A while back, I was building a knowledge base for one of my clients — a presentation skills coach who wanted his team to be able to access three years of his content without constantly asking him questions.
We took everything: podcast transcripts, course materials, blog posts, frameworks he'd developed. Loaded it all into a single knowledge base. Now, when his team needs to write something, or create a presentation, or check what he's said about a particular topic, they query the knowledge base instead of interrupting him.
The power wasn't just the retrieval. It was the centralization. Everything in one place, always current, accessible to everyone who needs it.
That same principle applies to your AI agents — and most people are doing it the inefficient way.
The Problem with Isolated Agents
When most people set up AI agents, they configure each one separately. The email agent gets the company background baked into its system prompt. The scheduling agent gets preferences and guidelines. The client support agent gets product details and pricing. Each agent has its own little pocket of context, isolated from the others.
This works fine at first. But it breaks down the moment anything changes.
A new team member joins. You move to a new office. A pricing policy shifts. A client changes their contact info.
Now you have to update every agent that contains that information. Hunt through system prompts. Remember which agent had which version of the details. Edit them one by one and hope you didn't miss anything.
I've seen people managing 8-10 agents start to dread making any changes because the maintenance burden is so high. They let the context go stale rather than face the update process.
The Shared Doc Solution
The fix is simple: one Google Doc.
You create a single document that contains all the context your agents share — business background, team roster, current policies, preferences, addresses, anything that applies to more than one agent. Then instead of pasting that information into each agent's system prompt, you load the Google Doc into the knowledge base of every agent.
Now when something changes, you update the doc. That's it.
Every agent that references the shared doc automatically has the new context. One update, everywhere, instantly.
I walked through this with Evan Baehr last December. He was running several agents across different areas of his business, and every time something changed — a new building came online, the team structure shifted — he'd have to touch each agent individually. Switching to the shared doc model was an immediate unlock. Update the doc, done.
What to Put in the Shared Doc
Think of this as your master context document. What does every agent need to know?
The basics usually cover:
- Business name, address, and contact info
- Team member names and roles
- Core services or products and how they work
- Communication preferences (how you like emails written, what tone to use)
- Current active projects and their status
- Key clients and relevant context about them
- Any recurring policies or rules that agents need to follow
You don't need every detail in the shared doc — only the things that multiple agents need. Specialized context that only applies to one agent can still live in that agent's individual knowledge base.
The shared doc is for the common layer. The stuff that should be consistent everywhere.
This Is What Centralized Context Means
There's a broader principle here that I call Centralized Context.
Most agent failures aren't about prompting or model selection. They're about fragmented, outdated, or missing context. The agent doesn't know what it needs to know. Or it knows something that was true six months ago but isn't anymore.
The solution is a durable memory layer that all agents can read from — and that's easy to update. Not a dozen separate system prompts you have to maintain. A readable document that acts as the single source of truth.
This is why Google Docs work well for this. They're readable. They're editable without any technical setup. Any agent platform that supports knowledge base loading (Lindy, CustomGPT, and most others) can reference a Google Doc. And you can keep it current without touching any agent configuration.
The decision hook I use: if multiple agents need the same truth, centralize it.
Setting It Up
The setup takes about 20 minutes.
Create a new Google Doc. Write out the shared context — business details, preferences, anything that applies across agents. Make it readable and specific. Don't write it like a legal document. Write it the way you'd brief a new assistant on their first day.
Then go into each of your agents and add the doc to the knowledge base. In Lindy, this is the Knowledge section of the agent settings. Most platforms have an equivalent. You're pointing the agent to the doc rather than pasting the content directly.
Test it. Ask an agent a question it could only answer if it had read the shared doc. Confirm it pulls from the right context.
Then the next time you need to update something — a phone number changes, a new service launches, a team member leaves — you update the doc. One place. Done.
The Longer-Term Payoff
The real value isn't the first update. It's the tenth.
The first time you update the doc and watch all your agents stay current without any additional work, it feels almost like a trick. The fifteenth time, it's just how you operate.
Centralized context is one of those foundational patterns that doesn't have a dramatic demo moment — it's just consistently better infrastructure for everything you build on top of it.
If you're managing more than two or three agents and still configuring them individually, this is worth your afternoon.
The 4-Day AI Sprint covers how to build agents that stay in sync, including the shared context doc setup. If you're building your first agent stack, that's a good place to start.
