生成AIによる技術的負債の低減

In a previous life I was responsible for a portfolio of mission critical banking platforms. They spanned a diverse range of technologies from a few IBM mainframe applications that were almost twice the age of many on my team, to the latest machine learning models running in Kubernetes containers on Red Hat OpenShift – with a host of Java-based applications in between.

At one point there was business interest in migrating a couple of mainframe applications (each with about one million lines of code) to Java on a more modern platform. A code migration vendor provided estimates of a few million dollars – just for their costs – to lift and shift the code over an 18 month period. That was a short conversation about ROI, relative risk, and developer productivity that I’m honestly pleased did not move ahead. The right solution to was to break-down the application components and replace/rebuild any needed functionality into the progressively modern suite of applications. That was still not an inexpensive proposition, but it serves to highlight one of the big dilemmas for technology executives – eliminating technical debt cost effectively.

Celent’s Technology Insight and Strategy Survey (CTISS) 2023 found that, on average, only approximately one-third of “change” investments are allocated to growth initiatives—that’s less than one-sixth of overall technology spend. Since mandatory projects consume proportionately more spend at the expense of growth initiatives, there simply aren’t enough available funds or resources to deliver competitive change quickly. We delved into this further in our report, Corporate Banking Global IT Priorities and Strategy in 2023.

Drivers of Technology Budgets

Base: All Corporate and Retail Banking respondents (sample: 442)
Question: What are the top three drivers of your institution’s IT spending strategy for 2023?

Source: Celent Technology Insight and Strategy Survey 2023

Banks must shift gears and reallocate investment to technology that supports business growth through more agile, efficient platforms and methodologies. This is not just for innovation projects, but for all technology solutions. As bank IT spending increases by an average of 5.6% globally in 2024 their technology portfolios will expand. Banks must find ways to deliver projects faster—either through more effective DevOps, efficiency in code development, and in technology lifecycle management. This is not just about DevOps and agile process improvement. There are many ways to develop greater speed and agility in technology transformation, and there is certainly no “one-size fits all” prescription.

Taming Technical Debt

For all the exciting and legitimate use cases about the impact of Gen AI on business productivity, let’s also consider developer productivity. Gen AI has a huge potential to transform the way software is developed and refactored. Gen AI co-pilots can not only help developers write code, but they can also be used to migrate/upgrade to newer versions of the code platform. This has the potential to reduce technical debt, overall cost of technology, and deliver software projects more quickly. Not all technical debt initiatives are as extreme as decommissioning mainframe applications. Often the challenge is modernization or migration within established code platforms (Java/Java or .Net/.Net) – often driven by end-of-life deadlines for old versions, compatability, and related patching/security vulnerabilities.

At the ReInvent conference last month, AWS showcased how Gen AI with the new “Amazon Q” co-pilot can support developer productivity. Although still only in “preview” mode, AWS shared the results of proverbially eating their own dogfood with Amazon Q Code Transformation. As an example, 1,000 internal Amazon applications were upgraded from Java 8 to Java 17 in just two days. In the past, it took an average of at least two days per application.

Of course – such generic statements don’t directly translate to the efficiencies that might be gained in banking applications, but this does change the conversation about technology efficiency. As noted in my earlier blog - Viva Las Vegas! A Quick Recap of AWS ReInvent 2023 - banks must be aware of the developer initiatives the hyperscalers are investing in and “reinvent” how technology is delivered. If banks can’t find better ways to deliver “run the bank” and mandate initiatives faster, they are unlikely to truly “change the bank.”