Your team keeps asking the same questions and re-creating the same documents because your institutional knowledge isn’t connected to your AI tools. This guide shows you how to build a team knowledge base using Claude Projects, Custom GPTs, and Notion AI in an afternoon.
You’ve probably noticed this already. You ask your team’s ChatGPT or Claude about your company’s standard process, and you get a generic answer that doesn’t match what you actually do. You ask it to draft a proposal in your brand voice, and it sounds like every other company’s proposal.
The AI isn’t broken. It’s working exactly as designed. It was trained on publicly available internet data. It doesn’t know your company exists. It doesn’t know your brand voice. It doesn’t know your internal processes. It doesn’t know which tools your team uses or how your customer workflows actually happen.
This is a massive opportunity cost. Every time someone asks a question that’s been answered before, they’re rebuilding institutional knowledge from scratch. Every time you draft something, you’re starting from generic templates instead of your company’s proven patterns. Every time you onboard someone new, you’re not pointing them to a single source of truth about how things work at your organization.
The solution is a team knowledge base: a collection of your company’s actual documents, processes, guidelines, and FAQs that gets connected to your AI tools. When your team asks Claude a question, Claude can reference your company’s real policies. When someone asks about brand voice, the AI knows exactly how your company communicates. When you need to draft something, you’re building on your actual best practices, not generic internet advice.
A team AI knowledge base is just your institutional documents fed into an AI system. That’s it. No setup with data engineers. No complex integrations. Just files.
Good documents to include: onboarding guides, brand guidelines, style guides, product documentation, process documentation, decision frameworks, customer case studies, frequently asked questions, pricing and terms, sales templates, and anything you notice people asking repeatedly.
Start with the documents that live in your heads. What do you have to explain to every new hire? What do you have to repeat to every new customer? What decision framework do you use repeatedly? What’s your pricing philosophy? What’s your hiring process? These become your knowledge base.
The magic happens when your AI system has access to these documents. Instead of giving a generic answer, it gives your answer. Instead of suggesting a standard template, it suggests how you would do it. The AI becomes a co-worker who actually understands your business.
Claude Projects is Anthropic’s tool for this exact problem. It’s the simplest way to get started because it requires exactly zero technical setup. You upload documents. You describe what you want. You talk to Claude.
Here’s how it works in practice. You create a Claude Project and give it a name like “Company Knowledge Base” or “Brand Guidelines.” You upload your documents: your brand guide, your pricing document, your product FAQ, your onboarding guide, whatever you want Claude to know about your business. You write a brief system prompt describing what the project is for. Then everyone on your team can access that project and ask Claude questions, and Claude will reference your documents in answering.
The limitations are real but manageable. Claude Projects is designed for ongoing work with detailed documents, not for sharing widely outside your organization. If you have a team of five people using Claude, this is perfect. If you need to give a client access to your knowledge base, you’d want Custom GPTs instead. Claude Projects also has a context limit, so if you have more than a few thousand pages of documents, you might hit the ceiling.
The advantage is speed. You can have a functional knowledge base working in less than an hour. Upload files, add your system prompt, you’re done. No waiting for IT, no integration headaches, no DevOps required.
Custom GPTs (built in ChatGPT) are the answer if you need to share your knowledge base with people who aren’t going to log into Claude. You can build a Custom GPT, set it up with your knowledge base, and then share it via link. Anyone with the link can use it.
Building one is straightforward. Go to ChatGPT, create a Custom GPT, upload your documents, write your system prompt (the instructions for how the GPT should behave), and save it. You can set it to “only me” (private), “link sharing” (anyone with the link), or “public” (anyone can find and use it).
Custom GPTs are better for distribution but slightly more limited on the AI side. You’re working with ChatGPT’s model, not Claude’s. If your team is already using ChatGPT, this integrates naturally into their workflow. If your team uses Claude, Custom GPTs feel like a different tool.
Custom GPTs also let you add “actions” (API integrations), so you can build a GPT that not only knows your company’s information but can also look things up in your CRM, pull documents from your file system, or send information to other tools. That gets more complex, but it’s possible.
If your team already uses Notion to store documentation, Notion AI is worth considering. You don’t need to set up a separate knowledge base tool. You just enable Notion AI and give it access to your workspace. When you use Notion AI, it can reference all your Notion documentation and help you draft, summarize, and search.
Notion AI is the least flexible of the three approaches but the most integrated if you’re already deep in Notion. The AI is more limited in what it can do (it’s built into Notion rather than a separate tool), but it’s seamless if your team’s documentation lives in Notion already.
The main limitation is that Notion AI is narrower in scope. It’s built for specific Notion workflows (summarizing, drafting, Q&A within Notion), not for building a custom business chatbot. If you want a more sophisticated AI knowledge base, Claude Projects or Custom GPTs are stronger choices.
Day one: gather your documents. Don’t overthink this. Make a folder and drop in the documents that your team refers to repeatedly. Your brand guide. Your standard operating procedures. Your product FAQ. Your pricing document. Your onboarding guide. Your customer persona document. Start with five to ten important files. You can add more later.
Day two: choose your platform. If your team uses Claude and is small (under 20 people), use Claude Projects. If your team uses ChatGPT or needs to share the knowledge base widely, use Custom GPTs. If you live in Notion, use Notion AI. This should take you 30 minutes.
Day three: upload and test. Create your project or Custom GPT. Upload your documents. Write a simple system prompt: “You are an AI assistant for [Company Name]. You have access to our company documentation. When answering questions, reference our documents whenever relevant. Maintain our brand voice.” Upload your files. Test it by asking it questions you know the answers to. Does it reference your documents? Does it get the tone right? Tweak the system prompt if needed.
Day four: train your team. Show three people on your team how to use it. That’s it. People will figure out the rest. Your job is just to break the ice. Here’s an example: “When you need to check if something aligns with our brand voice, ask the knowledge base. When you need to understand our pricing policy, ask the knowledge base. When you’re drafting something, ask it for an example.”
This is the hard part. Technology adoption is a behavior change problem, not a technology problem.
The most effective approach: use it yourself, visibly. When your team sees you referencing the knowledge base in meetings, asking it questions, using it to speed up work, they’ll ask how you did that. That’s your moment. Show them. Don’t make it a training session. Just show them quickly and let them try.
Reduce friction to zero. Make sure the knowledge base is immediately accessible where your team actually works. If your team uses Slack, put the knowledge base link in a pinned message. If you use email heavily, include it in your signature. Make it the path of least resistance.
Measure what’s changing. After a month, ask your team: “Are you spending less time searching for information? Are you asking fewer duplicate questions? Is documentation more consistent?” These are the metrics that matter. If the knowledge base is working, you’ll see these shifts. If you’re not, adjust what documents are in there or update the system prompt.
No. The three approaches covered in this guide, Claude Projects, Custom GPTs, and Notion AI, all require zero coding knowledge. You’re essentially uploading documents and describing what you want. If you can create a folder and type an instruction, you can build an AI knowledge base.
Start with the things people ask about most often: onboarding guides, brand style guides, company policies, product FAQs, process documentation, and anything your team consistently re-creates from scratch. Avoid including anything genuinely confidential unless your platform has enterprise-grade security.
Both let you upload documents and give the AI context about your organisation. Claude Projects is better for ongoing work with the same AI model and long conversations. Custom GPTs (via ChatGPT) are better for sharing a tool with your whole team, since they can be distributed as a link. The choice depends on whether your team uses ChatGPT or Claude.
Schedule a quarterly review where someone on the team updates the key documents. In Claude Projects, you can simply replace old documents with new ones. In Custom GPTs, you re-upload updated files. Most platforms don’t auto-sync with your internal systems yet, so manual updates are the current reality.
This is the right question to ask. For sensitive information, check your AI platform’s enterprise terms. Claude’s Team and Enterprise plans, ChatGPT Team and Enterprise, and Notion’s Business plan all offer data protection agreements that prevent your data from being used for model training. Consumer-tier accounts often don’t have the same protections. Always check before uploading anything confidential.
AI strategist and business advisor
Sana helps business leaders understand AI, translate between technical teams and executives, and build sustainable AI strategy. She’s particularly focused on the business, legal, and ethical implications of AI for companies outside the tech sector. She runs AI Bootcamps for non-technical professionals and leaders who need to make AI decisions.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.
