tl;dr Don’t wait for one tool that can do it all—build your own AI squad and assign each role. You’ll move faster, stay in control, and spend more time on the parts of product development that actually excite you.


One AI to rule them all? That’s a trap.

Everyone’s waiting for the one: a magical AI tool that designs, writes, codes, and deploys while you sip cold brew.

That’s not how I work. And honestly, I’m not sure I want it to be.

Instead, I’ve found a much more realistic—and scalable—approach: treat each AI tool like a teammate. Give it a specific role. Let it shine in the right part of the workflow. Stay involved in the things only you can do.


Here’s how I orchestrate my “AI team”

When I kick off a new feature or idea, I think like a PM first.

1. ChatGPT = Solution Architect

I always start by thinking like a product manager:
“As a user, I want [X] so that [Y].”

Then I describe the current context—what’s in place, what’s missing, and what the desired experience should feel like. I plug that into ChatGPT (or Claude) and ask it to play prompt engineer:

“Turn this user story and background into a detailed technical brief.”

ChatGPT excels here: it identifies edge cases, suggests data models, and sometimes even catches things I missed.

I’ll often prime the model with screenshots, style guidelines, or example interactions from Dribbble or my own app. This is where I shape the big picture.

2. Cursor = Senior Developer

Once I have a solid brief, I move into the codebase with Cursor.

Cursor helps me: - Find relevant files across my Rails monolith - Scaffold new components or endpoints - Write tests I’d otherwise forget - Catch inconsistencies or unhandled logic

It works best when I’ve already thought through the logic. Think of Cursor like a skilled dev—you still need to lead.

3. Me = Product/UX Lead

Here’s where human me steps in: - Review and refine UI in Figma - Pull references and patterns from the design system - Spot where brand, voice, or user flow needs love - Time-box feedback loops so I don’t endlessly iterate

The goal isn’t perfection on pass one—it’s momentum. Good enough is often great enough to ship, test, and then evolve later.


Why orchestration beats over-reliance

When you assign roles, a few good things happen:

Benefit Why it matters
Each tool stays in its lane ChatGPT thinks big. Cursor stays close to code. Figma keeps the UI grounded.
You don’t lose control You’re not handing the whole project to a single black box. You’re staying in the loop.
You can upgrade in pieces Swap Claude in for ChatGPT? Easy. Test new AI dev tools? Go for it. You’re modular.
You can scale faster Each tool adds leverage. Together, they 10x you—without replacing you.

Pain points and workarounds

Yes, even orchestration has bumps.


Recent “wow” moment

I’d been avoiding a Stripe subscription bug—nervous it might mess with money.
But this time, I walked through it with Cursor + ChatGPT.

We reviewed docs, explored the code, scaffolded a fix, and tested it—all in a single focused afternoon. No panic. No rabbit hole. Just fast, accurate work with great backup.


Advice I’d give a junior teammate

If you’re just getting started with this stack, here’s what I’d say:

  1. Learn what makes a good user story. Break ideas down into small, testable pieces.
  2. Use different tools for different hats. Treat ChatGPT like a PM, Cursor like a dev, and Figma like your UI pair.
  3. Create chat “roles.” One conversation might be your solution architect. Another is a tech writer. Another is your QA. Label them. Revisit them.

The clearer you are, the better your tools can support you.


Final thought

Don’t wait for one tool to do it all.

Orchestrate the tools you’ve already got.
Assign them roles.
Stay in the loop.
And use your best human skills—judgment, taste, curiosity—to guide the whole show.

Orchestration illustration


Got your own AI tool stack or orchestration tips? I’d love to hear what’s working for you—DM me or reply on LinkedIn or X. 🚀