AI Tools for Product Teams – Part 4: Delivery (And Why I Skipped Discovery 😅)

How the new way of coding tools could help you a lot as a PdM.

Hey there 👋

This post was supposed to be about Discovery.

But, like in most product journeys, reality stepped in. We've been neck-deep in the Delivery phase lately. Not because everything was perfectly validated, but because I wanted to see how these new sort of AI coding tools could help me in my job as a PdM.

So instead of sticking to the plan, I figured I’d share what’s been working for me in real life. Especially the AI tools I’ve been using every single day to keep momentum going. This one’s from the trenches — not the theory.

---

Why PMs Should Care About Delivery (Even If You Don't Write Code)

Let me tell you about the first time this hit me:

We had an idea worth testing (our first personalization algorithm), but all our engineers were tied up in a massive refactoring. No one had bandwidth to jump in on this small discovery initiative. We risked stalling.

So I asked myself:

How can I test what I want, make progress, and show it to the team in a way they can instantly understand — without slowing them down or needing them to build it first?

I used a few AI tools to build a first version myself.

What happened next surprised me. That version wasn’t ready for prod — not even close. But it gave us:

• A high-level direction that made sense

• A real discussion starter

• Insights into performance trade-offs and architecture decisions

Talking through the actual code (not specs or slides) helped the engineers spot opportunities to make it better and more scalable. It created alignment without stepping on toes.

I’m not aiming to replace engineers. I’m aiming to unlock better conversations.

This isn’t about proving I can code. It’s about showing intent, learning from the process, and helping us move faster toward valuable outcomes.

---

My Delivery Workflow (How I Build Without Blocking Anyone)

Here’s my current loop:

✈️ Step 1: v0 (by Vercel)

Write a prompt. Get a working UI layout.
I use it to generate quick, functional frontends when I don’t want to open Figma or describe wireframes. This helps me to translate what I think is valuable and get an idea of how they could use it.

Yes, this is something that I use to discuss with our designer as well. We have started to having discussions over the concept directly.

🧐 Step 2: Claude Code

Feed in the v0 output and describe logic directly as a Claude project (for coding use cases I’m not using ChatGPT). Claude turns it into clean code

✨ Step 3: Cursor

This is my main editor. It feels like pair programming with someone patient and context-aware. I ask things like:

"Refactor this." "Why is this failing?" "Make this more readable."

⚙️ Step 4: (again) Cursor

Run and test the code in-browser. Great for catching quick errors or demoing flow.

By the end of this loop, I have something I can:

• Show to engineers

• Use for alignment

• Test assumptions

The Tools (What They Do, How I Use Them, and Where They Fit)

🧠 Devin

Devin is an autonomous AI engineer that can set up repos, write code, test apps, and even plan tasks. I haven’t used it yet — it’s invite-only and more dev-focused. But the vision is intriguing.

Where it fits: Early testing by engineering teams who want to delegate full workflows to AI.

⚡ Cursor

My #1 recommendation. Cursor is a VSCode-based AI editor where you can talk to your code. It understands context and doesn’t hallucinate as much as others.

You can start wither creating the things yourself or using a prompt to build the skeleton o an App, or the entire App itself.

Why it works: It lets me work with code like a collaborator, not a gatekeeper. It’s great for quick wins, small tweaks, or getting unstuck.

🌊 Windsurf

Windsurf is like having an assistant IDE. I can ask it to run flows, handle bugs, or optimize code while I focus on something else.

I haven’t used it in a real project yet, but I explored it. It’s cleaner than some competitors and works well for more dev-heavy flows.

Where (if I ever wanted to jumped into that) I Deploy the Work

Once I get something working-ish, here’s how I ship it:

• Vercel — Great for frontends. Deploys React/Next.js apps in seconds.

• Replit — Good for backend logic, full-stack demos, and sharing live versions.

More PMs Are Starting to Build (And Talk About It)

This isn’t just me.

There’s been a huge wave of conversations lately around PMs learning to build faster, not because we should ship prod code, but because we should understand what’s possible and create better collaboration.

Try It Yourself: Your First Step

If you’re curious but haven’t tried any of these yet, here’s my challenge to you:

Try this today:

1. Download Cursor

2. Create a file called hello.js

3. Paste this prompt: *"Create a simple HTML page with a counter and a button to increment it."

4. Ask it to explain the code.

5. Change something. Break something. Ask it to fix it.

That’s how I started.

And if you get stuck? DM me. I’ll share how I use Cursor for other PM tasks too — like product analytics, sparring, and even as a design partner.

Final Thoughts

The Delivery phase isn’t just for engineers anymore.

For PMs, these tools aren’t about proving we can ship code. They’re about:

• Moving faster from idea to test

• Understanding trade-offs early

• Helping teams build better things, faster

AI is now part of our product toolkit.
Start small. Explore. Build. Learn.

Until next time 👋
— Fede