Breaking down the walls: Enterprise architecture gets its day in the sun

The head of technology at a large US bank has a strained look on their face. They’re trying to explain the importance of enterprise architecture and the pushback they received upon joining the bank a few years ago. 

“I’d say that project teams initially found it to be annoying,” they say. “When I got there, shadow IT had popped up all over the place. More than that, technology itself was so locked down by bureaucracy that it could barely move. As a result, the business would just go work with whoever could solve their problem—they’d do whatever they wanted.”

All of a sudden, there was a new function (or layer) that had to be adhered to, and a new person explaining why this new setup was necessary. Fortunately, senior management—starting with the CEO—was all-in on a tech overhaul, and for good reason. 

In 2018, complaints from the bank’s various business units were rising to the level of the CEO. As a result, they commissioned an informal survey where they sent a senior technologist and a senior business leader around to ask employees their thoughts about the bank’s technology. 

Can you imagine? The head of capital markets has no idea how things get done. That’s very bad.

Head of technology at a large US bank

“They got an earful—it was terrible,” says the head of tech, who also heard a disturbing story. “When I was interviewing, the head of capital markets said to me, ‘I have no idea how technology projects get prioritized.’ Can you imagine? The head of capital markets has no idea how things get done,” they say with extra emphasis. “That’s very bad.”

All of the feedback led to a three-year project that focused on implementing, or at least formalizing, three functions: application identification and data management, tied together with the enterprise architecture layer. Most important, though, was how tuned into the strategy senior executives were. During the course of the project, the key business and IT leaders—along with a large consulting firm providing guidance—would meet with the C-suite for 90 minutes every two weeks.

The head of tech, who was at the time still new to the bank, says it was “fairly wild” for IT to have that much access to the CEO and major business leaders for such a long time, that consistently. 

“The C-suite got way deeper in the weeds than they ever should have to; but in this case, they had to—they had to see where decision-making was a challenge.”

Interop by any other name

While it has yet to be explicitly stated in this Voice of the CTO series, if there’s one major theme that ties each of the five stories together, it’s interoperability. But depending on who you talk to, that word takes on different meanings. 

There’s the desktop app interoperability crowd, which, as the name implies, is all about application connectivity inside a desktop browser. There’s interoperability of previously walled-off third-party systems, outside of a browser. Then there’s the seemingly simple act of making sure that for, say, basic reporting needs or cutting data costs, internal systems are talking to one another, and that there’s an oversight structure around those applications. And then there’s some combination of all of those. 

Ultimately, the goal is to get systems and applications to connect, break down siloes, and reduce reliance on legacy, monolithic platforms. This helps reduce technical debt and free up budget for more innovative projects. But equally as important is the concept of bad data in, bad data out. Large language models and generative AI are impressive evolutionary steps, but they’re only as effective as the data being fed to them. Without that data management layer, data duplication costs skyrocket; applications spit out erroneous information; regulatory reporting becomes more difficult; unique insights are tougher to uncover; and risk grows.

This is why the US bank’s head of technology is so passionate about the topic of enterprise architecture—which, at its core, is all about creating structure, governance, and interoperability around applications and data. 

“You need that enterprise architecture perspective to ensure all the bits and pieces come together at the enterprise level,” rather than in a silo, they say. “It’s really data management and enterprise architecture together to help ensure that each piece you add to the puzzle has a known place.” 

The chief operating officer of a global systemically important bank (G-Sib) echoes this sentiment. As part of “a sizeable investment” to create a golden record of data—a master of securities, products, and customers—the bank realized that systems were not relaying data as efficiently and accurately as they wanted. This led the bank to rethink its definition of interoperability.

“We have a vast sea of products,” says the COO. “We’re not just quant trading or investment banking—we’re a supermarket with so many different products. So we needed these platforms to connect more seamlessly.”

The COO then mentions a word that was echoed by several technologists spoken to for this series: reuse. 

What’s old is new again

“Reuse—for data and for breaking down monolithic systems—is used to make things more nimble and portable across the organization,” says the COO.

They point to an internally built wealth management platform as an example of reuse’s success. After identifying that there were user experience and data visualization issues with the old platform, they decided that rather than build a new one from scratch or buy off the shelf, they could first recycle what they already had.

For instance, what they found was that pieces from its treasuries system and index data services system could be repurposed and stitched together with working components of the old platform. The COO says that a new platform was built within the year.

“The reason it was able to come from an MVP stage, to test trial, and then to a fast launch in such a short amount of time was because of reuse. We basically took the ecosystem of our ‘like’ areas across businesses and reused. I now bang the drumbeat for reuse.”

And, they say, there is an added benefit of breaking away from walled-off, monolithic systems. Applications using modern tools like cloud, open-source, and APIs can be more easily repurposed, thus cutting down on technical debt. But just as there are multiple definitions for interoperability, the term “reuse” means different things to different people, depending on the situation.

The chief information officer at a second G-Sib says that when it comes to building an enterprise architecture and creating interoperability between systems, this is where “cloud reuse” becomes valuable.

Major enterprise architecture projects take years, so budget has to go to those endeavors, or else they become quagmires. But at the same time, banks can’t afford to fall behind on innovation discovery. This is where the CIO tries to identify what they do well, and then partner with major cloud providers to augment those skills or fill in skills gaps. 

“It’s difficult to get consistent budget for innovation. So the hope is that over time, with the other aspects of cloud—like AI—what they provide will get better,” says the CIO, pointing to open-source tools like TensorFlow for machine learning, or specialist offerings like OpenAI’s ChatGPT for generative AI, which is based off of tools and libraries for natural language processing and artificial intelligence research that were open-sourced. 

“The reality is, if we can put applications and datasets in the cloud, we can leverage, say, Google’s amazing capabilities when it comes to AI construction and data science, and all that is interoperable with our cloud architecture. But we have to get our data to the cloud first to use it,” the CIO says.

Endings as beginnings

Going back to the head of technology at the large US bank, the three-year revamp is over and yielding results. But the project, technically, never ends once the enterprise architecture is in place. The firm has since created an annual planning process that involves the C-suite from the start.

Every month, there’s a meeting where IT, the business leaders, and the C-suite go over status reports for projects set to end within the year. They go over the roadmap for regulatory initiatives that could take upwards of three to five years. They identify strategic priorities. And they decide whether older proposals should be brought back to life as other initiatives are completed.

The feedback is then triaged, and charts are created that detail how much of each business silo’s budget is going toward operational efficiency versus customer service versus technical debt reduction.

“So there might be 50 projects—you can’t look at 50 projects and make heads or tails of it,” says the head of tech. “But you can sum up the spend and say, ‘Yeah, I agree that’s the right allocation of spend,’ or not.”

They say that while it felt like they were building a foundation “for a really long time,” it was exciting to see the fruits of that labor bloom.

For example, the head of tech attended a board meeting last year, and the bank’s audit team did a presentation on how it was able to use a recently incorporated data visualization tool called Qlik Sense to analyze data and support its audits. “I didn’t even know that was happening. When stuff is happening and you don’t even know, that’s great. There’s a certain amount of end-user self-service that we enable.”

When asked if that setup is also what enables pernicious shadow IT to grow, they explain that once the enterprise architecture structure is in place, everyone is working off the same platform, and they can be self-sufficient in a governed environment. 

“There’s no way for IT to keep up with what the business is trying to imagine, so you put some of that in their hands. You enable them to make decisions within the confines of an enterprise and data management structure.”

The head of tech says this new process has also helped business unit heads and the C-suite to understand IT’s cloud, interop, and data strategies.

“This three-year process would’ve only worked top-down. If it had come out of technology, you can have some kind of persuasive skills that can get you some of the way, but for the harder decisions, you’re just not going to get there,” they say.

“I can assure you that, now, no one will ever say, ‘I don’t know how projects get permissioned.’”

Maybe so, but it’s also important to note that none of the concepts expressed by technologists for this series are particularly novel, and technology is evolving at an increasingly rapid rate. It takes just one change at the top or one key champion leaving, followed by budget cuts and reorgs and questions of “why are we doing this and not that?”—and before you know it, you’re back to where you started, and the walls have gone back up.