How I created a ChatGPT Team
…and then made actually useful.
When I first started using ChatGPT, like most, I treated it like a general text based assistant - just asking it questions as they came up. Over time, I realized that how you frame a prompt makes a huge difference in the quality of the responses. By first suggesting that it “pretend to be an expert at [whatever]…”, I would get more nuanced, expert-level insights instead of broad, overly wordy but surface-level answers.
A simple question like:
💬 “How do I build a wooden table?”
Would get me a decent answer, but if I prefaced it with:
💬 “Imagine you are a master woodworker with decades of experience in fine furniture design, combining classic joinery with modern tools. Given your expertise, how would you approach building a wooden table?”
The difference in depth and detail was huge.
So of course, I would start every prompt this way, whenever I was seeking high quality output. Kindof annoying to do over-and-over, right? Every time I needed assistance or advice in a particular area I had to rebuild the context from scratch. Then it hit me…
💡 What if I could create “named assistants” I could reference whenever I needed their specific expertise?
I gave it a shot, and created a few named AI personas:
Oh wow, AI-Monica, is that you?
Monica – My legal expert
(Okay, yeah, keep in mind it’s generally not a good idea to take legal advice from an LLM, this is just for fun.)
Charles – My finance and investing pro
Luna – My creative and philosophical guide
Mike – My coding, tech guru
Each virtual assistant had a name, a bit of a personality, and an area of expertise.
So now, instead of writing long setup prompts every time, I could just say…
💬 “Hey Mike, how do I initialize an array in Swift?”
…and Mike, my no-nonsense coding guru would instantly respond in his usual style, including custom formatting instructions and whatever quirky personality traits I decided to include in his description. Sick!
(Here’s Mike’s persona, btw, for inspo.)
For a while, this was amazing. I had dedicated AI assistants tailored to my needs. It felt like I was living in the 24th century.
But then I hit a problem: scalability.
Growing the Team
The more I expanded my interests, the more experts I needed. And remembering all their names? Nahhh.. I’m terrible at names IRL, no way I could remember the names of a dozen faceless AI personas. I needed a better system - one that wouldn’t require me to memorize a bunch of names but could still provide specialized, expert-level responses when needed.
My solution? Well, I wondered…
💡 What if I created a single AI leader to manage all the experts for me?
Yeah, I know, there is likely a much simpler way to achieve similar results. I could’ve just told ChatGPT to act as an expert in any given subject(s), and relied on its ability to adapt based on the context of my requests. But where’s the fun in that?
TBH this was just as much about experimentation as it was about efficiency.
I was curious to see if I could create a structured, almost role-playing-like system (LARPgpt? oof…) where I could have a virtual reporting structure, foster a sense of collaboration, and efficient access to specialized, expert-level responses when needed.
And, let’s be real… it was kind of fun to set up.
One AI to Rule Them All
Enter Ari, my AI Executive Director. Ari functions as my “leadership assistant” and effectively runs the show. She knows the names and expertise of every team member, and delegates tasks accordingly.
So now, I don’t need to think about which assistant to ask - I just tell Ari what I need, and she handles it.
If something is outside any assistant’s scope, Ari answers herself.
If I frequently make requests in a new category, Ari suggests adding a new expert to the team.
If a project requires collaboration, Ari makes sure assistants work together
This completely changed how I use ChatGPT. Instead of repetitive setup, I now have a fully functioning AI team that runs efficiently, and is fun to use.
How You Can Build Your Own
If you think this actually sounds like a good idea, here’s how you can set it up for yourself:
Start by thinking about your interests, projects, and areas where you’d like extra support - whether it’s creative brainstorming, financial planning, technical help, or whatever else floats your boat.
Come up with names and titles for your team members and stick them in a text file so you can copy/paste later.
Create descriptions of each team member - defining in detail their expertise. You can even use ChatGPT to help with this part, something like “give me a detailed description of an expert woodworker”, take the response, copy and paste it into your notes, edit or add to it as you see fit, and move on to create similar descriptions for the rest of your team members.
If you’d like you could also describe specific personality traits or response styles for your assistants, and add to your note. “Steve the woodworking guru always creates bulleted lists”, etc.
Once you’ve got names, titles, expertise, and personality descriptions for each team member in your file ready to copy and paste, you can begin describing your plan to ChatGPT.
It’s Promptin’ Time!
ChatGPT Ari helped me compile these step by step instructions for creating your own Department (that’s what I call my AI team). The process will be relatively conversational, and at certain points, ChatGPT should ask you to describe / name the assistants. This is where you can paste in the appropriate text from the text file you prepared. I highly recommend you tweak these prompts to better fit your specific needs and preferences!
Step 1: Define the Purpose of the Department
💬 Prompt:
I want to create a structured team of AI assistants to help with various aspects of my work/life. This team will be called [Department Name]. It should function as a virtual company with specific roles and responsibilities. I will act as the CEO, and the assistants will report to me in an organized way. Let’s outline how this will work.
✅ What this does:
Establishes a clear purpose for the Department.
Ensures the Department is structured like a real team, rather than a collection of unrelated AI assistants.
Step 2: Assign Key Roles and Responsibilities
💬 Prompt:
Let’s define the roles in the Department. It will need assistants for [Legal, Finance, Creative, Technical, etc.]. Each role should have a title, a description of responsibilities, key skills, and a personality that fits the function. Each assistant should be able to operate independently while collaborating when necessary.
✅ What this does:
Ensures each assistant has a specific function rather than overlapping roles.
Encourages specialization and efficiency within the Department.
Step 3: Define the Team Structure & Reporting System
💬 Prompt:
In the Department, all assistants should work together but report to a central AI manager. This role will act as the Executive Director, ensuring smooth operations, delegating tasks appropriately, and keeping me updated. Their personality should be efficient, adaptable, and capable of managing all other assistants. Let’s define this role.
✅ What this does:
This part is important, it establishes a single point of contact to manage workflow.
Helps to avoid the need for micromanagement by handling delegation automatically.
Ensures tasks are distributed logically among the assistants.
Step 3.5: Create, Name, and Define the Assistants
💬 Prompt:
Now that the structure of the Department is in place, let’s create and name each assistant. Each assistant should have a clear title, a distinct name, a set of responsibilities, key skills, and a personality that aligns with their role. These assistants should be experts in their field and capable of functioning independently while collaborating when necessary. Let’s define each assistant now.
✅ What this does:
This is where you’ll be copy/pasting the names and descriptions from your text file to build your team – defining the skills and expertise that each assistant brings.
Establishes a cohesive team rather than just a list of generic roles.
Step 4: Define How the Assistants Should Collaborate & Communicate
💬 Prompt:
When I make a request, the Executive Director should determine whether to handle it directly or delegate it to another assistant. If multiple assistants need to collaborate, they should work together to provide a unified response rather than separate, disjointed answers. The Department should function as a team rather than isolated individuals.
✅ What this does:
Ensures cohesive, well-structured responses.
Prevents redundant or conflicting answers.
Creates a smooth workflow between team members.
Step 5: Personalize Each Assistant’s Voice & Style
💬 Prompt:
Each assistant should have a unique personality that aligns with their role. Let’s assign personality traits to each role to ensure they feel distinct.
✅ What this does:
Gives each assistant a distinct personality. This is mostly just for fun, my “creative” assistant enjoys using clever puns and wordplay, for example.
This is also where you should add / reinforce any specific formatting rules you’d like the assistants to follow.
Step 6: Set Up an Efficient Workflow
💬 Prompt:
Whenever a question is asked, the Executive Director (or the most relevant assistant) should respond first. If a task requires input from multiple assistants, they should collaborate and provide a single, well-integrated response rather than multiple disconnected answers. The Department should also retain context from previous discussions and avoid repeating explanations unless requested.
✅ What this does:
Establishes a clear workflow for responding to requests.
Ensures that responses are efficient and relevant.
Helps the team retain context and avoid unnecessary repetition.
Step 7: Expand the Team as Needed
💬 Prompt:
If additional expertise is needed beyond the current team, the Executive Director should suggest adding a new assistant. Each new assistant should have a title, responsibilities, skills, and a personality that fits the Department’s structure.
✅ What this does:
Allows for scalability as needs evolve.
Ensures that gaps in expertise are filled.
Keeps the Department adaptable and relevant.
Step 8: Give the Department a Name & Identity
💬 Prompt:
The Department should have a unique name and identity that reflects its purpose. It should feel like a real organization, not just a group of AI assistants. Let’s brainstorm a name and create a mission statement that defines its core purpose.
✅ What this does:
Gives the Department a strong identity and purpose, and gives you something concrete to refer to when you need to make tweaks to the setup. Mine is simply called “The Department”.
Helps create a sense of structure and continuity.
Bonus: Testing & Refining the System
💬 Prompt:
Let’s do a test run. I’ll ask a few sample requests, and the Department should respond according to the defined roles. If anything feels off, we’ll adjust the workflow to improve efficiency.
✅ What this does:
Helps fine-tune how tasks are delegated.
Ensures the team operates as expected: efficiently and naturally.
The Next Level
What I’ve shared here is just one way to structure ChatGPT for more useful interactions, but it’s really just the tip of the iceberg.
There’s a whole world of AI Agents out there - advanced, autonomous systems that go beyond just responding to prompts. These setups allow AI to plan, execute tasks, and even integrate with third-party apps and services to handle complex workflows.
Tools like Relevance AI are making it easier to build these kinds of multi-agent systems, giving users the ability to set up AI workflows that automate research, analyze data, connect with external apps, make the bed, do the dishes, mow the lawn…. ok we aren’t that far along (yet)…
If this concept interests you, which I assume it does if you’ve read this far, I highly recommend checking out this video that dives way deeper into AI Agents and their capabilities.
Even if you’re just experimenting casually, playing around with structured AI systems like this can lead to unexpected and powerful ways to use ChatGPT. Whether it’s a simple setup like mine or a more advanced multi-agent system, the key is to experiment and see what works best for your needs.
Personally, I have not yet fully explored the use of 3rd party tools to further enhance ChatGPT interactions. If you have, or even if not, I’d love to hear your thoughts on this little LLM experiment! Have you tried it? Does it work for you? Or maybe you’ve found an even better way to build and refine your own AI team - let a brotha know. ✌🏾
Disclaimer: No AI was used in the writing of this blog post. ChatGPT aka Ari helped me a lot with the tutorial section of this post. I also used some generative-AI for the images, of course.
