Ahead in the Cloud: Smashing Data Centers and Embracing Public Cloud with John Wei

Chad Watt: Welcome to Ahead in the Cloud where business leaders share what they've learned on their cloud journey. I'm Chad Watt, Infosys Knowledge Institute researcher and writer, here today with John Wei, Chief Technology Officer at Comerica Bank, a top 25 US bank. John joined Comerica in 2021 to smash data centers and bring the bank into the public cloud. John, how many data centers have you closed now?

John Wei: Well so far we've closed two.

Chad Watt: And how many more do you have to go?

John Wei: One more to go.

Chad Watt: One more to go. And how does amateur hockey play a role in closing your data centers?

John Wei: Well there's a funny story of that. And as you can imagine, closing a data center is not a trivial task, especially when you have had the data center for, you know, almost 30 years and plus, right? And in the middle of the migration activities, and we have chosen to move much of the workload into the public cloud, and there were difficult days. And there were days when we walked through the data center with dreams, wouldn't it be wonderful that one day this data center would be completely empty and we would be playing ice hockey. And fast forward toward the end of the year and the data center did entirely become empty and we're able to organize a tour for the entire business to play ice hockey in the data center. And more importantly we actually organized the smash party.

The smash party in the spirit of bringing the business technology team together and we actually saved some of the servers, took the battery out, storage out so it's no data, no fire issue, we got to smash some of the equipment, still responsibly dispose of them of course, but really the spirit of that playing the hockey, have a smash party is really the idea that we can achieve things above and beyond the daily routine. We can take bold actions and in our experience, shutting down a data center requires bold action and ambition.

Chad Watt: So this whole shutting down a data center, how many times did you hear someone say, "Oh we can't do that. We still need to have A, B, C." Did you hear that a lot?

John Wei: I would say almost every week. At most I probably will say every day. Because especially part of that is just habit. The truth of the matter is being a bank, by the time you get into the details, there are regulatory controls, there are security controls, access controls, there's also series of procedures we need to follow, more than 200 controls we have to follow. So it is a significant work effort. But as we have demonstrated over the past 18 months, once you set a goal and you're able to mobilize business and technology together and really understanding the why, it is a mission possible. I think the hardest part is getting the organization mobilized and marched toward a shared mission.

Chad Watt: Why is Comerica smashing data centers and moving to public cloud in 2022?

John Wei: I think there are two answer to this question. So there's a business answer to this, right? One is our CEO actually put in 2019 annual report where deploying cloud technology to become faster, agile, more resilient, and it was greater effectiveness and efficiency, but then there is more digital technology to this. If you look at the bank industry, we are living three digital waves all the same time. Let me elaborate. The digital one really is about access. It's your, you know, mobile banking, web banking. Conceivably if you don't have a large, fast growing customer base, you probably could put that in your traditional data center.

But the next wave which we're living in the middle of it now is this notion of unlocking the enterprise capabilities we have developed for many commercial customers. Turning their black box into a gray box, then turning that into an open box, moving toward an embedded banking. All of a sudden you're really dealing with a large ecosystem of partners. That is extremely difficult to do with a traditional non-composable architecture because you have to bring a lot of experiences, services, composed, you know, workflow setting to deliver that simple and convenient experience.

And the same with the digital 3.0 which is essentially AI machine learning abled to create that intelligence and predictive guidance to the customers and to automate intelligently all these capabilities requires huge amount of data enablement. The speed of data access is simply not possible economically with a traditional architecture.

Chad Watt: You talk about embedded banking. Is this something you guys have in practice now? Can you give me an example of embedded banking in action at Comerica?

John Wei: Yeah. So there are some details I cannot really disclose but I can give you some what we do, right? We have some commercial automotive client and if you think about the way how they pay, the dealers and the dealers pay the suppliers, it's a huge volume of payments that happens across the board, right? The embedded banking, if you think about how commercial banks it works, is they go through a specialized process dealing with checks or bank being a place you go to, to do it. We really want our banking to be deeply embedded into the customer workflow, right? What causes a payment to take place between the dealer. Well dealer sold a vehicle, is ready to pay back to the bank and ready to pay back to the OEMs, or the OEM is fine on seeing that the dealers building some special marketing programs. And all these activities actually happen inside of the OEM, inside of the dealer space.

So all this workflows happens today making those capability available so the API so that they can call upon those services and have it executed so they can actually focus on real business, which is the interaction with the dealers is what really provides value. And then frankly, this also makes the banking much more stickier, right? Because instead of, you know, kind of the single threaded interface into the bank, now we are deeply embedded way how the business is executed.

Chad Watt: Right. And that first horizon kind of context, those are all discrete transactions that take time and you wait, and you check, and you clear, and you balance. But sounds like this is much more at least to that third horizon about AI, automation, and I mean frankly speed. Talk a little bit more about that for me.

John Wei: When it comes to risk and think about what bank does. Bank really does three things across all the lines of business. One is how you deposit money into the bank. Second is payment. And third is giving out loans on the deposit side. Knowing customers you want to reach out to, because you don't really want to send out a lot of spammers and they're not effective. Knowing the family relationships, knowing a child is going to college and beginning that financial services journey and having targeted, you know, messaging and multichannel outreach. And then that really creates a much high [inaudible 00:07:53] willingness to go to maybe online bank to three, four clicks and they have account being opened. And our ability to really predict who are the likely customers but also instead of having people fill out this crazy long forms. For us to really understand all the data, you know, if this child is living with the parents, well there's a fair amount of information we can already collect and prepopulate, right? So that's under account side.

On the payment side, there is a significant fraud that are insured by the banks, right? And if you think about a fraud case, and we always think about fraud being somebody, you know, very sophisticated and doing a lot of, you know, unexpected things. But sometimes somebody comes in, it's obviously asking the wrong question, but their whole goal is this one called to figure out who the customers or the payers name is. Next call to figure out the address, next call is figure out maybe the account number and so on, so forth, right? And then think about in the human sense, why would agents not catch it? Because in retrospect we listen to many of the phone calls, it's obvious these are fraud.

Well it turned out we have this bias to demonstrate empathy to our customers. It's really hard when you listen to somebody that is not speaking logically, perfectly logically and the agent is wondering like is this person just because of medical conditions is struggling and demonstrate high degree of empathy or this is a real intentional fraud going on. It turned out through AI machine learning, we can actually bring quite a bit about behavioral science into the conversation and being objective, being- alerting the agents of the higher probability of fraud. And so that we still can honor our tradition of valuing our customers but still reduce the fraud significantly.

Chad Watt: So some instance where you're using the AI to basically help the agent do a better job of being an agent.

John Wei: Exactly. Exactly. Right. And I think the third element is giving up the loans. So thinking about giving up the loans, what happens, right? It's underwriting risk and usually there's just tremendous amount of information, uh, historical behaviors and a lot of contributing factors, lots of documents frankly. And it's not uncommon a business comes to us and says, "I need a $5 million loan to import X, Y, Z from overseas." Going through the bill of landing the materials and knowing that the money indeed was spent for that purposes, historically it's a huge manual effort. Today, AI machine learning, we can literally take a PDF, drop through AI machine learning and turn it on the other end through natural language processing a fairly good understanding of what goes on that ship and what the risk we're taking on. So there are many, many creative use of AI machine learning to make it much more intelligence and going actually beyond just prediction.

Chad Watt: I wanna come back to a point you made earlier. Your investors were told in 2019 that cloud was gonna help Comerica get better at being Comerica. Let's do some calibration here. How much at that time did Comerica really rely on the cloud?

John Wei: We maybe enrolled in cloud at that point but majority of our workload were still in our traditional data centers. So cloud is something we have a story for, but it's not where our primary workload is being hosted on.

Chad Watt: So is their goal to make that the case down the road? Or is that your end goal? Part of the answer to the why.

John Wei: I'm actually pleased to say that we're not no longer describing the intentions. Today majority of our business workload already sits in the public cloud. So I think it's useful to define what public cloud means, right? Number one it certainly means not private cloud because I think if you go back to the definition what Gartner laid out, what cloud is, it's really about leveraging the internet, shared infrastructure, security, easy access, but more importantly an economic model where you can pay by the drink, if you will, right? Private cloud simply does not really deliver that. So the way I would describe it, private cloud, this may be a little bit controversial, private cloud is almost like I bought a better power generator at home, now I'm claiming I'm on the public grid. It's maybe the generator's a little bit quieter so I don't hear it, but the fundamental economics is not the same as being hooked up to a public cloud.

Chad Watt: Kind of that first wave, first generation cloud kind of situation.

John Wei: Exactly. So one advantage being a late mover is we could frankly leapfrog our competitors by awarding some of those private cloud type of technology prototypes we are able to look at their experiences and frankly avoid some of the mistakes. So I give you concrete examples. Mistakes are only doing cloud migration lift and shift and as opposed to taking a modernization approach. Mistakes are focusing purely on the infrastructure and not looking at the modernization from the lens of the business. Meaning number one I can describe how we're actually doing the cloud modernization today.

The question really starts with why do we need this application. And that's a serious question. It's not a check box to move to second. It's a serious real conversation. The second question is do we have a replacement software as a service. So I give you an example how serious it is. Our general ledger for the bank, our ERP system, we have migrated that to the Workday general ledger. And there's a lot of people that use Workday for functions. We actually successfully migrated that into Workday for general banking, general ledger. And then frankly, you know, if you look at the entire cutover process there's literally a thousand task that we go through. It's been amazingly smooth.

Chad Watt: That's really telling because just back to your point about the nature of banks, the experience has to be 10 out of 10, 100 out of 100, and available whenever that one customer comes in because one bad experience can cost you that customer. And so, between being typically just kind of conservative as it comes to change, banks also, you don't take a system out if it's working. Talk to me a little bit about that, and from a culture perspective, why not just lift and shift? I mean this is a great system. I can just lift and shift and we won't miss a beat and it'll be running on Monday morning. Why should I have to rethink this, sir? Convince me otherwise.

John Wei: That's really a great question. Here's the curve kind of a mental model that I have in my mind. Think about a diagram and horizontally it's technology and vertically is the value, right? In these days if you have zero technology, it's hard to generate any value. If you have too much technology, definitely you have too much complexity, the value will drop eventually to zero because all your time is spent on just managing that complexity. So the relationship between the value and technology has to be a bell curve. And here's the trick of the bell curve, because it's impossible to find where the pinnacle is, right? Where the kind of peak is. What you really have to do is play on both sides. So really as we innovate to bring the digital capabilities into the banking environment, it's easy to add a complexity to the environment. Which means in order to pay for that added complexity, you have to be much more intentional in the complexity out of the environment.

The challenge of the traditional lift shift approach is it kinda shifts the problem but does not really fundamentally increase the resiliency and it keeps the same complexity of the operating environment. And for many banks, if you are kind of taking the lift and shift, you have cloud, you have on-premise, you're not shutting down the data centers. The overall footprint is actually becoming more complex. Now on top of that you're adding the digital capabilities and no wonder, you know, there is a constant struggle with the budget and a constant struggle of spending the ever, you know, slightly increased budget and a much, much larger IT footprint and still be able to take care of the platform.

Chad Watt: Okay, John. Let me, uh, take you through a lightning round. A few questions about your own behavior here. In your wallet right now, do you have more plastic or more paper?

John Wei: Oh definitely more plastic. I hardly have any dollars left, physical dollars left. (laughs).

Chad Watt: Do you reconcile your bank statement each month?

John Wei: Uh, I don't. I do trust the bank. But I have to say, in my younger ages, I used to do budgeting. I have developed that intuition, so today I'm more informed by intuition than looking at a physical statement.

Chad Watt: And your personal data, John Wei. Do you keep your personal data on-premises, in a private cloud, or in a public cloud?

John Wei: It is in a public cloud and I wouldn't disclose which one. (laughs).

Chad Watt: (laughs). Fair enough. Fair enough. Got it. I really appreciate. What's the banking app you use the most?

John Wei: Well of course Comerica's mobile banking and we actually launched a couple mobile banks. I have to say I'm very pleased to say it's 4.5 so one of the highest. I'm those people that actually use both the web banking on the mobile device and the mobile bank roughly equal 50/50%.

Chad Watt: You know, as a CTO you could probably find some other beta testers to do that for you, I think.

John Wei: (laughs). That is true.

Chad Watt: John, thank you very much for your time today. And thank you for your insights.

John Wei: Well Chad this has been very enjoyable. Truly been a privilege and I enjoyed listening to your questions and really made me think, so thank you.

Chad Watt: This podcast is part of our collaboration with MIT Tech Review and partnership with Infosys Cobalt. Visit our content hub on technologyreview.com to learn more about how businesses across the globe are moving from cloud chaos to cloud clarity. Be sure to follow Ahead in the Cloud wherever you get your podcast. You can find more details in our show notes and transcripts at infosys.com/iki in our podcast section. Thanks to our producers Catherine Berdette, Christine Calhoun, and Yulia De Bari. Dode Bigley is our audio technician and I'm Chad Watt with Infosys Knowledge Institute. Until next time, keep learning, and keep sharing.