Attune

A case study in product development with AI tools

I spent 15 years building education products at Adobe and Skillshare. I’ve led teams of engineers, designers, and PMs. I managed roadmaps and sprints, A/B tests, and rollout plans.

I’ve been reading so much about the new role of a product manager in the age of AI. I wanted to see it for myself and internalize the new normal. So, I set out to ship a product and get real users: just me, a set of AI tools, and 15 years of product instincts.

Here's what I learned — and why it matters.

THE PROBLEM I WANTED TO SOLVE

In August 2024, I attended a retreat convened by a small organization that helps people reconnect with purpose and meaning in their work. The program was powerful. We sat in small groups, reflected on guided prompts about struggle, hope, imagination, and wonder, and then shared what we'd written. The experience created a real connection and shared understanding. That connection is needed more than ever as people feel lonely, caught in the polarization of our society, and struggle to pay attention to the impact it’s all having on their emotional well-being.

But the retreat experience doesn't scale. You need a facilitator, a physical room, and everyone's calendar to align. I left wondering whether there was a solution to bring an experience like this to many more people.

So I decided to build one.

THE BUILD

I started the way I'd start any product: with a concept brief and competitive analysis. I mapped the landscape — Calm, Headspace, Day One, dozens of journaling apps — and identified the white space. Plenty of solo mindfulness tools. Plenty of journaling apps. Nothing that combined structured reflection prompts with community — the ability to read what others in your small group wrote in response to the same prompt, anonymously, or not.

That community layer was the insight. Reflection without witness is a diary. Reflection shared with others becomes a connection point.

From there, I scoped an MVP. Authentication, a group architecture, a prompt database, text reflections, email reminders, and basic analytics. I designed the full experience in Figma to be mobile-first with a simple, clean UI. I used ChatGPT for strategic thinking: refining the value prop, drafting email copy, developing a brand kit, and working through the information architecture.

Then I set out to build it. I tried a few different AI tools. I landed on Replit, and stood up a full-stack web app: user auth, a CMS for managing groups and prompts, a reflection submission and community viewing system, email integrations via Resend, a PostgreSQL database, and a basic analytics dashboard. I called it Attune.

I recruited a beta group to try out the experience. They are about to wrap up the first test.

The entire journey from concept to live product with real users took weeks, not months. Not because I cut corners on the product thinking, but because AI compressed the execution dramatically.

WHAT I LEARNED

Three things stand out:

1. Product judgment was the challenge, not code. AI can generate code, write copy, and scaffold architecture. What it can't do is decide what to build, what to cut, and what matters to the user. Every hard decision on Attune (the character limit on reflections, anonymous-only community sharing, the daily cadence) came from product instinct built over years of watching users in online communities.
2. The "full-stack PM" is real. With AI tools, a PM like me, with strong product sense and basic technical literacy, can prototype, test, and iterate at a speed that used to require a cross-functional team. I’d still need engineers to really scale this. But for 0-to-1 validation, I’ve proven the capability to do it myself. (And I really enjoyed it!)
3. The “messy middle” still matters. It was difficult to find a beta group, and then they didn't engage as much as I'd hoped. The quality of the prompts (in generating high-quality reflections) mattered more than the UX. They didn’t want their reflections to be anonymous. These are classic product problems — activation, engagement loops, habit formation — and AI didn't solve them for me. Product-market fit still requires the slow, unglamorous work of listening to users, adjusting, and iterating. Bonus that I was able to make adjustments, like toggling off the anonymous-only setting, in minutes, not weeks.

WHY THIS MATTERS

I’ve been feeling anxiety about whether AI will replace me (maybe you have too). After building Attune, I'm convinced it won’t (ask me again next month 😜). But the ability to use AI tools in product development is a must-have.

The product leaders who will thrive in this era are the ones who can do what I did with Attune: combine deep product craft — user research, strategic framing, competitive positioning, MVP scoping — with the ability to move from idea to live product using AI as a multiplier.

For me, Attune was also a personal experiment. After years of leading teams at Adobe and Skillshare, I wanted to know: can a senior product leader, working solo, use AI tools to go from insight to shipped product? The answer is yes. And the skills that made it work weren't the AI skills. They were the product skills I'd spent a career developing.

I'm continuing to iterate on the concept and exploring new use cases for the same core idea — small groups, shared reflection, human connection. I’m seeing the power of Claude Code in real-time. If you're building at the intersection of AI and community, or rethinking how product leaders work in this new era, I'd love to hear from you.