I’m dropping a workflow today that solves the most expensive bottleneck in consulting: The Discovery Phase.
We all know the drill. You sign a client, they dump 50 PDFs, contracts, and SOPs into a Google Drive, and you (or your junior staff) spend 2 weeks reading them just to figure out what’s broken.
So I built an n8n workflow that does that 2-week "Document Review" in 5 minutes.
This isn't just summarizing a PDF. It is a multi-agent RAG (Retrieval-Augmented Generation) system designed with a "Split Brain" architecture to handle heavy data loads without hallucinating.
How it works:
The Input: You submit a Google Drive Folder ID containing the client's "Data Room."
The Brain (Vector Store): The workflow ingests binary files, runs them through a Recursive Character Text Splitter (to maintain semantic context across chunks), and upserts the vectors into Pinecone using OpenAI embeddings.
The Investigation (The Secret Sauce): instead of dumping everything into one massive prompt (which degrades quality), I set up the Vector Store as a "Tool" and split the logic into 3 Parallel Agents. This allows each agent to query only the specific vectors it needs:
- Operations Agent: Queries for process bottlenecks and workflow inefficiencies.
- Risk Agent: Scans specifically for financial liability and contract loopholes.
- Culture Agent: Retrieves data related to employee sentiment and toxic patterns.
The Deliverable: It aggregates the outputs from all three branches into a structured "Gap Analysis Report" in Google Docs.
The ROI:
Speed: Walk into the kickoff meeting knowing their problems better than they do. Accuracy: Because I used a "retrieve-as-tool" logic, the AI cites the specific document for every finding. Scale: You can run this asynchronously for multiple clients without hitting token limits.
JSON attached below. Requires an OpenAI key and a free Pinecone index.