I run a SaaS biz that is low transaction value / high volume. This meant when Stripe introduced their latest policy where you get charged $20 per card dispute no matter if you win or lose it, I decided to try out Stripe Radar's 30 day free trial to help me detect credit card fraud on my site.

It worked well, and I managed to prevent several fraudulent cases using it. However, once the trial period ended I was being charged £0.06 per screened transaction(!). This was becoming really expensive with many transactions.

I noticed during my free trial with Stripe Radar that 90% of the time it's quite easy to eyeball if a credit card payment is fraudulent based off of the number of blocked card attempts from the customer, and the number of cards tried - so I wondered if I could build my own version using the free information available through Stripe's API.

Stripe's docs can be confusing, so I decided to try ChatGPT with GPT-4 for the first time to help write code. I described my task, and it created some beautiful Typescript functions that take in a paymentIntent id, and determines if the payment is fraudulent taking into account failed payment attempts, number of cards, and even IP address, country of origin, and email address - all using Stripe's free API!

I've been using this code for a day and it has already picked up a couple of sketchy transactions, but completely free. It's missing things like "Early Fraud Warnings" and info from "Stripe's Network", but I'm okay with that since it's saving me £0.06/transaction.

This is the first time I've used AI in a professional coding sense for a real problem (not just for fun). I'm actually blown away, not only did it save me a bunch of time coding, but the result is saving/making me money - hence my excitement writing this!

I've been coding for 10 years, but this experience using AI felt like something totally different. It's like gaining a team of developers that you can orchestrate whilst you do less of the typing. It's made me more excited to build things than ever before.