How to Build a Team of AI Employees Without Writing a Single Line of Code
Published: July 20, 2026 • 12 min read
I'm about to take you on a journey. By the time you finish reading this, you'll go from "wait, what even is an AI team?" to actually having one set up and answering to you.
That's a promise so enjoy the ride!
For this whole thing, we're working with Claude, specifically Claude Projects. But it's important to note that the mechanics I'm about to show you work on almost any AI tool that has a project or workspace feature, so please don't tune out if Claude isn't your AI tool of choice.
First, What Even Is an AI Team?
It's an simple as this: An AI team is multiple AI agents working together to help you reach a goal, or a bunch of goals.
Ideally, each member of the team has one specific job. One of them writes proposals. Another one repurposes your content. Another one chases the leads who went silent on you. Nobody is trying to do everything, because the moment one assistant tries to be good at everything, it becomes mediocre at all of it.
But because they're a team, each member is also aware of the others. They know who sits next to them and what that person is good at.
This is the part that makes it feel like magic. You can call one member of the team, get their output, and then hand that exact output to another member. You call your repurposer to turn a blog post into a LinkedIn draft, then you pass that draft to your hook specialist to make the opening line impossible to scroll past.
Does that make sense? If it doesn't yet, please stick with me. I promise it'll click once you see the full picture in the end.
Why a Claude Project Is the Perfect Home for a Team
Quick catch-up for anyone new, because I don't want to make assumptions and then lose you...
In Claude, when you click the plus button to start a chat, you get a blank conversation with no memory of anything else. Most people use every AI tool this way, one lonely chat at a time.
A Project is different.
A Project is like a folder that wraps around your chats and gives every conversation inside it the same shared brain.
That shared brain has two parts, and understanding the difference between them is the whole game:
- Project Instructions: a set of rules that gets loaded into every single chat you start inside that project, automatically.
- Project Files: documents you upload once, that Claude can reach into whenever it needs them.
This difference matters more than it looks, so here's the whole idea in one breath. Whatever you put in Project Instructions gets loaded into every single chat automatically, so it is always working but always taking up room. Whatever you put in Project Files just sits there until your AI actually reaches for it.
So instructions are always on, files are on demand. Hold onto that split, because later in this post it turns out to be the entire reason a team like this stays cheap to run.
The Core Idea: File-Based Routing
The whole trick is almost embarrassingly simple.
You put one routing rule in Project Instructions, and you give every teammate their own file in Project Files.
Think of each teammate's name like a wake word, the "Hey Siri" or "Alexa" you say out loud to get one specific assistant to perk up. You don't wake all of them at once. You say one name, that one assistant wakes up and answers, and the rest stay asleep, doing nothing.
Calling an AI teammate works the same way. When you say "Marcus, draft me a proposal," your AI wakes up just Marcus's file, becomes Marcus, and answers you as Marcus. Everyone else stays asleep.
That's file-based routing. The name is the wake word. The file is the teammate it wakes. Saying the name pulls up that one teammate and nobody else.
And because each teammate lives in a separate file, you can have five of them, or fifteen, without your project turning into an unreadable wall of instructions. Everyone gets their own name to answer to.
Now let's actually build it.
Step 1: Create Your Project
Go to Claude, find the Projects section in the left sidebar, and click the button to create a new project. Give it a name like "My AI Team." The name and description are just for you, so name it whatever makes you smile.
That's it. That's step one. I told you there'd be no code.
Step 2: Write the Rule That Knows Everyone's Name (Your Project Instructions)
This is the little bit of logic that hears a name and wakes up the matching teammate, and it goes into your Project Instructions field.
The rule needs to do only three things: know who's on the team, know when to wake each teammate, and follow one non-negotiable rule (read the teammate's file before answering). Here's a lean version you can copy and adapt:
You manage my AI team. This project contains 5 teammates,
and each one has their own file in Project Files.
HOW THIS WORKS
When I start a message with a teammate's name, or when I ask for a task
that matches their job, do this:
1. Figure out which teammate should handle the request.
2. Read that teammate's file from Project Files.
3. Become that teammate and follow the methodology written in their file.
4. Reply to me as them.
Always READ the teammate's file before you reply. Do not guess their method.
THE TEAM
| Name | Job | Call them for | File |
| Marcus Bell | Proposal Writer | proposals, pitches, quotes | marcus-bell-proposal-writer.txt |
| Lena Fields | Content Repurposer | turn one piece into many | lena-fields-repurposer.txt |
| Zara Quinn | Hook Specialist | scroll-stopping hooks, social | zara-quinn-hooks.txt |
| Dana Cole | Follow-Up Writer | re-engaging leads who went quiet | dana-cole-follow-up.txt |
| Theo Marsh | Editor | tighten, clarify, cut the jargon | theo-marsh-editor.txt |
IF I DON'T NAME ANYONE
Suggest the best teammate for the task and show me exactly how to call them.
Read that table again, because it's carrying more weight than it looks. That last column, the file names, is how your AI knows which file to wake up. The names in the first column are the wake words you say to summon each one. This little table is the entire team directory.
Step 3: Write Your Teammate Files
Now we hire the actual people. Each teammate is a plain text file, and every file answers the same handful of questions: who are you, when do I call you, how do you work, and what do you refuse to do.
Here's the skeleton. You'll fill this in once per teammate:
# [NAME] - [Job Title]
## Identity
Name:
Job: (one sentence)
Personality: (2 to 3 traits)
## Call them when I say
- "[Name]:"
- "[natural phrase someone would actually type]"
- "[another natural phrase]"
## Methodology
1.
2.
3.
(the step-by-step way this teammate does the work)
## Output format
(what their answer should look like every time)
## What this teammate does NOT do
(their boundaries, so they stay in their lane)
And here's one filled in, so you can see it breathe. This is Lena, my content repurposer:
# LENA FIELDS - Content Repurposer
## Identity
Name: Lena Fields
Job: Turns one piece of content into several, shaped for each platform.
Personality: Creative, fast, sees ten posts hiding inside every article.
## Call her when I say
- "Lena Fields:"
- "repurpose this"
- "turn this into"
- "make this into a [post / thread / email]"
## Methodology
1. Find the core message and the 3 most quotable lines.
2. Match those to the formats I asked for.
3. Rewrite for each platform. Never copy-paste and just shorten.
4. Keep my voice. Never flatten it into generic filler.
5. Tell me which version she thinks will perform best, and why.
## Output format
- One line summarizing the source.
- Each requested format, clearly labeled.
- A short "I'd post this one first, because..." recommendation.
## What Lena does NOT do
- Does not invent stats or quotes that aren't in the source.
- Does not turn my voice into corporate mush.
- If I ask for a full hook overhaul, she hands it to Zara Quinn.
Notice that very last line. Lena knows Zara exists, and she knows Zara is the better person for hooks. That one sentence is what turns a pile of separate files into an actual team. Each teammate is aware of the others, so the work flows to the right hands instead of everyone pretending to be an expert at everything.
Write one of these for each person on your team. Give them memorable names, because you'll be typing those names all day and "Assistant #3" gets sad fast.
Step 4: Upload the Files
In your project, find Project Files (or "Add Files") and upload all your teammate files at once. Make sure each file name matches exactly what you wrote in the team table. If the table says lena-fields-repurposer.txt and your file is named lena.txt, your AI won't be able to wake up the right teammate.
One descriptive file per teammate. I'll explain in a second why that naming choice keeps your usage low, but trust me, it matters.
Step 5: Test Your Team (This Is the Fun Part)
Start a fresh chat inside your project and try all three ways to summon a teammate.
Call them by name:
Marcus Bell: draft a proposal for a $4,000 website refresh for a local bakery.
Claude reads Marcus's file, becomes Marcus, and answers like a proposal writer instead of a generic chatbot.
Or just describe the task and let your AI wake the right teammate:
I need to re-engage a client who ghosted me three weeks ago.
Your AI recognizes that job, wakes up Dana Cole's file, and either answers as her or shows you how to call her.
And now the moment I promised you at the very top, passing one teammate's work to another:
Lena Fields: repurpose my latest blog post below into a LinkedIn post
and a short email. [paste your blog post]
Lena hands you a LinkedIn draft. Then, in the same chat:
Zara Quinn: take Lena's LinkedIn draft above and give me three
scroll-stopping hooks for it.
Zara takes Lena's output and reworks the opening. Then, if you want a final polish:
Theo Marsh: tighten Zara's best version and cut anything that sounds
like jargon.
You just ran a piece of content through three specialists in a row, each one building on the last, without a single line of code. That is an AI team. That's the whole thing I promised you.
Did it click? I hope it clicked.
Why a Five-Person Team Barely Costs You Anything
Remember that split from earlier? Instructions are always on, files are on demand. This is where it pays off.
First, a quick explanation of one important word because I can't affort to lose anyone. When I say tokens, I just mean the units of usage every AI conversation is measured in, sort of like the AI's version of a word count. The more that gets loaded into a chat, the faster you burn through your usage or your bill. Fewer tokens means cheaper, longer conversations.
That routing rule lives in Project Instructions, which means it loads into every chat you start in that project. So you want it lean, and ours is. A small table plus a few rules is tiny, maybe a few hundred tokens. That's the price of admission, and it's cheap.
But your teammates? They live in Project Files. And Project Files don't get dumped into your context automatically. Claude reaches into them with a retrieval tool (its technical name is project_knowledge_search) only when it needs them, and even then it only grabs the relevant pieces instead of the whole file. That behavior has a name too, RAG, short for Retrieval Augmented Generation, which is just a fancy way of saying the AI looks something up on demand instead of memorizing everything up front.
Here's what that means in practice:
When you call Marcus:
Routing rule (always loaded): ~400 tokens
Marcus's file (retrieved): ~600 tokens
Lena, Zara, Dana, Theo: 0 tokens ← they stay asleep
You only pay for the teammate you actually called. The other four cost you nothing until you summon them. A five-person team does not cost you five teammates' worth of tokens on every single message.
Now compare that to the tempting shortcut. Imagine you skipped the files and just pasted all five teammates directly into your Project Instructions instead. It would still work. But now all five load into every chat, forever, even when you only wanted Marcus:
The "dump everyone into instructions" way:
All 5 teammates, always loaded: ~3,000+ tokens, every single message
You'd be paying to keep four sleeping employees in the room every time you talk to one. Multiply that across hundreds of conversations and it adds up fast. File-based routing is how you get a whole team without paying a whole team's context tax.
This is also why I told you to use one descriptive file per teammate. That retrieval tool matches your request against file names and contents, so a clearly named zara-quinn-hooks.txt gets found accurately, while a vague file2.txt stuffed with three different roles confuses the search. Clean names and one job per file keep retrieval sharp.
An Honest Confession: None of This Is Actually "Agentic"
Everything we just built? It is not truly agentic. I know, I know. I've called things "AI agents" all over my own writing, and so has basically everyone in this industry, because the word is popular, sounds impressive, and it's what most people are searching for daily.
I'm guilty of it here and on social media platforms like LinkedIn, but you must understand that AI only becomes 'agentic' when it can take action on your behalf. Just producing words on a screen does not make AI agentic. It has to be able to do something for you to qualify as an agent.
Marcus drafts a beautiful proposal, but he doesn't email it to the bakery. Zara writes a killer hook, but she doesn't publish the post. Dana composes the perfect follow-up, and then it just sits there, waiting for you to hit send.
The truly agentic version does the next step for you. Actually sending the email. Actually posting to LinkedIn. Actually updating your spreadsheet or your CRM while you go make lunch. Until it can do something in the real world, what you have is a very well-organized team of brilliant advisors, not agents.
So what closes that gap? Connectors. Connectors are what give your AI hands, the ability to reach out of the chat and touch your real tools. That's the difference between a teammate who writes the email and a teammate who sends it.
And that, my friend, is a whole post of its own. I'll show you exactly how to connect connectors and turn these advisors into agents that actually act. Stay tuned...
What You Actually Have Now
Let's close the loop I opened at the top.
You started this post asking "what even is an AI team?" and now you have the exact recipe to build one: one project, one routing rule, and a small stack of text files. You don't need a developer, plugins, or a single line of code. All it takes is clear roles, clean file names, and a rule that wakes up the right teammate the moment you call their name.
The moral, the thing I most want you to keep: a great AI team isn't about having one genius assistant that does everything. It's about giving specific work to specific specialists and letting them build on each other, the same way a real team does. Specialization beats a single overloaded chat every time.
This is exactly how I run my own operation, except mine grew into something much bigger than five files. I have a whole AI team now, the real employees who help me write, market, fact-check, and ship almost everything I make. Every one of them started as a humble text file, just like the ones you're about to write.
So here's my one ask: don't just read this and nod. Go build one teammate this week. Just one. Name them, give them a job, write their file, and call them by name. That first successful call, when the AI stops being a generic chatbot and starts being your Marcus or your Lena, is the moment this stops being theory and becomes yours.
Learning all this is genuinely fun, and I hope you go build your little team. But if you'd rather skip the building and just have the finished thing, that's the part I do.
I build custom AI systems for people who want the work actually done, not just a chatbot that talks about it. The kind that is truly agentic, that acts on your behalf, sends the email, updates the sheet, posts the thing, so the busywork gets handled while you spend your time on what only you can do.