I've been tasked with researching AI knowledge base options for our company, and I'm honestly a bit overwhelmed by the choices. We're a mid-market retailer with about 50 stores, and our current "knowledge base" is a chaotic mix of Google Drive folders, old SharePoint pages, and tribal knowledge in people's heads. The goal is to unify product info, SOPs for store operations, and vendor details into something our corporate and store teams can actually query.
I started my review with NotebookLM because of its focus on grounding answers in your own documents. I uploaded some of our product catalogs and a new visual merchandising guide as a test. The immediate appeal was how it could summarize the key points from the long guide and answer specific questions like "What are the fixture requirements for the new spring apparel line?" It pulled the right specs from the document. That's promising for training new store managers.
But I'm running into a wall when I think about scaling this for the whole company. Our data lives in many places: our CRM (HubSpot), our inventory system, and thousands of past vendor emails. NotebookLM works on documents you give it, but I'm not sure how it would connect to live data sources or be accessed by hundreds of employees. The chat interface feels more for individual analysis than a company-wide reference tool.
So my question is, has anyone successfully implemented an AI knowledge base in a similar scattered, multi-location retail environment? What actually works? I'm looking at tools like Guru or Slite with AI add-ons, but I'm curious if a solution like NotebookLM can be the core if we structure our documents better, or if we need something built more for business operations from the start. My next step is to map out all our key information sources, but I'd love to hear about real workflows.
You've hit on the core challenge. NotebookLM is great for a clean, document-focused pilot, but it's not built to connect to your live systems. That integration piece is where most companies get stuck.
For a setup with data in HubSpot, inventory systems, and email, you need something that can either pull from those sources directly or sync with a central repository they all feed into. The tool itself is less important than the data pipeline you build for it.
I'd suggest pausing the tool evaluation and mapping out exactly which data sources are "must-haves" versus "nice-to-haves" for your initial rollout. Trying to ingest everything at once is a recipe for that overwhelmed feeling. Maybe start with structured docs like SOPs and product catalogs, prove the value, then tackle CRM integration in phase two.