Return to page


Building a Customer Obsessed, AI-Driven Organisation by Dan Jermyn, Chief Decision Scientist, CommBank

Watch now


Video Synopsis:


  • Dan Jermyn, Chief Decision Scientist, Commonwealth Bank

Read the Full Transcript



And been in the industry for almost 20 years in the financial space. And we have Dan German, chief data scientist responsible for machine learning that powers the customer engagement platform. So welcome, Nitty, Dan.

Dan German:

Thank you. Nitty will be joining me in a little while. I wanted to say just before I start, thank you to our friends from AT&T, I mean, incredibly inspirational to us what you've achieved. It's really incredibly impressive. AT&T everybody. We live in a real time world. Thanks also for making my KPIs for the year ahead. Twice as difficult in the space of 20 minutes. I was getting some texts from Andrew that were quite prescriptive about what we need to achieve. But I'm confident we will. It's great to be back in India. I've been here quite a few times. My first trip to Bangalore, Bengaluru. Well, I feel like it's my spiritual home in India. I've been a Royal Challengers Bangalore fan since the start of the IPL.

Mixed response, mixed response to that. As I was... Andrew said that anyone with a banking app, I got my app out. I started looking to see what I got. I got advice to stop spending so much money on thumbs up. I can't help it, I love the stuff. It's a terrific drink. Andrew's done a lot of this already, but I wanted to talk about the Commonwealth Bank of Australia and how I've been working there for six to seven years and really the culture and purpose is what keeps me incredibly excited to work every single day. Do we... We have a few people, I think from the Commonwealth Bank of Australia. Maybe you could make yourself known in the audience, stand up. So that's right. That's right. I mean, much love to our friends at Westpac, but come on.

Commonwealth Bank's Purpose

It's an incredibly purpose-driven organization. I'm so proud to work there. And in the entire time that I've been there, the steer from the leadership right down has been incredibly simple. Do good things for customers. The only thing I've ever been asked to do and so I'm going to talk today about some of the things that we've been able to achieve there. Actually, you see that building a brighter future for all, the purpose that we're driven to, but for all is important. It's a special place and we have a commitment to our customers, but also to our communities and the wider world. We also have incredible leadership. I'm very lucky to work for the best digital analytics leader in the world, Dr. Andrew McMullen. Very much the Veract Coley of the analytics world. But also Matt Common, our CEO and the board.

Inspirational leadership in terms of what AI and data can do. Capability. We've talked about incredible talent in the room here in all of our locations. But also diversity of talent, diversity of experience. I'm very proud of the number of people in our team who started their career at the CommBank, serving customers on the frontline in our call centers online. A diverse, sensible, distributed skillset, set of talents in all areas is very, very important. Probably something the RCB want to think about in terms of their bowling attack. Anyone watch the game on Monday? Interesting approach. And partnerships, I mean, goes without saying. We're here talking about what we've been able to achieve with H2O. We've been a customer for many years and the closer the partnership has gotten, the more we've been able to do together.

And it's a key part of what I'm going to talk to you about today. And Nitty's going to follow with. We have a vision to become the global leading AI bank. Incredibly committed to doing that. I'm incredibly excited about the journey that we're on to do that. One of the things that we did was create the branding of CommBank AI around that to really drive forward what it means to become an AI first bank. And actually on this slide, you can see my one key contribution to the whole endeavor, which is the logo on the right hand side. Here. You see this, I had the brilliant idea to put just .AI on a yellow. I don't know where that idea came to me from, just magic-ed it up, I guess. And our partnership with H2O has obviously been incredible in terms of what we've been able to achieve so far, but where are we going to go next?

Use Cases

And I want to talk about a few of the use cases that have been key to that in building a brighter future for all. We're always thinking about the customer experience. We have some incredible data scientists, but what does it mean? What's it going to feel like to the customer impact at the end of the day? Fraud. We heard the folks from AT&T talking about how incredibly powerful AI can be in fraud detection capability. This is within the Commbank app. We use AI modeling to try to detect where the fraud is occurring. Now, if this has ever happened to you, you may get a message that says, was this fraud? Was this not fraud? Sometimes banks will stop a transaction automatically because it looks suspicious. That's to try and keep you safe. But using AI in our Commbank app and the customer engagement engine, we're able to actually in real time look at a suspicious transaction and give the customer the choice and say either, yes, this is unusual, but this is true.

It's correct, or no, this is not correct. This is fraud. And when you do that, if it is a suspicious activity, you click straight through right away through AI, through our customer engagement engine, immediately to an agent who knows exactly why you're calling in that moment, that high stress moment. And because it's in the app, you're already pre-authenticated. We know exactly who you are and get straight to the matter. No queuing up, no press one to do this, two to do this. It's an incredibly customer-centric approach to helping our customers and one of the most high stress things that can happen. Bill Sense, just another example of how we're using AI to help our customers with their financial well being. People worried about the cost of living, worried about managing their finances day-to-day. We provide a range of tools for the Commbank app that use AI to help forecast what's going to happen and how you may wish to stay on top of your finances.

And we augment that experience through using AI curated in the app. Benefits Finder, Australia's a wonderful country. There are many different ways in which the state federal governments provide benefits, rebates to members of society. Perhaps there are allowances for childcare, education services, all sorts of rebates, hundreds in fact. But they're quite complicated to know which ones a customer is entitled to. So we decided to create something called Benefits Finder, which puts all of those various government rebates in one single place and makes it really easy for our customers to get access to money that was already there through the government through a seamless process. And AI helps us to curate that and make it simple for customers. It's something that I think was a really good example of what our bank stands for.

This was something that isn't part of our core banking service, but it was just a way to use our incredible opportunity with customers to provide a service to them that is just generally helpful and the right thing to do. Andrew talked briefly about natural disasters which is another obvious case in which AI can be incredibly helpful. Australia is a huge country and bush fires, we've heard about, as one of the examples of natural disasters. If there's a bush fire in your local community, that's an incredibly high stress environment. If you were to go to the local branch, everybody would know. You could smell the smoke in the air, but actually the branch is likely to be closed too, because obviously it's in the area of danger. So in that high stress moment, you may be calling up and speaking to somebody who's a thousand miles away.

So AI here connects our customers to our frontline staff at the end of the phone immediately knowing that this is the most important thing to talk about right now at that moment. And finally, just in terms of a sort of summary of some of the use cases that we've done to date, transaction abuse monitoring. We're going to talk about this a little bit more later when Nitty comes onto the stage. Now, this is how we're using AI to try to detect customers who are using transaction descriptions to send abusive messages. It's an incredibly powerful use case, and Nitty's going to talk about that a little later. Some of the incredible things we've been able to do lately with unstructured data analysis with our friends from H2O. And you can tell, get a sense from, those are the four or five examples.

How Services Are Expanding

We have hundreds of different ways that we communicate to our customers and offer support. And the propositions that we're providing to them in our ecosystem is getting broader, more complex, more helpful for our customers. But there are lots of them. And so you see here, just a summary of some of the ways in which the Commonwealth Bank of Australia is providing services to our retail business institutional clients, and to help make that a simple experience for our customers. Building a recommendation engine with H2O that allows us to be customer-centric in the way that we reveal those propositions to our customers and coordinate the experiences. And this is something I'm particularly proud of, the work that the team has done on this. You think about recommendation engines, they've been around for a while, and what they will typically do is try to predict and work on what customers will tell you that they want.

But often the way that a customer will tell you what they want will be dictated by the interface that you are providing to them. And what you'll find is what we really should be optimizing for is what do the customers really want? Not necessarily what has the interface allowed them to tell you that they want. And so by being able to be incredibly bespoke about the way that we use our amazing data asset, I'm really proud of the work that the team is doing to create curated experiences that we think provide a better outcome for our customers across the whole ecosystem. I'll talk about a couple of these examples because we've heard a little bit about driverless and democratization. It's something I'm really passionate about and it was obviously one of the key reasons why in the first instance, H2O was such an incredible partner for us.

These are a few more examples of some of the ways in which we try to predict what's most helpful for our customers and present within our Commbank app messaging that we think through AI is relevant to them. So you've got three examples here. One is for customers for whom we think maybe a savings account is a good tool, a good vehicle for them to help manage their financial well being. Bill Sense, which is the one that I mentioned about earlier, which is helping to use machine learning to predict customer finance and Benefits Finder, which is the tool through which we coordinate those various government benefits. Now, each of these was a conversation within our ecosystem that we had created maybe three to four years ago, using AI with tremendous results to be able to be more relevant and timely to our customers.

But we're always pushing to do more. And the more data that we get, the more bespoke we can get. So we realize that at scale, the people throughout our organization who are responsible for these assets needed to be able to plug into our data much in the same way that we've already heard about with regards to the community of practitioners having a lower bar to entry for AI. And we were able to get some of the analysts or product owners who had no data science background to start to fine tune some of the AI models that were personalizing these experiences. So in the first instance for the savings accounts, within the first two weeks of applying a new H2O model to something that already had an AI model powering it we tripled the number of incremental savings accounts that were opened as a result of the personalization targeting here.

Now there's a way to look at that that says, I've been doing a very bad job of this for the previous three years. I choose to think that the incremental success of the the citizen data scientists that we had is really just the most inspiring thing about this because of the scale that we get from the thousands of people in our organization who are committed to serving customers, being able to access this incredible capability. Bill Sense doubled the number of customers onboarded and Benefits Finder triple the number of claims started. And as of December of last year, we hit 1 billion of money returned to customer accounts as a direct result of using that Benefits Finder. Tooling an incredible response I think for Australia.

One More Use Case of Fraud

I wanted to just talk about one more use case before I hand over to Nitty, who's going to talk about some of the cooler things that we've been doing recently. We've heard about fraud, obviously a classic use case for AI. Some of it was very subtle, but you may have noticed earlier in Sri's introduction that there was a slight hallucination from the GPT around the biggest bank in Australia. Very subtle. I think most people didn't notice. Westpac. Actually Sri is and always has been a visionary genius. And that was a deliberate mistake because what that point highlighted was the importance of being able to, in a large language model world, where generative AI is going to become increasingly important to be able to train that and weak it to individual circumstances across boundaries. Now Australia, wonderful country that it is, generally is importing historically over a large period of time, software packages, services from overseas large tech vendors and our fraud system's no different.

We have a number of people doing a brilliant job trying to protect our customers from fraud. But the off-the-shelf packaged models that we were using that are provided by vendors were simply not fit for purpose for the Australian market. I don't know if anybody has ever banked in Australia versus India versus the US. There's very, very different patterns of behavior. And therefore, the need to have models that bespoke to that particular geography is essential. And so using H2O, we're able to create a really tailored version of a fraud model for the Australian market and our very first attempt at this incredibly successful in terms of being more bespoke and accurate. 35% uplift in the first one in terms of card not present fraud detection. And that's why as we partner with H2O and think about the ecosystem that we're building with the incredible data asset that we have, our ability to create tailored AI products and services for our customers that are bespoke to the incredible data asset that we have become a massive impact enabler for not just the Commonwealth Bank of Australia, but for Australia at large.

And now to talk a little bit more about some of the things that we're unlocking and seeing right at the cutting edge in terms of unstructured data, I'm going to hand you over to my brilliant colleague, Nitty Sinah. Please welcome her.