Product management in practice: app ratings, user feedback, PRDs, data vs. instinct, and launching quickly without raising risk.
Guests can now order food through ChatGPT, Claude, Gemini and Google Maps. What a restaurant brand needs in place before an AI agent orders on a guest's behalf.
Rick Rubin says he's paid for his taste. In product, that's product sense: knowing what to do, when, why, and when to say no. It's what I'm hired for.
Most restaurant loyalty programs pay guests for visits they'd have made anyway. How to match points, rewards and offers to the right guests and protect margin.
How to turn a PRD into a working prototype with AI in an afternoon, why the PRD-first order matters, and what an AI prototype should and shouldn't be used for.
A guest who opens your app is betting it's faster than the line. Sign-up fields, buried menus and slow APIs break that bet. How we cut ordering time in half.
Restaurant app offers rarely fail because the deal is weak. They fail at redemption. Five places offers break, with fixes from Jack in the Box and Popeyes.
AI can draft a PRD in minutes, but a polished PRD built on vague thinking is worse than none. The workflow I use with Claude, and where the human judgment goes.
I rolled my eyes at agentic AI until I built an MCP for ordering fast food. Why agents cut the effort to value, and why company buy-in is the real test.
A step-by-step way to turn hundreds of app store reviews into a ranked list of product problems, using nothing but a spreadsheet and an afternoon.
Replying to app reviews helps in a few specific cases and wastes time in most. When a response is worth writing, what to say, and what moves the rating instead.
I put an MCP server on my family's chores app, and now Claude updates chores in plain conversation. What MCP is, how it works, and what it taught me.
How we took the Jack in the Box app from 2.6 to 4.8 stars: read the reviews as data, fix the jobs that fail most, then ask for ratings at the right moment.
How my consulting site gets built with Claude, ChatGPT, GitHub and Vercel, and where I show up in the flow.
When code takes minutes, your ideas speed up to match your tools, not your judgment.
AI made content cheap to produce. When volume is free, quality becomes the scarce resource, and the people who can enforce it become more valuable.
AI let me build prototypes faster than ever, until I couldn't say what problem they solved. Why clarity on the problem is the work, not overhead.
Not a prototype. Not a rough sketch. A documented, visual, working proof-of-concept in under 20 minutes.
The PRD isn't the bottleneck; the effort of writing one is. The first in a series on using AI to lower the barrier to writing good product requirements.
AI makes it tempting to skip straight to prototyping. Why the PRD isn't dead: the thinking it forces keeps fast prototypes pointed at a real problem.
Payment friction is invisible until you remove it. Then the money is obvious.
The instinct to rebuild everything is almost always wrong, and expensive. The framework I use to make the call.
One-star reviews aren't complaints. They're failed jobs.
A 2.6-star rating looks like a verdict. It's actually a diagnostic.
Guest complaints aren't an ops problem to manage. They're a precise map of where your product is failing.
Polished PRDs, diagrams and dashboards are evidence of collaboration, not a substitute for it. Why the real product work happens in the conversations.
I shipped a feature the metrics said was working, and nobody believed it. Why dashboards don't build stakeholder confidence, and what does.
Shipping fast without user feedback doesn't lower risk, it raises it. How to launch quickly while keeping customers in the loop and the bets small.
I'm an instinct-driven product manager in a metrics-driven job. How I use customer intuition to find the problem, then funnel data to prove it.