Main content

Deutsche Bank doubles down on generative AI after laying foundations with Google Cloud

Derek du Preez Profile picture for user ddpreez September 5, 2023
What started as a project to build a ‘one cloud platform’ with Google Cloud has resulted in Deutsche Bank’s management team getting very excited about the prospect of generative AI.

An image of a poster in front of two large buildings showing someone using a Deutsche Bank app
(Image sourced via Deutsche Bank)

Germany-based Deutsche Bank, a world-renowned investment bank and financial services company, has been on a multi-year journey with Google Cloud to build a new technology foundation for its business in the cloud. The bank’s work with Google Cloud was fortuitous, according to Deutsche’s Chief Innovation Officer Gil Perez, as the organization is now able to double down on generative AI and target use cases that have low regulatory compliance requirements, but need high levels of manual intervention. 

diginomica spoke with Perez at Google Cloud’s recent user even in San Francisco, where he explained that Deutsche Bank’s ‘One Technology, Data and Innovation (TDI)’ strategy began back in 2019, when the organization wanted to consolidate its technology infrastructure - bringing it out of the control of individual business units, creating one company-wide platform. So, instead of having corporate banking, investment banking, private banking all running their own individual technology estates, Deutsche Bank would instead have a centralized technology platform (hence ‘One TDI’). 

However, Perez said that simply doing that was the equivalent to “just moving boxes”. Instead, Deutsche wanted a more transformative approach by moving the company’s technology infrastructure and services - everything except end-user computing (which is staying with Microsoft) and third party applications (such as Salesforce and Workday) - to the cloud. Perez said: 

That galvanized everybody, for the first time, to work across the organization end to end. 

After running an RFP, Perez said that the company decided to enter a partnership with Google Cloud, for a number of reasons. Deutsche Bank felt that it could handle most of the technology requirements or challenges, but that it needed a partner that would be understanding of the regulatory pressures facing such a financial institution. It wanted a partner that would be flexible. 

Perez said that AWS basically hands over a set of APIs and tells you to get on with it, whilst Microsoft adopts a “somewhere in the middle” approach. Whereas Google Cloud was willing to be flexible and adapt. Perez added: 

Google knew that they needed to adapt. They know that we have requirements to meet in our regulated industry and they said that they would work with us to adapt their environment in order to meet our requirements. 

It's a 10 and a half year deal, so that was the most important thing. From a technology perspective, we would have managed to do well with any one of them, but I think it's the partnership, it's the adaptability, it's the willingness to take on that view of collaborating with customers - as opposed to ‘here's a one size fits all’. That was the key element in making our decision.

Regulatory challenges

This requirement for a flexible partnership was ultimately driven by Deutsche Bank’s intense regulatory environment. For instance, Perez explained that whilst some organizations may have a requirement to host their data in a specific region, financial organizations such as Deutsche are required to be in a certain region, but also have their data centers regularly audited. And that’s just one example, there are of course other requirements that are focused on documentation and internal auditing too, which the Bank has to give feedback to some or all of its 46 different regulators. 

This means that even when a vendor like Google Cloud makes a new product generally available, it can be six months before Deutsche Bank can use it because of all the box ticking required. And even then, there’s more work. Perez said: 

When a product comes out, for us, in many cases it's a non-starter, because we need all of these additional elements before we can put it in our data center. 

And after we're allowed to use it, we now have to go through our process of onboarding it and verifying that all of the evidence is actually accurate…and then we can actually deploy it in a manner that is consistent with the way the regulator has approved it. 

One of the projects that Deutsche Bank is working on may actually automate some of this regulatory reporting in the future - particularly around observability. For instance, if something goes wrong internally, there’s currently a process for recording that, fixing it and reporting the fix with documentation to the regulators. Perez explained:

Assuming something is not okay, then we need adaptability and automation. If there's a problem, we want to be able to quickly change it. Then the third thing is the evidence, In a regular organization, you move on. But no, we have to evidence everything and to communicate it. There's an entire additional step of paperwork and documentation at the station. 

Things take so long because you have so many steps, with humans in the loop. And what we're trying to figure out is how do we automate everything as much as we can. And so, the perfect situation - which we’re not there yet -  will hopefully be that the evidence will go directly to the regulators. 

This will be real time, a dashboard where you can see an incident happen, and they can see the remediation, what happened, and the steps taken with the evidence. That's the Nirvana, what we're trying to do on the cloud - to construct it in that kind of stack. 

The Next Big Thing 

Although much of the work with Google Cloud back in 2020 was still a work in progress, laying the foundations for a future cloud platform, Perez said that he was already asking the organization: ‘What’s the Next Big Thing?’. This undoubtedly raised some eyebrows, but Perez already had his sights set on future possibilities beyond cloud. 

Perez was looking to bring in another partner and decided to engage with NVIDIA, which proved to be a serendipitous move given the work now that NVIDIA is doing in AI and in partnership with Google Cloud. NVIDIA CEO Jensen Huang played a pivotal role in the work with Deutsche Bank and was focused - in the middle of 2021 (pre ChatGPT’s public release) on AI and ML. Perez said: 

We looked at who we thought had synergy, and I guess we picked right. And we then worked for about six months with NVIDIA. In early 2022 we brought Jensen to the Supervisory Board of Deutsche Bank to talk about AI, M, generative AI, avatars. We got the entire board excited. And so in 2022 began working on these things that we're seeing right now. And then probably at the end of Q1, we got more and more of the Google folks back engaged.

2022 ended with Deutsche Bank deciding to double down on generative AI. However, whilst the management board was excited about its possibilities, it still viewed it as a ‘technology thing’, according to Perez. However, this all changed with the public release of ChatGPT. He added: 

We went to the board on the 7th December…on the 30th of November ChatGPT happens. Ever since then, the big thing that has happened is that for the first time business leaders want to transform their business. We were fortunate enough to do a lot of the groundwork, prep and priming of the pump to seize the moment. But the biggest thing that has happened, it's not the technology, it's that it has transcended the technology community. 

Perez is now focused on capturing this excitement amongst Deutsche’s business leaders. He added:

This is phenomenal. It’s never happened before. We need to somehow bottle this energy, I don't know how long it will last. That's why we have a sense of urgency, and we're fast tracking use cases - we have about 25 of them. 

Of these 25 use cases, most of them are focused on areas where Deutsche Bank can make quick progress, whilst not facing huge regulatory hurdles. He explained: 

I'll give you a couple of examples. First of all, software development, why? There’s no data privacy issue with client data and all of that. So the use cases we're now targeting are ones that don't have the regulated challenges. 

I’ll give you another one. We call it adverse media. For instance, all of a sudden there's some news about you that you got caught in some bribe with Colombia or something like that. You might be our client. If that happens, we need to basically scour all kinds of sources. And if that happens, we need to immediately take that input and apply it to our regulatory relationship with you. 

Another interesting use case…we get lots and lots of documents from all kinds of entities that actually ask us to put a hold on a client because you know, whether it be a telephone company, or that somebody hasn't paid. We're getting about 6,000 of those a day. The minute that we get them, it's our responsibility, the burden is now on us. That's again a manual process that we’re completely automating. And it's only possible right now because of AI. 

A grey colored placeholder image