AI-Powered Real-Time Fraud Detection
Key Points
- AI is reshaping business by unlocking massive productivity gains and trillions in economic value, with IBM Z’s high‑throughput, secure, encrypted environment forming the backbone for these transformations.
- Traditional credit‑card fraud detection relies on simple rule‑based checks that miss nuanced, out‑of‑pattern behaviors because only a tiny fraction of transactions can be scored in‑line within the tight processing window.
- By embedding AI directly on the IBM Z system, organizations can evaluate every transaction in real time, identifying anomalies—like an unexpected guitar purchase—that would otherwise slip through the rules engine.
- This integrated, inline AI approach enables near‑100 % coverage of transaction scoring, dramatically improving fraud prevention while maintaining the speed and security required for high‑volume processing.
Sections
- AI-Driven Fraud Detection on IBM Z - The speaker explains how AI combined with IBM Z’s high‑throughput, secure architecture can enable real‑time credit‑card fraud scoring, overcoming the latency limits of traditional rule‑based systems.
- AI-Powered Real-Time Fraud Detection - The speaker explains how integrating AI directly into IBM Z mainframe transaction processing enables inline, 100% coverage fraud detection, eliminating false charges without offloading data to external systems.
Full Transcript
# AI-Powered Real-Time Fraud Detection **Source:** [https://www.youtube.com/watch?v=OSRXo56R5Ts](https://www.youtube.com/watch?v=OSRXo56R5Ts) **Duration:** 00:04:50 ## Summary - AI is reshaping business by unlocking massive productivity gains and trillions in economic value, with IBM Z’s high‑throughput, secure, encrypted environment forming the backbone for these transformations. - Traditional credit‑card fraud detection relies on simple rule‑based checks that miss nuanced, out‑of‑pattern behaviors because only a tiny fraction of transactions can be scored in‑line within the tight processing window. - By embedding AI directly on the IBM Z system, organizations can evaluate every transaction in real time, identifying anomalies—like an unexpected guitar purchase—that would otherwise slip through the rules engine. - This integrated, inline AI approach enables near‑100 % coverage of transaction scoring, dramatically improving fraud prevention while maintaining the speed and security required for high‑volume processing. ## Sections - [00:00:00](https://www.youtube.com/watch?v=OSRXo56R5Ts&t=0s) **AI-Driven Fraud Detection on IBM Z** - The speaker explains how AI combined with IBM Z’s high‑throughput, secure architecture can enable real‑time credit‑card fraud scoring, overcoming the latency limits of traditional rule‑based systems. - [00:03:05](https://www.youtube.com/watch?v=OSRXo56R5Ts&t=185s) **AI-Powered Real-Time Fraud Detection** - The speaker explains how integrating AI directly into IBM Z mainframe transaction processing enables inline, 100% coverage fraud detection, eliminating false charges without offloading data to external systems. ## Full Transcript
Today AI is a huge business disruptor.
It has the opportunity to boost productivity,
to unlock trillions in economic value.
We think about IBM Z and the systems we have,
we think about speed of transaction processing.
We think about the scale that IBM Z can handle
with millions of transactions processing through the system.
And we think about the secure IBM Z system
that provides pervasive encryption to ensure our data is secure and locked down.
And all of this today provides the foundation for our
high-throughput transaction processing systems.
And that's just a bunch of facts.
So let's get to a real-world example about how
AI can transform our business world today.
Let's think about fraud detection and about credit card processing.
When we think about our credit card processing today,
very few transactions are actually scored in-line
because we don't have time.
And what do I mean by “we don't have time”?
If we look at a transaction first
it has to get there, it has to get processed,
and then it has to return.
And so what I have is this limited time window
to actually get the processing done.
And in that processing window,
I need to decide if this transaction is valid or not.
And so how do I decide that?
Well, traditionally it's been a rules engine.
Let me explain.
Let's say I swipe my credit card here in North Carolina.
Yep, that's where I am, North Carolina.
I'm buying something at the store.
Everything's fine -- it goes through.
My same credit card number gets swiped in Turkey.
Yeah, that's not possible.
I can't be in North Carolina and in Turkey.
So the rules engine handles it without a problem.
But let's try a slightly more complex scenario.
I love cooking, so I have to buy food and I enjoy buying food.
Maybe not the grocery stores all the time,
but I enjoy food and working with food.
And clothing.
I enjoy buying clothing. I enjoy making clothing.
This is normal. This is my normal kind of processing.
The normal dollar amounts that I will spend.
Now, all of a sudden, a guitar shows up [as a charge].
Me? That doesn't fit me.
So that needs to not be processed.
With AI, I can do that.
I can look at the “normal” and say, this is out of norm.
But remember, I have to do this in this small time window.
How do I do it?
Well, today, the transaction processing is sitting on the IBM Z system.
And many people are sending a few transactions out to be processed on an AI system.
Okay, I can process a few.
I might catch this guitar.
But if instead, I take that
and put it all together on my Z System,
so I'm processing my transaction
and while in-line, doing the AI processing,
I can get up to 100% coverage of my transactions.
And then I'm guaranteed that guitar's not going to go through
and it's not going to be charged and be a fraudulent charge.
It allows me to do that processing in the environment,
in-line where all my data is,
where all my transactions are currently being processed.
AI is disrupting our industry.
How are you going to take advantage of it?
Bring AI to your existing workload and data on IBM Z and Linux One,
and let's make this a better place.
Thanks for watching.
For all you mainframe fans out there,
don't forget to click like and subscribe so you won't miss my next video.