Generative AI is beginning to help business technology leaders with the long-awaited challenge of modernizing their information technology systems. Some of them are filled with code from programming languages that predate the Beatles.
Generative AI is beginning to help business technology leaders with the long-awaited challenge of modernizing their information technology systems. Some of them are filled with code from programming languages that predate the Beatles.
“The big problem in our field, and in traditional companies, is that we have COBOL running around,” said Amin Benjara, chief data officer at ADP, a 75-year-old payroll processing company. “How many people know about finding a Cobol engineer?” he added.
Hello!Reading premium articles
“The big problem in our field, and in traditional companies, is that we have COBOL running around,” said Amin Benjara, chief data officer at ADP, a 75-year-old payroll processing company. “How many people know about finding a Cobol engineer?” he added.
The Roseland, N.J.-based company is developing its mainframe code from Cobol, a language first designed in the 1950s and still commonly used by banks and financial services companies, to a relatively new language in programming. We are considering using generative AI to “translate” to Java. Language dating back to 1995.
This translation reduces the need to find and train Cobol specialists. Cobol specialists have become relatively rare as generations of Cobol specialists retire and new programmers choose modern languages such as Python or, in some cases, languages other than Cobol.
Upgrading a company’s core technology, such as updating software or moving to a cloud computing platform, has always been a priority for chief information officers, but now, as CIOs seek to improve efficiency and contain IT costs, Its importance is increasing.
Over the past year, generative AI-based coding assistants from Microsoft-owned GitHub, Amazon, Google, and IBM have emerged to help developers with tasks like auto-completion of code snippets and creating code documentation. Some developers estimate that coding assistants increase productivity by about 25%. This is primarily a task similar to spell checking or autofill when assisting with document creation.
These days, companies like ADP are experimenting with using generative AI and similar coding tools to not only write new code but also upgrade old code. This is by far the most common use.
IBM remains reliant on its mainframe business and supports a wide customer base that relies on large-format computers, but it has launched the WatsonX AI Coding Assistant to help customers quickly and easily address the shortcomings of legacy technology. It’s on sale.
“Our clients haven’t always invested that much in their applications over decades, so they face risks, skills challenges and knowledge gaps for very extensive applications containing tens of millions of lines of code. ” said Skyla Loomis. Vice President of Z Mainframe Software at IBM.
IBM’s Coding Assistant, released last fall, uses generative AI to help developers migrate code from Cobol to Java or stay with Cobol. According to the company, companies have no plans to retire Cobol anytime soon. Cobol powers the computer mainframe, a large data server pioneered by IBM. It requires a lot of maintenance, but companies say it still works well.
Like other AI-based coding assistants, IBM’s tool provides developers with new code suggestions and allows users to ask questions in plain English.
Loomis said the hope is that IBM Coding Assistant will allow companies to update their traditional systems in just one to two years, rather than years. Compared to existing tools, generative AI can “understand the intent of the code” and translate it into ready-to-use Java, she said.
At Boston-based Wayfair, a generative AI-based coding tool just started helping the online furniture retailer’s 2,000 developers and data scientists update outdated code. Wayfair primarily uses Google’s Coding Assistant, said Fiona Tan, the company’s chief technology officer.
Founded 20 years ago, Wayfair doesn’t use Cobol, but does have “legacy code” in languages like PHP, old database code in languages like SQL, and code written by retired developers. Masu.
“For years, we’ve been building code that isn’t well documented. It doesn’t matter what language it’s in. It still takes a long time for someone to learn it.” Mr Tan said.
Wayfair is banking on AI tools to help reduce “technical debt,” or the defects and costs that arise when companies try to quickly resolve technical problems. Using AI, engineers can learn new languages faster, reducing technical debt, Tan said.
“As we get better and better, there will be some really great benefits for people whose digital transformation efforts have stalled or been postponed,” she said.
San Francisco-based Databricks uses generative AI to help engineers understand data storage and management companies’ code bases faster.
“Trying to get an old codebase is a pain, so being able to quickly understand what that codebase does is very helpful for engineers,” said Naveen Zutshi, the company’s CIO. states.
Arun Chandrasekaran, a Gartner analyst specializing in AI and cloud, cautions that code upgrades are just one part of a larger task of modernizing technology systems.
“If you’re thinking about this as a really complex workflow automation tool, you’re going to need multiple AI models, some focused on code generation, others focused on dependency mapping and impact analysis. There are also some models that have been designed to fit the needs of our customers,” says Chandrasekaran. . ”
However, implementing generative AI tools comes with its own risks, including incurring additional technical debt. Some technology leaders say the ease of writing code quickly also makes it easier for poorly documented or irrelevant code to sneak in, requiring greater human oversight.
“If you want to move quickly, you’re going to take on some technical debt. We’re in that balance right now,” ADP’s Benjara said.
Email Belle Lin at belle.lin@wsj.com.
