The Anatomy of a Modernization Failure and What We Can Do About It

Every financial institution is at some stage of modernization. The reasons are clear, the economic justification crystal. And after years of trying, we have accumulated best practices and various centers of excellence.

None of these explains why the financial services landscape is still littered with modernization failures and under-delivering initiatives. So, what’s holding modernization back and what’s contributing to the failures?

John Duigenan, general manager of financial services for global industries at IBM Technology, has one word: “mindset.”

Mindful modernization

Duigenan, a distinguished engineer, sees mindset at the root of a successful program. In his authored whitepaper “Fit for Purpose: Modernizing With Impact,” he calls for a new way of thinking, working, and architecting for modernization.

“For the most part, institutions are under competitive pressure. Competitors are coming from unanticipated companies, often by surprise. Equally, consumer expectations have never been higher. Addressing customer expectations is no longer a choice, even for the most incumbent organizations,” says Duigenan.

To survive and thrive, financial institutions need to innovate at speed. Duigenan points out that “the conventional delivery cycle of two product release cycles a year is no longer viable.”

Yet, our minds seem to be telling us otherwise. Duigenan offers two extreme examples to show how we are being led astray by conventional thinking.

One is the view that there’s nothing new to learn after deploying new technology, leading to a period of absent investment. “A lot of firms have been complacent, even arrogant, in their technology investments,” says Duigenan, and have not invested in “technical vitality or innovation.”

“Arguably, they have even under-invested in their core platforms and the skills required for those platforms to evolve. Those firms find themselves at a pressing modernization crossroads,” Duigenan explains.

Instead of evolving their current IT investments, these same financial institutions take the more time-consuming “bold rip and replace approaches.” The result: the creation of untested future products and an ROI they cannot achieve.

Another familiar example is the belief that “everything goes to the cloud,” often with a single cloud provider.

This approach is flawed because moving everything to a cloud or doing a rip-and-replace does not accomplish ROI. “Often, other motivators are at the heart of this approach, such as the cult of the celebrity CIO. In doing so, firms do not sweat or evolve their existing stable assets or leverage innovation, scale, and inherent security already available to them,” says Duigenan.

IBM advocates a hybrid approach that leverages existing and new assets on a fit-for-purpose infrastructure to unlock and accelerate innovation and simplify operations. But more importantly, it highlights the need for a new mindset, one that is rooted in an objective, comprehensive, and honest view of ROI.

Clarifying the murky technical debt

The term “technical debt” is often used synonymously with the modernization journey. Many journeys are framed as exercises in refactoring or lifting and shifting applications to modern, more agile systems so we can eradicate technical debt.

 “That’s a fallacy,” says Duigenan. “Technical debt is a favorite mythical topic. I happen to love it. Firstly, the myth is that technical debt can be eliminated. Eliminating technical debt is a fundamental untruth.”

Why? “Today’s technology decisions will incur a future rectification cost. The second the code is written and deployed, it is aging and incurring a future debt,” says Duigenan. In other words, technical debt is a necessary evil, and financial institutions should do better in mitigating its risks rather than waiting to get rid of it with a single initiative.

Waiting is also ill-advised as other not-so-apparent risks may manifest quickly over time. It’s like financial debt interest — technical debt risks compound over time. And it’s not just agility or performance that gets impacted. Duigenan points to cybersecurity, where aging code and libraries need remediation as risks are identified.

“Failing to take action, or as we like to say, ‘kicking the can down the road,’ increases the cost of rectification over time, fueled by multiple factors including inflation, the scarcity of technical skills, vendors exiting business, and so on,” says Duigenan.

So, taking action now rather than waiting for an opportune moment is well advised. Sadly, kicking “the can down the road” is more common. “That happens simply because the ROI for addressing technical debt is not compelling,” Duigenan adds.

Old is not bad

We also need to understand that because technologies are mature does not mean they are out of fashion. Mainframes such as IBM Z are prime examples.

“It is one area ripe with misperceptions. Some would suggest it is old technology, part of the legacy — it is far from that. IBM z16 is the most modern hardware infrastructure available,” says Duigenan.

It can also be more economical to keep older and more baked technology. “Dollar for dollar, when measured by a technology economist, the cost per transaction is lower than distributed systems, and the revenue per transaction is higher,” says Duigenan.

Besides, innovation for aging software platforms is already available. So there’s no reason financial institutions should not explore alternative routes to modernization.

“Firstly, 60-year-old software is among the cheapest to operate, as the majority of its development cost was amortized in prior generations. However, innovation for aging software platforms is readily available through code-refactoring, AI-infused development tools, DevOps toolchains, and software that opens the data and transactions to new-generation code,” says Duigenan.

AI needs a different mindset

The much-needed shift in our mindset becomes more urgent with AI.

While financial institutions are no strangers to AI, many focus on operational benefits. The timing of the ChatGPT consumer release turned this gradual process of bolting AI onto modernization initiatives on its head.

Consumers can now see what’s possible with advanced AI and, for the first time, can’t get enough of it. Subsequently, it is forcing financial institutions to move AI to the center of their modernization journeys, what Duigenan calls a shift “from +AI to AI+.”

“Business stakeholders and development organizations are at a pivotal point on the AI journey. We call the starting point ‘+AI’ in which AI capabilities such as machine learning decisions are infused into existing applications. Moving forward, we’ll see ‘AI+’ in which AI capabilities, such as virtual agents, taking questions and instructions in natural language, are the core, and application capabilities are built around them,” he explains.

While this shift may seem obvious on paper, it needs a proper rethinking in practice.

Take data placement, for example. Conventional wisdom says that data and processing should be co-located. While this doesn’t always have to be true, Duigenan thinks, as a rule, it makes sense, most often for reasons of latency and risk.

“However, arbitrary migration strategies have forced the issue of data placement. While I might not want to move very private data, if the only place I can run AI is in a public cloud, I have to move the data and take on the risk of leaving that data in a cloud,” says Duigenan.

“That’s a big step when it comes to both the (regulated) private and proprietary data/algorithms that are the value driver in most firms. If the only way I can use a service is to move the data, I have to accept the risks,” he adds.

How we manage production data placement directly impacts machine learning and, in turn, how well we build a model. In financial institutions, especially retail banks, this type of data is mostly controlled and least likely to leave the premises.

“The training challenge is a key reason why firms are struggling to scale even their most basic AI efforts,” says Duigenan.

It is one reason IBM advocates a multi-tiered approach that layers data access and data governance with AI development and deployment. This is a key differentiator with IBM’s Watsonx promise and the company’s partnership with HuggingFace.

At the same time, data fabric and lakehouse capabilities are essential along the modernization journey as they provide connectivity, virtualized data access, and open storage format.

“When mapped with foundational models, large language models, and generative AI, the impact will be profound in how existing and new applications will be built and delivered,” says Duigenan.

Then there are regulations, which consumers don’t have to worry too much about.

“While all of the public awareness of Generative AI is coming from consumer-based approaches, many of these will not fit with the enterprise requirements for trustworthy AI grounded in need for security, privacy, freedom from bias, and explainability,” says Duigenan.

Financial institutions not actively addressing their regulatory risks of using AI are living on borrowed time. Regulators are waking up and demanding better explainability and showing that the model’s output was bias-free.

“For either eventuality and those in between, IBM advocates a hybrid approach that leverages existing and new assets on a fit-for-purpose infrastructure to unlock and accelerate innovation and simplify operations,” says Duigenan.

It takes a village

Changing mindsets for modernization is a big challenge. We know now that modernization cannot be seen as an IT project to succeed. But to shift our mindsets, we need better stakeholder alignment.

Duigenan suggests starting with the senior stakeholders. “For example, if LoB and Technology stakeholders sit in different organizations, they must align around business goals and outcomes. Doing technology programs without buy-in from the business results in doom and missed expectations.”

Next is to share the program priorities across the organization. “For example, procurement should not negotiate with partners and vendors in isolation of business and tech objectives. If poorly aligned, they’ll buy technology and services that are not appropriate to the required outcomes,” says Duigenan.

Essentially, stakeholder alignment is where modernization journeys should start. Getting this part wrong will mean the journey is doomed for a restart.

“Without cross-organization business and technology stakeholder alignment, there’s little chance of success,” says Duigenan.

Winston Thomas is the editor-in-chief of CDOTrends and DigitalWorkforceTrends. He’s a singularity believer, a blockchain enthusiast, and believes we already live in a metaverse. You can reach him at [email protected].