for a long time, my workflow was standard product management: i have an idea, i write a ticket, it gets assigned to a developer, and eventually, it goes live. that cycle works, but it has friction.

recently, that changed. instead of just writing requirements, i started writing software. using tools like cursor, claude, and chatgpt, i went from "having ideas" to implementing them myself. in one week, i built bots that solved real operational problems.

here is how i replaced the "junior dev" gap with ai, and what i built.

the stack

i use chatgpt and claude for planning, documentation, and high-level architecture. they help me turn vague requirements into clear technical specs.

for the actual code, i use cursor. it's the engine. i can talk to it in plain english, paste errors when things break, and iterate instantly. i don't need to memorize syntax; i just need to understand the logic.

1. the slack bot: protecting assets

working in car rental, we have expensive physical assets moving around. we had an incident where a vehicle was driven thousands of miles away—from california to the east coast—and we didn't catch it in time.

i realized most problems are just database queries waiting to happen. i built a slack bot that runs a cron job every hour. it checks our database, compares vehicle locations to their home base, and alerts the team if a car is more than 100 miles away.

what would have been a $5k-10k risk is now an automated alert. it took me a few days to build, not weeks of sprint planning.

2. the toll dashboard: killing manual work

we used to spend hours reconciling toll bills. we'd get a list of charges and have to manually match timestamps and license plates to bookings to charge the right user.

i built an internal dashboard where the team simply uploads the csv from the toll authority. the script parses it, matches it against our database records, and identifies exactly who was driving. weeks of manual data entry turned into a 30-second drag-and-drop task.

the new pm

i'm not a senior engineer. but with ai, i can operate like a mid-level developer for specific tasks. i can build, test, and deploy tools that would otherwise sit in the backlog for months.

this doesn't replace engineering teams—it frees them. while i handle internal tools and automations, they can focus on the core product and complex architecture. but the line is blurring. as a pm, being able to build what you spec is a superpower.