How to Build Your Own Personal AI Assistant

I quit my six-figure job at GitHub to become a full-time Creator. The problem? I still needed a way to create studio-quality content while traveling the world.

Every new place came with the same question: how do I make this random room look like an actual studio?

So I built a personal AI assistant for exactly that. I call it my Studio Engineer.

Whenever I’m in a new space, I send it a few pictures of the room. It looks at the space, checks the gear list I gave it, and tells me where to put my camera, lights, and everything else. I get a setup that works with the gear I already have instead of buying a bajillion new things every time I land somewhere.

And before you hear “personal AI assistant” and assume I built some complicated agent with APIs and a bunch of code, nope. I made this one with Projects in Claude. You can do the same thing with Projects in ChatGPT.

No code. You just need one problem you want help solving.

What is a Project?

A Project in ChatGPT or Claude gives you one central place for a specific task. You can add instructions and upload useful context, then every new conversation inside that Project starts with the information your assistant needs.

Think of a regular chat like explaining a task to someone new every time. A Project is more like working with the same person who already knows the job, your preferences, and what you’re working with.

That’s what makes it useful for building a simple personal AI assistant.

Step 1: Pick one specific problem

Start with one task. Please do not try to build an assistant that manages your entire life on day one. You’ll have a much easier time testing and improving it if its job is clear.

My problem was setting up a filming space while traveling.

Your problem could be:

• Turning your weekly brain dumps into newsletter drafts
• Planning meals around the groceries already in your kitchen
• Reviewing brand contracts against your usual requirements
• Creating lesson plans in a format you already use
• Helping you pack for trips based on the destination and weather

Pick something you do often enough that remembering your context would save you time.

Step 2: Brain dump everything into a chat

Open a new chat in Claude or ChatGPT and explain the problem like you’re talking to a friend who is about to help you solve it.

Tell it:

• What you’re trying to do
• What keeps going wrong
• What a good result looks like to you
• What limitations it needs to work around
• Any tools, materials, or information you already have

For my Studio Engineer, I explained that I’m a Creator who films long-form and short-form content for brands and for fun. I told it that I’m traveling, so every filming space will be different. Then I sent over my entire gear list and explained what I struggled with when trying to make a temporary setup look polished.

Do not worry about making this part pretty. It’s a brain dump. Have the conversation, answer its questions, and give it the messy context it needs.

Step 3: Test the solution in real life

Do not turn the conversation into a permanent workflow yet. Test it first.

I sent Claude pictures of the room I was staying in and asked it to walk me through the setup using my actual gear. Then I moved things around, checked the camera, and told it what worked and what didn’t.

The results surprised me. I used this process to film in a spare room at my mom’s house in Jamaica, an apartment in Croatia, and a random spot in Bali.

You cannot tell me those don’t look studio-like! Even the outdoor setup did what it needed to do.

Testing matters because the first suggestion might be almost right, but “almost right” is not the workflow you want to save. Keep trying the solution and giving feedback until the assistant consistently helps you get the result you want.

Step 4: Turn the conversation into Project instructions

Once the workflow works, ask the chatbot to turn the conversation into detailed instructions for a Project.

Here is the prompt I use:

Based on our entire conversation, create a detailed set of Project instructions for a personal AI assistant that can repeat this workflow in future chats.

The instructions should include:

1. The assistant's role and its single primary job
2. The context it should always remember about me
3. The information it needs to ask for before giving a solution
4. The exact process it should follow
5. The format its answers should use
6. The constraints, preferences, and available tools it must consider
7. What it should do when information is missing or unclear
8. A final checklist it should use before responding

Make the instructions specific enough that I can paste them into a new Claude or ChatGPT Project and get a consistent result without explaining the full workflow again. Do not invent details we did not discuss. Ask me questions about anything important that is still missing before you write the final instructions.

Read what it gives you. This is important. Make sure the instructions match the workflow you tested, remove anything it invented, and add anything the chatbot missed.

Step 5: Create your Project

Now create a new Project in Claude or ChatGPT.

Give it a name that makes its job obvious. Mine is called Studio Engineer because that is exactly what it helps me do.

Copy the instructions from the last step into the Project’s instructions section and save them.

Step 6: Add the context it will need again

Upload any information the assistant should have available every time you use it.

I added my gear list so my Studio Engineer never recommends a light, lens, tripod, or microphone I don’t own. If I buy something new, I can update the file.

Depending on your assistant, your Project knowledge might include:

• Your brand guide
• Examples of work you like
• A pricing sheet
• A checklist or template you already use
• Product information or business policies

Only add information that helps with the assistant’s specific job. More context is not automatically better. A folder full of unrelated documents can make the output worse, not smarter.

Step 7: Use it like an assistant

That’s it. You just built your first personal AI assistant.

Now, when I need to film, I open my Studio Engineer Project, start a new conversation, and upload pictures of the space. It already knows what kind of content I make, what result I’m going for, and which gear I have with me. I don’t have to explain my whole life again.

I’m heading to Vietnam next, and this assistant is about to put in WORK, especially as I start my YouTube channel.

Building a personal AI assistant doesn’t have to be complicated. Start with one annoying, repeatable problem. Solve it in a regular chat, test the solution, and save what works inside a Project.

I have a bunch of these Projects helping me run my Creator business now. If you build one, tell me what you made. I wanna know what task you’re finally taking off your plate.

This post was written with the assistance of AI from a human written script.