3 weeks. 1 person. Real product. Here's how AI changed what's possible.
Three weeks ago, I started building HumDono — a finance app for couples .
In just three weeks, I developed HumDono, a finance app designed for couples, while simultaneously searching for a job. Today, it's live with real users, featuring auto-parsed bank statements from HDFC, SBI, and ICICI, all while prioritizing privacy and being tax-aware for India.
As a Senior Project Manager, not a full-time engineer, I achieved this rapid development by leveraging Claude, Anthropic's AI, as my coding partner. Every architectural decision, debugging session, and inquiry about the right approach involved in-depth discussions with the AI.
Five honest observations from the experience:
AI didn't replace product thinking. The hard parts were "should I open this to NRIs?", "is the couples framing right?", "why aren't users activating?" These required human judgment, user research, and conversations with real people. AI accelerated my code, not my thinking.
Debugging is where AI shined. A user uploaded an HDFC PRIME account statement that broke my parser. Three iterations of console logs + analysis + targeted fixes — 30 minutes total. A year ago, this would have been days of digging.
PM skills became more valuable, not less. Knowing what to build, breaking it into sprints, prioritising, listening to users — these are exactly the PM skills I use professionally. AI amplified them, didn't replace them.
My first real user was my brother. He uploaded his bank statement and said "you'll see all my data, so I won't add expenses manually." Privacy fear was the real blocker — not features. I added a privacy banner that explains client-side parsing, showed him DevTools to verify, and he tried again. Lesson: user feedback > assumptions, always.
The bar for "non-developers can build software" just moved. If you've been waiting to build the thing you've been imagining — the tools exist now. Just don't skip user research. The AI can't do that part for you.
Building HumDono has made me a better PM, not a worse one. I now understand viscerally what engineers go through. I can scope better. I can have sharper conversations with technical teams.
I would love your feedback: humdonoapp.com
#BuildInPublic #IndieMaker #AItools #ProductManagement
What stood out to me is that AI compressed the execution cycle, but your biggest lessons still came from user behavior.
The parser wasn't the bottleneck. Understanding why people hesitated to trust the product was. As AI makes building faster, that kind of insight becomes an even bigger competitive advantage.
Thanks — you're right that user behavior was the harder problem.
3 weeks in: 18 signups, 0 couples formed, 0 replies to 12 personal
outreach emails. The "couples finance app" premise didn't survive
first contact with reality.
I pivoted to "personal finance, partner sharing optional" and quietly
scoped HumDono down from a startup attempt to a family tool I use
with my wife (she earns in GBP, I'm in INR — real use case now).
The AI made it feasible to build alone. Reading the data honestly was
the harder skill.
I think that's exactly where the next strategic decision begins.
Your reply made me think there's one business decision sitting underneath that pivot which becomes much more significant as HumDono evolves, but I don't think I can explain the reasoning properly in a thread without oversimplifying it.
If you're interested, what's the best email to reach you on?