Joshua Poehling

speaker

90 appearances 1 recordings 1 series first heard Jul 2026 last heard 15 Jul

Joshua Poehling’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 1 in all, peaking in Jul 2026 with 1.

Appearances

newest first · ▶ plays the moment

FEAR & GREED | Business News · Q+A: The $300,000 attack that changed this company's fraud strategy · 15 Jul 2026 podcast

It was okay, but I couldn't say it was a significant step up on the previous approach.

And then in September that year, that's when we had the big fraud attack.

And during that, we had this sort of aha moment and we switched to this AI-based approach, which is what we've been using now, which has been very successful.

So I put together this page that lists just in chronological order, all the user behaviors that could relate to fraud.

So all the user logins and what IP address they logged in and what country that was from and when they changed their password and when they tried to purchase and whether it failed or not and things like that.

And when you see all that stuff in one chronological list with a human eye, you can often very easily say, ah,

That's fraud or I can tell that's genuine.

And I was looking at that page trying to think, okay, what should our point system be?

Describing the point system that we spoke about earlier.

What should the point system be?

How do we get the right mix of points for each behavior?

And I just copied and pasted the whole text of the page into ChatGPT and just said, hey, does this look like fraud?

Literally just that, nothing else.

And it was amazing.

It just knew all the stuff that I had already learned.

It already knew.

And so I tested that on another 10 orders and gave it a bit more information about our business and the type of fraudulent behaviors we see versus what is typically legitimate for us and so on.

and it was consistently really good out of the box.

That's our new approach.

We say, take that list of behaviors which we were previously trying to generate a points-based score on and just send them all as texts to OpenAI and say, hey, do an analysis on this and come back with a fraud risk out of five.

Showing 41–60 of 90 · page 3 of 5