How to Build an AI Chief of Staff in Notion (We Did It. Here's the System.)
Every founder I talk to is searching for an AI chief of staff. They want something that triages their inbox, routes decisions, compiles the weekly status, and stops them from being the person everyone waits on for answers. So they go looking for software. Fifteen tabs open. Twelve SaaS options. And somehow, they end up more overwhelmed than when they started.
Here's what I figured out: you don't buy an AI chief of staff. You build one. And Notion is the platform that makes it possible -- because the agent already lives where your team's knowledge, tasks, and projects are. We built ours at Modern Operators. We call it Vision. Here's exactly how it works.
An AI chief of staff is an agent-powered layer that handles the operational, coordination, and information tasks a human chief of staff would own -- triage, routing, reporting, Q&A from company docs -- so the founder can stop being the human search engine.
What a chief of staff actually does (and why founders need one)
Strip away the title. A chief of staff does five things: triages information before it hits the founder, routes decisions to the right person or system, follows up on open loops, synthesizes status across the business, and runs the weekly reporting cycle. If you're a founder with 5 to 30 employees and no COO, you are doing all five of those things yourself. That is the definition of the bottleneck.
A chief of staff exists to remove the founder as the single point of failure for information and decision routing in a growing business.
The problem is that a human chief of staff costs $150,000 to $300,000 per year fully loaded.[1] For most founder-led businesses under $10M in revenue, that hire doesn't pencil. So the founder stays the bottleneck, and the business grows slower than it should.
That's where the build-your-own case starts.
AI chief of staff vs. hiring a human chief of staff
Let's just put the comparison on the table.
Human Chief of Staff | AI Chief of Staff (Notion) | |
|---|---|---|
Annual cost | $150,000 to $300,000 fully loaded | ~$20/user/month (Business plan) + agent credits |
Availability | Business hours | 24/7 |
Political judgment | Strong -- reads the room, manages relationships | Limited -- data-driven only |
Context on your business | Needs onboarding, ramp time | Already inside your Notion workspace |
Handles repetitive coordination | Yes | Yes, at higher speed and volume |
Strategic advisory | Yes | No |
Here's my honest read: roughly 60% of a chief of staff's task load is information processing, coordination, and repetitive follow-up.[2] That 60% is exactly what an AI agent handles well. The other 40% -- sensitive negotiations, relationship navigation, judgment calls in ambiguous situations -- still needs a person.
For a 10-person founder-led business, build the AI layer first. You'll free up your own time, you'll surface the actual work that needs a human, and you'll have a much clearer picture of what you're hiring for if and when you bring someone in.
What an AI chief of staff can realistically handle (and what it can't)
The AI layer handles the operational 60% -- information triage, routing, reporting, and Q&A from your company docs -- so you and your team can focus on the judgment 40%.
Concrete things it handles well:
Weekly status digests. Pulls from open tasks and project databases, compiles a summary, routes it to the right people.
Q&A from company docs. Team member asks "what's our refund policy?" or "who owns the client onboarding process?" The agent answers from your Notion knowledge base instead of routing that question to the founder.
Inbox triage routing. Flags what needs a response, what needs a decision, and what can be handled without the founder touching it.
Meeting prep. Pulls relevant context from the workspace before a client call or team sync.
Task routing. New request comes in via Slack or email. The agent categorizes it, assigns it to the right person or queue, and logs it.
What it doesn't do: replace the judgment layer. A client relationship going sideways, a personnel decision, a partnership negotiation -- those stay human. And if your Notion workspace is chaotic, the agent will be chaotic. A disorganized system produces a disorganized AI. That's not a knock on the technology. It's just honest.
How we built ours in Notion: the Vision system
Modern Operators runs its entire content, research, and operations layer through a set of custom AI roles built directly in Notion -- a system we call Vision.
Here's the architecture in plain language.
We have distinct AI roles, each with a specific job. SCOTT is our SEO director. SCOUT is our researcher. MIMIC is our copywriter. Each role lives as a page in Notion with its identity, responsibilities, decision framework, and a memory layer. When a task comes in, it routes to the right role, and that role has everything it needs to do the work -- not because it was briefed in the moment, but because the context is already there.
Each role connects to our actual databases -- the blog pipeline, the keyword database, the task system. When SCOTT runs a brief, it pulls from the SEO Keywords database. When MIMIC writes a post, it references the brand avatar and voice guides that live in Notion. The agents don't need to be spun up and handed context every time. They already have it.
Then we have Notion Custom Agents running on triggers. One agent compiles weekly reporting from open tasks. One handles Q&A from our internal knowledge base. When a new task comes in, routing happens automatically.
This is what Notion meant when their co-founder said in May 2026: "Use your Notion database as a sheer canvas to power both your workflows and your agents."[3] The knowledge and the agent are the same system. That's the difference between a bolt-on tool and a compound system.
Notion's February 2026 release confirmed Custom Agents can run autonomously across Notion, Slack, Mail, Calendar, Figma, Linear, and custom MCP servers.[4] At Notion internally, 2,800 agents run around the clock. One Vercel Technical PM said at launch: "We have a Slack-mentionable agent that grades and rewrites drafts intended for the CEO, using our Chief of Staff's guidance on how to communicate with them." That's a chief-of-staff function, built in Notion.
How to start building your own AI chief of staff in Notion
The three prerequisites are documented processes, a shared source of truth in Notion, and defined routing rules. Get those in place, and the agents are the next natural step.
Here's where I'd start if you're building this from scratch:
Build your knowledge base first. Your agent is only as good as what it can access. If your SOPs, processes, and policies aren't documented in Notion, the agent has nothing to work with. Start here before anything else. how to organize your entire business in Notion
Define one routing rule. Pick the most common question your team asks you or routes through you. Write a simple decision rule: if X, then Y. That becomes your first agent instruction.
Set up one autonomous agent on a recurring trigger. Weekly status digest is the easiest starting point. Connect it to your task database, set a Friday morning trigger, and let it run. You'll see immediately what the agent can and can't do -- and that will tell you what to build next.
Don't add the agent until the system exists. I can't stress this enough. 40% of agentic AI projects are expected to be canceled, not because agents don't work, but because they were deployed on top of broken processes.[5] A well-organized Notion workspace is the foundation. The agent is the next floor.
AI agents for founder-led businesses no-code AI agents agentic OS for business owners
Frequently asked questions
What is an AI chief of staff?
An AI chief of staff is an agent-powered system that handles the operational coordination layer of a business -- information triage, decision routing, status reporting, and Q&A from company documents -- without requiring the founder or a human hire to manage that work manually. It handles the repetitive 60% so the humans can focus on the judgment 40%.
How much does an AI chief of staff cost vs. hiring one?
A human chief of staff runs $150,000 to $300,000 per year fully loaded in the US.[1] Building one in Notion costs roughly $20 per user per month on the Business plan, plus agent credits at approximately $10 per 1,000 credits (pricing as of October 2026 -- verify at time of build). For most founder-led businesses under $10M in revenue, the build-your-own approach delivers most of the operational value at a fraction of the cost.
Can I build an AI chief of staff without coding?
Yes. Notion Custom Agents are built in plain language -- you describe what you want the agent to do, set the trigger, and connect the sources. No code required. The more structured your Notion workspace is, the more effective the agent will be.
What are the limitations of an AI chief of staff?
AI handles the coordination and information layer well. It doesn't replace human judgment in sensitive or ambiguous situations -- personnel decisions, client relationship repair, strategic pivots. It also depends entirely on the quality of the system it runs on top of. A disorganized workspace produces a disorganized agent.
If you're the bottleneck, this is the fix
Most founders will keep searching for the right software. They'll find a list of 15 products, pick one, bolt it onto their existing chaos, and wonder why it didn't work.
The founders who get ahead of this build the system first. They document their processes in Notion, define how decisions get routed, and then deploy agents that run on top of that foundation. The agent gets smarter the longer it runs. The system compounds. The founder stops being the person everything routes through.
That's what we built with Vision. And it didn't start with the agents -- it started with building a real operating system first.
If you want to build yours, start with CompanyOS. It's the foundation we build for founder-led businesses: the systems, the structure, and the AI layer that runs on top of it.
About the author: Mark Malian is the Co-Founder of Modern Operators, where he helps founder-led businesses build operating systems that run without them. He writes about AI, operations, and what it actually takes to scale a business past the founder bottleneck.
