What I learned trying to Weave AI into My Product Workflow (Not a Playbook)
This is not a “10x your productivity with AI” post—it’s what I’ve tried so far, what I still want to try, and what I’m learning along the way.
Hey, I’m Joshua 👋 I’ve led product at startups and public sector platforms. I’m now founding PM and head of product at a 1 year old AI startup building the last piece of software for seniors. My hope is that seniors will never have to learn how to use another new app because they know how to use our one app powered by AI-agents for all the needs in their daily life.
This post is part of the ongoing conversations we’re having in Curious Product, a small Telegram community of product folks who care more about craft than clout. If you’re curious, we jam weekly on things like “Which feature would you kill?”, “How would you validate X?”, and more here on Telegram.
Over the last few months, I’ve been experimenting with using AI tools in my day-to-day product work. Not because I want to save one click here or there, but because I’m curious about how AI can help me think better, move faster, and communicate more clearly.
This post is a reflection on what I’ve actually tried, not what definitely works. It’s more like a sketchbook than a blueprint.
1. Using AI Prototypes to Drive Alignment
One of the most helpful ways AI has shown up in my workflow is helping me validate ideas quickly—with users and with my founder.
My main workflow:
idea → AI-generated prototype
Depending on the form factor, I use:
What this helps me do:
Align with stakeholders without needing full Figma designs (I have pow wows with my founder twice a week, and speed is quite critical)
Get early feedback from users without burning too much time (I meet different elderly users about 2-3 times a week so putting a tangible prototype in their hands and watching them struggle with it is such a rich source of insights)
2. From Prototype → Figma Screens
Once the prototype gets a thumbs up, the next step is to generate design screens in Figma. Here I’ve tried two tools:
My workflow:
AI-generated prototype → mid/high-fidelity Figma screens
✨ Stitch: For generating high-quality screen designs that can be pasted into Figma with minimal editing
🎛️ Magic Patterns: For more mid-fidelity UI mockups—useful when I’m still early in the design-thinking phase
Neither is perfect. I still spend time tweaking. But it gets me unblocked faster and gives me something concrete on Figma to work off. Our company has been designer-less for about 2 months now, and I am also running point on designing screens on Figma - therefore any help I can get with Figma, I’ll take it!
What I Still Plan to Try
I’ve only scratched the surface. Here are areas I’m excited to explore next:
A. Ideation & Research
I’d love to go from a single problem statement to:
AI → Search complaints, forums, competitors → Suggest roadmap of ideas
Still experimenting, but thinking of using:
📓 NotebookLM + Gemini to handle the internet trawling and synthesis of information
🤓 Remio to collect links (e.g., reddit posts) and content (e.g., research papers) I come across about either my users (seniors) or my industry (e.g., eldercare, elder tech) and then use AI to get clear insights from all these messy info
B. Handoff: From Design → Dev Tasks
I’d love to shorten the gap between Figma and Flutter (we build for mobile).
My dream workflow:
Figma file → Flutter code → Dev tasks auto-generated on Notion (this is where we manage our sprint)
Some tools I’m eyeing:
🤖 Manus AI: reads Figma files → outputs Flutter code + task breakdown
🛠️ Figma MCP → Cursor: another potential flow to test
C. Data Analysis to Track Product Performance
Not much progress here yet, but this is a big one. If I can get AI to analyze metrics and flag issues or suggest next steps, that could be a game changer. Right now I still go through the data manually.
What I’ve Learned So Far
Don’t optimise just for the sake of it.
A one-click gain doesn’t matter if it doesn’t improve thinking, clarity, or speed of alignment.
Context matters more than tools.
Someone’s “best practice” might work for them, but it might be a total mismatch for you or your team. Think about your workflow, team, and product lifecycle. Then consider how each part / section / block of work can be improved!
Just start somewhere.
When I started thinking about how to improve my workflow, I had a clear idea on the phases of product management work: discovery → alignment → design → development → data. But instead of waiting to build the perfect AI-augmented pipeline, I’m just starting with one piece at a time. I’ll link them together later.
Final Thought
This AI stuff isn’t magic, but it is promising. If you’re also a PM dabbling with AI in your workflow, I’d love to hear what’s working for you (and what’s not). Maybe we can learn from each other’s sketches.
If this sparked something in you — a tension, a disagreement, a parallel experience — I’d love for you to join the conversation inside our Curious Product Telegram group. It’s a small, thoughtful corner of the internet where we reflect, challenge, and sharpen our product management craft together.
Most of our members found their way in via this newsletter. If you’re reading this, you’re already halfway there. Feel free to join the Curious product Telegram community anytime.
Thanks for reading The Curious Product Manager! Subscribe for free to receive new posts and support my work.