source file with the answer. A GPT based model and a fallback model handle the chat, while a file upload path adds new knowledge to the search index.
What we delivered.
The finished assistant combines a front end, n8n workflows, and the required integrations. It answers with citations, retrieves supporting files, and accepts uploaded documents. A fallback model keeps the chat available when the primary provider is unavailable.
How it came together.
- Mapped the audit questions, file sources, and evidence requirements.
- Connected Google Drive as the source and designed the RAG retrieval flow.
- Built the n8n backend for model routing, document retrieval, and fallback handling.
- Built the chat interface for answers, evidence, and files.
- Added a file upload path so the knowledge base can grow without a rebuild.
System plates.
THE STACK
Reflection
The most important rule was simple: the assistant must show a source for every answer. That keeps the tool useful for audit work and gives the team something they can verify.
Work delivered for a real client or organization, produced by the Dignify studio.
