Agility at enterprise scale means building systems that not only make millions of micro-decisions, but also keep humans at the center — creating customer experiences that are timely, relevant, and respectful of the customer.
What if your data knew your customers better than your frontline employees — and used that insight to serve, not sell?
Today we are here at PegaWorld 2025 at the MGM Grand in Las Vegas, and we’re going to talk about how National Australia Bank has built what they call a “Customer Brain” — a centralized, intelligent decisioning layer that unifies engagement across service, sales, and relationship-building at scale.
To help me explore this, I’d like to welcome Christian Nelissen, Chief Data and Analytics Officer at NAB.
Resources
National Australia Bank: https://www.nab.com/au https://www.nab.com/au
The Agile Brand podcast is brought to you by TEKsystems. Learn more here: https://www.teksystems.com/versionnextnow
Catch the future of e-commerce at eTail Boston, August 11-14, 2025. Register now: https://bit.ly/etailboston and use code PARTNER20 for 20% off for retailers and brands
Online Scrum Master Summit is happening June 17-19. This 3-day virtual event is open for registration. Visit http://www.osms25.com and get a 25% discount off Premium All-Access Passes with the code osms25agilebrand
Don’t Miss MAICON 2025, October 14-16 in Cleveland – the event bringing together the brights minds and leading voices in AI. Use Code AGILE150 for $150 off registration. Go here to register: https://bit.ly/agile150
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://www.theagilebrand.show
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
Greg Kihlstrom (00:00)
Agility at enterprise scale means building systems that not only make millions of micro decisions, but also keeps humans at the center, creating experiences that are timely, relevant, and respectful of the customer. What if your data knew your customers better than your frontline employees and use that insight to serve, not sell? Today, we’re here at PegaWorld 2025 at the MGM Grand in Las Vegas, and we’re going to talk about how National Australia Bank has built what they call a customer brain.
centralized, intelligent, decisioning layer that unifies engagement across service, sales, and relationship building at scale. To help me explore this, I’d like to welcome Christian Nelissen, Chief Data and Analytics Officer at NAB. Christian, welcome to the show. Yeah, looking forward to talking about this with you. Before we dive in though, why don’t you give us a little background on your role at NAB.
Christian Nelissen (00:43)
Thanks very much. Thanks for having me.
So, well, as you said, I’m the Chief Data and Analytics Officer. And what that means in reality is I always say I do four things. So I have two bosses, which means that’s because I need a lot of supervision. But for the bank’s CIO, I run all the tech for our data estate. And also, I’m responsible for all the risk-related activities around data. So I have the Chief Privacy Officer, the head of the data risk management function. We have records retention risk. So I do all of that for the bank.
And then for the chief operating officer, I run all the analytics for the bank. So I have all the analysts that work in the bank and that includes the work we do. I count on that all the work we do on the decisioning, pegged decisioning engine, which we call the brain. And then also I’m responsible for the bank’s gen AI, agentic AI. I think we’re gonna have to rename it soon given the way the industry is going. I run that program for the bank as well.
Greg Kihlstrom (01:40)
So a few things then. Yeah.
Christian Nelissen (01:42)
Yeah,
it keeps me busy and somewhat out of trouble.
Greg Kihlstrom (01:44)
Nice, nice. Well, yeah, so let’s dive in here. And I mentioned National Australia Bank’s customer brain in the intro. I was wondering, can you talk a little bit about that, break it down for us, and what exactly is it, and what’s it designed to do?
Christian Nelissen (01:58)
Yeah, so the brain is a peg, it’s based on a peg a decision engine and the idea is that we take everything we know about our customer and feed it into the brain and then the brain is connected to all of our channels that we see customers in. So whether it’s inbound or outbound, human or digital, we’re in every channel where we see customers. And so what it means is when we see something that we want to talk to a customer about, we can make that decision centrally.
And then in every part of the bank where we see that customer, we can have that one conversation with them. So if we think you’re interested in a home loan, we can send you a text message, send you a push notification. We can put that in a queue for the teller to talk to you about when you go into a branch. We can make it appear when you open your mobile app and we can also, when you log in online, we can also make it appear there. once we see, and we can also push it out to an agent to call you.
So if we think that you’re interested in something and we see something in the data that tells us that’s important to you, it all comes through the brain. And the idea behind the brain is every human just has one brain. The bank just has one brain that makes a decision about what’s most important to talk to a customer about.
Greg Kihlstrom (03:03)
Yeah, and so I mean that that sounds a lot of people say omni-channel, but that actually sounds omni-channel, right? And and just to give some some scale here, so you’re managing about 8 million customers, 16 channels, 300 actions, 2000 adaptive models. So, you know, how from a complexity standpoint, how do you even begin to orchestrate something at that scale?
Christian Nelissen (03:27)
Yeah, it’s an interesting challenge. think there’s a couple of dimensions to it. Firstly, I think it’s really important that you give people a way of understanding the framing for what it can do because it’s a very common use case in the bank. You want to talk to customer about something, but the ability to have that capability means you really have to help the whole bank understand what the capability is. And I’m routinely in conversations with people that say, well, you could do that. Yeah, of course. Once you have the data plugged in,
you have this decision layer and you’re connected to the channels, what you want to do with the customers is relatively straightforward. The second thing then I think is, know, how do you keep it simple is we deal with, we have a very good team and we deal with the complexity. So we try not to bring the complexity into the organization. So we’ve learned over, this isn’t my first go at this and some of the people that work with me have worked with me for a while. We’ve learned over the years how to
position it with customers, with our customers, without the people that we support in a way that’s easy for them to consume. And really what we’re talking to, we want to talk to them about what they’re trying to achieve. So what are you trying to do with your business? Where do you want to take your business? And we then help them understand, we bring the thing to them and say, hey, look, if that’s what you’re trying to do, here’s what we’d suggest would be a great set of things to do. And then we deal with a lot of the complexity and the background, but also, I think one of the reasons.
One of the reasons we like PEGA is because it helps us simplify the complexity. We’ve been at this for a long time. We’ve been great partners with them over a number of different institutions. And we can see how their tools have evolved that make it easy for us to manage some of that complexity.
Greg Kihlstrom (05:03)
Yeah, and so another, there’s channel complexity and there’s decision complexity. There’s also competing priorities, right? There’s a customer calls for a service thing, but there might be a sales opportunity, but it’s not always the right decision to try to make that sale when a customer is calling about a service thing. And then there’s other things like just customer engagement and things. How do you approach that?
trade-off and you know with more intelligent orchestration.
Christian Nelissen (05:32)
Yeah, we start with the principle and I’ve been at this for a while and so I think it’s really important that you start with the idea that you have to talk to a customer about what’s most important to a customer. And I used to have a, I don’t do this as much anymore, but I used to, in the early days of doing this, I used to talk about taking banking back to the 70s. And the idea was in the 70s, your bank knew you as an individual. And when you walked in, the branch manager didn’t try to sell you a credit card every time they saw you.
They talked about your family and what’s happening to you and, well, if your kids are going to college, maybe one of the needs that are to count, you’re like, you know, I’ve got lots of interesting stories off the back of that over the years of telling that story. But in the 70s, we knew you as an individual. And so I used to use that conversation as a way of framing what we’re trying to do both externally and internally with the organization. The fundamental principle is, should be about what’s the most important thing to the customer, not what’s most important thing to the bank. And so…
That’s where you start that conversation. so, you know, within the bank, we have targets for how many service interactions we have and how many engagement interactions we have. And then what’s left over is the sales. So we’re constantly making sure that we’re trying to constrain how much stuff we do on sales. I think that’s one dimension. The other dimension is you start to get into an interesting place once you have the capability because
you can go to a much smaller cohort and get much better response rates. So I’ve got a live example, we were just talking about it before we flew out, where we went to 20,000 customers and we got a 15 % response rate. And so you’re talking about 3000 customers. And if I had a, and that’s a trigger targeted, very small set of activities. In an old school way, I’d go to 100,000 customers to get to that same population. But now I have the capability to run at,
at scale but also economically run much smaller, much more tightly focused campaigns. In fact, I’m trying to stop the organization from calling them campaigns, so let’s call them actions. And I can run multiples of them side by side. so in that coming, eventually coming back to the point, in that world, I can get the same volume of sales by going to a lot smaller groups, many smaller number of customers and use the rest of that space to have a different conversation with our customers.
I’d have a service or an engagement conversation. Yeah.
Greg Kihlstrom (07:50)
Because the economics of that, mean, one, you’re having higher conversion rates with the audience, but also there’s frustration. There’s loss there too, right, if you get it wrong with the wrong, if you just keep trying to sell everything to everybody.
Christian Nelissen (08:04)
100%. So I used to say to people like, we’d send out, when I first started doing this years ago at the Royal Bank of Scotland, we’d send out 100,000 credit card mailers and we’d get super excited because we got 3,000 responses to it. And I say, you’ve got to remember, 97,000 people opened that letter and was somewhere between somewhat and very disappointed in what we’d sent them, right? So we managed to annoy 97 % of the people. And the interesting thing for me is always,
and we do it, we’ve done various bits of tracking on this in the past. When you go to a cohort with a more targeted, where you’ve got a better reason for why you’re having the conversation, people, if they don’t do whatever you’re asking them to do, they still understand why you’ve done that. And so their response to that, actually makes sense to them while you’re having that conversation. Whereas if I’m just getting a credit card mailer every second month, because that’s what our targeting rules are, like they’re you’re wasting my time and you’re not listening to me, you don’t understand me.
And so I think the whole, you’re in this sort of fundamental shift from campaign driven to really, and people have been talking about this forever, but really truly always on and very focused with what’s around the customer.
Greg Kihlstrom (09:14)
you
And so part of the understanding the customer and I guess with the best action right now, I was going to say campaigns, but I’m trying to use your.
Christian Nelissen (09:23)
Yeah,
yeah. We want to be friends, so it’s important that you…
Greg Kihlstrom (09:25)
Right. So for a human to understand what the best action or potential actions are, they need insights, right? So how do you look at, you everybody’s talking about bringing customer insights to the forefront. And certainly there are many ways to do that, but it’s still hard to do that at scale. So, you know, how do you look at operationalizing insights so that your teams and the other teams can
bring those opportunities and those actions to the customers.
Christian Nelissen (09:54)
We always start with triggers. And because actually customers do and share with us a whole bunch of information that you can either push into a propensity model and have it of blended away, or you can choose to respond to the actual trigger and the thing that the customer’s doing that is relevant. so, it turns out that when you really put your mind to it, there’s all these really interesting things that you can respond to and that are valuable to the customer. And so in Australia,
One of the things that you see customers do when they’re ready to, when they’re starting to think about borrowing money is they’ll often download all their credit, all their statements, because whoever they’re dealing with has asked for all their statements. And that’s a signal to us that they’re looking for credit and they’re probably not talking to us about it. That’s a really good trigger. If I see, we can see customers online using our mortgage calculator, that’s a really good signal. Now it’s not a perfect signal and if you’re under 18,
you’re probably not the sort of person we want to talk to. So we do some filtering and stuff like that. But it’s a really good trigger for what we want to talk to you about. Those sorts of triggers, actually, there’s a lot more of them when you start to really think about what happens in the data that you can start to respond to. And our primary focus is always, let’s figure out, obviously, let’s hoover up all the triggers we can, and then we’ll fill the rest with propensity models.
you keep, once you start on this trigger journey, once you really start getting people thinking, okay, if I see that, I should do that. There’s a lot more triggers than you ever think you’re gonna have, than you ever, can imagine to start with. And it’s a great place to start. And then the rest of it sorts itself.
Greg Kihlstrom (11:25)
Well,
those are leading indicators too, as opposed to trying to figure out things after the fight. There’s all kinds of win-wins there.
Christian Nelissen (11:34)
And it extends, if you’ll indulge me, it extends to all sorts of other, this trigger based idea extends to all sorts of other things. So at RBS, we found that half a million times a year, customers would do everything they needed to do to take cash out of an ATM, except they’d forget to take the cash. And I’ve told this story quite a few times, and I would say the audience neatly divides into two completely separate groups. People can’t believe that they could
anyone could ever do that and pick the people who’ve done it. Yep, yep.
Greg Kihlstrom (12:03)
I’m not going to say, maybe I’m in the latter group, but yeah, it’s okay.
Christian Nelissen (12:06)
But the cool idea is with this sort of setup and with what you have in this sort of environment is that’s a trigger, right? Like if we see that data, because what happens is the machine eventually swallows the cash back up and it deposits back into your account. The joke is always, because we’re in a UK bank, maybe not so much in London, that someone would come along and take the money, but everywhere else the money would be. And so the ability to take that piece of data and then in an ideal world, as you’re walking away from the ATM, you’re getting a text from the, and you look at your phone and it’s your bank and your bank says, hey,
you forgot to take your cash, but don’t worry, it’s back in your account, you don’t need to call us, but here’s the number that you need. It’s just one of those moments. And it’s a trigger. And it’s super relevant, and the customers immediately understand why you’ve done it. And even if they’re not worried about it, or they’ve already forgotten about it, it’s just a really good service moment. So those sorts of things. I think the ability to take those signals and do something with it is so much more useful than…
you at some point you start to have to think about propensity models and those sorts of things, but that’s way down the track.
Greg Kihlstrom (13:04)
Well, and with so many, I mean, we talked about several different scenarios here just now. I’m sure there’s countless others, you know, with so many data points, know, propensity models, adaptive models, all of these things at play. How do you think about transparency in terms of, you know, internally understanding what’s going on so you can do a better job and optimize, but also, you know, the trust factor with your customers? You know, how do you ensure that?
Christian Nelissen (13:30)
So we have a rule, which I think I stole from somebody who no longer uses it, which is don’t be creepy. So we want it, we want it, because again, the interesting thing about a mantra like taking back to the 70s as a baseline is you wouldn’t do things that were creepy if you were a branch manager or a staff member in the 70s. So that’s a test, right? Does this feel normal, natural to us? Can we explain why we’re not?
We’re not chasing you around the net with stuff, we’re being, we’re relevant to you. And when we understand stuff, it’s obvious that we understand stuff and we’re trying to be connected to you. say to the guys that work for me, the ability to use our customers data to help them based is based entirely on the trust that they have, that they trust us with the data in a way that they in most, they don’t unnecessarily trust Facebook or any of those other things. And we would never want to do anything to breed to that trust. That’s much more valuable than any given sale or any given.
any given opportunity. We’re somewhat unique and I’ve done this job in banks, this sort of working in my way at this level four times in different banks. The one thing I’m really super proud of where we are now is we have a data ethics policy. And that isn’t just a policy that says blah, blah, blah. It’s actually operationalized. before you know, before you do some, there’s a set of decision criteria when you want to do something new, there’s a data ethics assessment.
And the thing has to pass the ethics assessment. And we have stopped a few things, including some things that are really at all with my CEO, which I won’t talk about, but that we did them for the right reasons. And they just weren’t the right thing to do for the customer. I mean, to be fair, he understood it in the end. I’m obliged to say he’s a good guy. And for my own career sake, though he is genuinely a good guy. And when you explained to him why we did it, he wasn’t, it just took him a little while to come to the same conclusion as we did. those, that ethics assessment is real.
And it’s embedded in the way the organization work and it makes decisions. actually stops things from happening.
Greg Kihlstrom (15:25)
Yeah, and even beyond the ethics thing, obviously very important on a number of, mean, from a customer trust perspective as well as any regulated industry, there’s a lot of needs there. What about just the pure data volume? How do you look at, is more data always a good thing? How do you distinguish between useful insights and maybe just noise?
Christian Nelissen (15:47)
There’s a couple of things I think at play there. Like in the background, one of the other things we’re doing is building out a data lake house. We’ve got a couple of data estates and we’ve turned off a couple of legacy environments. So we’re progressively moving to a single estate. So the first question is we’re trying to get to a place where the data is available and we call that getting data like electricity. So if you need data, you can get data. But to your specific point about how do you pick things out, there’s an interesting, again, because the dynamics have changed so much in the way that we think about these things.
the, you can now have things that run at a much lower unit cost. It doesn’t take a lot of overhead for us to keep running them. And so the cost of experimentation, the cost of failure, the sum of that conversation has changed significantly in that sense of, so the example I would use is we have this thing that runs continuously and it’s about encouraging people to tell us when they’re traveling overseas so that.
we don’t block their credit cards when they’re overseas, right? So use relatively useful service. And it doesn’t do very well when there’s not school holidays and it does super well when school holidays are coming up because it turns out a lot of people travel during school holidays, right? So, and I would say if we’d been really overly rigorous about it under the older regime, we would have said, you know, we would have, if we’d had put it in the wrong time, we would said, it’s not doing very well. We’ll kill it. We don’t want to, whatever.
But we just let it run and it turns out it just, does a really good job for peak cycles, know, three or four times. I don’t have any kids, but I’m guessing they still have school holidays. vaguely remember every school holidays when I was a kid, but you know, four times a year it does, it does really well and well enough for us to leave it in because it’s helpful to our customers. And so I’m never too worried about the, to your question. I’m not that worried about managing the proliferation because the proliferation will somewhat sort itself out. I’m not, it’s not an unreasonable, I don’t have to
pre-filter a of the ideas. At the moment, we’re in a place where every time you add something to the system, it does better. We’ll get to a point where starts to, you need to start thinning it down, but the cost of failure is dramatically different than where it was before. I haven’t done the math, but it’s an order of magnitude cheaper to put something in the system now than it used to be.
Greg Kihlstrom (18:01)
Yeah, that makes sense. Well, Christian, thanks so much for joining. Two questions for you here. First, we’re here at PegaWorld. We’re about a day in. What’s either been something you’ve enjoyed most so far or something you’re looking forward to?
Christian Nelissen (18:15)
Look, the interesting thing for me because I run the bank’s Gen.ai program is actually seeing how much, Gen.ai is transforming, not just my direct interest base, but it’s really making me continue to think about where we’re going with Gen.ai or agentic AI now that is coming in. every conversation, I mean, I’m making more and more connections with what I need to do when I go back. And I think that’s been super interesting. I’m in a sort of fortunate position of…
Peg has asked me to meet with a few customers and stuff like that. And I just love, I love those conversations and playing backwards and forwards with where they’re at and what they’ve learned and sharing some of the stuff that we’ve gained.
Greg Kihlstrom (18:50)
Well, last question for you. like to ask everybody on the show. What do do to stay agile in your role and how do you find a way to do it consistently?
Christian Nelissen (18:58)
I say to my team, I’m someone who’s got a person with very fixed convictions that’s open to learning. So I think that it’s important once you decide to do something to really to get it done. And I think there’s, you know, in a large organization that makes a huge difference is if you can execute. But I read a lot. I read an enormous amount. The lucky thing is I don’t have any kids, so I have an enormous amount of spare time as it turns out. But I read a lot.
And I’m constantly making connections in terms of thinking about how this changes with that. so, and because I’m lucky enough in my job to be able to take that back and say, okay, we were doing this, we need to build on that and do that now and keep moving.
