In the bustling year of 1959, a vivid scene unfolds: President Ike Eisenhower is in office, while Elvis Presley, amid his time in the U.S. Army, is experimenting with amphetamines and karate. Fashion trends of the time highlight women in stylish capri pants and young greasers emulating the legendary James Dean.
Simultaneously, a group of largely unnoticed computer scientists, mathematicians, and other brilliant minds—many of whom are women, defying the male-dominated workplace—are developing what will eventually be known as the Common Business-Oriented Language, or COBOL.
Fast forward to today, COBOL may seem like an archaic language in the realm of modern technology, yet it continues to underpin essential operations in sectors such as banking, insurance, and government—industries notoriously slow to innovate.
Recently, the latest wave of technology—artificial intelligence (AI)—is being touted as a solution to help public agencies modernize their COBOL codebases, as indicated by a recent announcement from Anthropic.
In a blog post released on Monday, the company highlighted its tool, Claude Code AI, designed for “COBOL modernization.”
Despite the promotional nature of the post, an Anthropic spokesperson remarked to Government Technology that the capability has existed for some time, serving more as marketing than a groundbreaking announcement.
This blog serves to emphasize the ongoing necessity for COBOL expertise within the public sector, reminding us that what was revolutionary in 1959 is something that few are familiar with as we approach 2026.
“The developers who built these systems have long since retired, taking with them valuable institutional knowledge,” the blog notes. “Over decades, production code has been altered repeatedly, yet the corresponding documentation has lagged behind. Meanwhile, educational institutions are not churning out replacements—COBOL is taught at only a select few universities, making it increasingly challenging to find engineers who can navigate it.”
Determining the exact number of COBOL specialists still active in the public sector, or anywhere else, is no easy task. For instance, NASCIO, the leading organization for state CIOs, does not maintain specific figures, and estimates of COBOL programmers in the U.S. range widely—anywhere from as few as 24,000 to as many as 2 million.
The ongoing loss of COBOL programmers, due to retirements and fatalities, continues to shrink the available workforce. Nevertheless, COBOL remains vital; according to the Anthropic blog, a remarkable 95 percent of ATM transactions in the U.S. are facilitated by COBOL.
A 2022 paper from the Association for Information Systems observed that although COBOL has “mostly vanished from classrooms,” approximately 800 billion lines of the language still support various systems.
In the public sector, technology leaders and government suppliers continue to make strides toward upgrading COBOL-based systems for services like courts and more, with Anthropic being one of them.
“From SNAP to the DMV, state and local governments maintain essential services for millions, and we’re thrilled to assist them in modernizing the legacy systems supporting these services,” stated Michael Lai, the state and local government lead at Anthropic, in an email interview with Government Technology. “Claude can handle the labor-intensive components of IT modernization—like documentation and unit tests—that often render these projects slow and unfeasible, supercharging government IT teams to tackle previously insurmountable challenges while saving taxpayer money.”
The COVID-19 pandemic laid bare just how reliant government operations are on COBOL, prompting officials to pursue often costly alternatives to distance themselves from the language.
Take Oregon, for instance. The state’s Employment Department recently allocated at least $106 million for a new system to replace a COBOL-based operation—this decision followed a chaotic period during the pandemic when the state struggled to manage an overwhelming influx of jobless claims.
An increasing number of tech companies, including Anthropic, are eager to take on a larger role in these modernization efforts. Their appeal lies in the complex challenge of moving away from COBOL as compared to more familiar code updates.
“COBOL modernization is fundamentally different from standard legacy code refactoring,” the blog articulates. “You’re not merely updating existing code patterns; you’re reverse-engineering business logic from systems created during Nixon’s presidency.”
Anthropic asserts that its AI can automate much of this daunting process while also identifying potential risks associated with upgrades and documenting “workflows that have faded from memory.” With the aid of AI, teams can modernize their COBOL codebases in a matter of months instead of years.
As we navigate the complexities of modernization in today’s digital age, the legacy of COBOL serves as a reminder of both the challenges and the innovation opportunities that lie ahead. With AI stepping into the spotlight, there’s hope for revitalizing these systems while ensuring critical services can continue to operate effectively.
