AI Engineer 'Devin' Deployed to Modernize U.S. Government's Legacy Code

SAN FRANCISCO and RESTON, Va. – June 03, 2026 – In a significant move to accelerate the digital transformation of the U.S. government, AI software engineering firm Cognition AI has entered a strategic partnership with Carahsoft Technology Corp., the heavyweight of public sector IT distribution. The alliance will bring Cognition AI's much-discussed autonomous AI agent, Devin, to federal, state, and local agencies, targeting one of the most persistent and costly problems in government: the modernization of decades-old legacy computer systems.

Under the agreement, Carahsoft will act as Cognition AI’s Master Government Aggregator®, making the Devin platform accessible through its vast network of reseller partners and critical government procurement contracts, including NASA SEWP V and NASPO ValuePoint. The partnership aims to provide government agencies with a powerful new tool to untangle, translate, and transform the millions of lines of archaic code that underpin many critical public services.

The Billion-Dollar Problem: AI Takes on Legacy Code

The federal government's reliance on aging IT is not just a technical issue; it's a strategic vulnerability. A significant portion of the federal IT budget is consumed by maintaining these brittle systems, many of which run on COBOL, a programming language from the 1950s, and operate on massive mainframe computers. This diverts funds from innovation and leaves critical infrastructure exposed to security risks. Compounding the issue is a demographic crisis: the generation of COBOL-proficient developers is rapidly approaching retirement, with some studies projecting 92% will have left the workforce by 2030.

This is the challenge Cognition AI and Carahsoft are aiming to solve. The partnership is less about replacing developers and more about equipping them with a force multiplier to tackle a problem that has grown too large for human capital alone. “Our platform doesn’t just translate legacy code – it understands it,” said Gardner Johnson, VP of Partnerships at Cognition AI. “That distinction is critical when agencies are modernizing millions of lines of COBOL and decades of mainframe logic, where a surface-level conversion creates more risk than it removes.”

By deploying an AI that can comprehend the intricate business logic embedded in old code, agencies can potentially bypass the “chicken-and-egg deadlock” where legacy systems block the adoption of the very AI tools needed to modernize them. The goal, Johnson explains, is to help organizations “reduce modernization timelines, lower risk and scale AI adoption with confidence.”

Inside Devin: The 'Agentic AI' Teammate

What separates Devin from a myriad of other AI coding assistants is its autonomy. Marketed as the world's first AI software engineer, Devin is built on an 'agentic AI' framework. Unlike tools that merely suggest snippets of code, Devin can independently plan, execute, and iterate on complex, multi-step engineering tasks. It operates within a sandboxed environment complete with its own shell, code editor, and browser, allowing it to function much like a human developer.

This capability is powered by what Cognition AI calls a “compound AI architecture,” which orchestrates multiple specialized models, selecting the right one for each sub-task, from analysis and transformation to testing and documentation. This model-agnostic approach prevents lock-in to a single AI provider and allows the platform to adapt as new, more powerful models emerge.

While early demonstrations showcased impressive end-to-end capabilities, independent analysis suggests a more nuanced reality. Devin's performance on the SWE-bench benchmark, where it resolved 13.86% of real-world software issues unassisted, was a significant leap over previous models. However, experts often describe its current ability as akin to a “smart intern” or “junior developer.” It excels at well-defined, structured tasks but requires clear instructions and human oversight, especially when facing ambiguity. For government agencies, this means Devin is not a hands-off replacement but a powerful tool for augmenting skilled teams.

Critically for public sector adoption, the platform is designed with enterprise-grade governance. The company highlights its adherence to Federal Risk and Authorization Management Program (FedRAMP) High standards—a rigorous security framework for sensitive government data—and a zero data retention policy, ensuring proprietary code and data are not stored by the platform.

Carahsoft's Playbook: Bridging Silicon Valley and Washington

An innovative technology like Devin is only transformative if it can navigate the labyrinthine world of government procurement. This is where Carahsoft's role becomes pivotal. As 'The Trusted Government IT Solutions Provider®', Carahsoft has perfected a model for bringing cutting-edge commercial technology to the public sector.

By acting as a Master Government Aggregator, Carahsoft provides a streamlined procurement path, eliminating the significant hurdles that often prevent startups from engaging with government agencies. Its vast portfolio of pre-approved contract vehicles allows agencies to acquire new technologies faster and with greater confidence in their compliance.

“Carahsoft’s partnership with Cognition AI makes harnessing the full potential of AI a reality for our Public Sector customers,” said Natalie Gregory, a Vice President at Carahsoft. She noted that the platform's ability to modernize legacy systems with confidence allows agencies to free up staff “to focus on what matters most and drive measurable business outcomes.”

This partnership is a classic example of Carahsoft's playbook: identify a critical government need, find a best-in-class technology solution, and leverage its massive distribution and contracting engine to deploy it at scale. The collaboration will include joint go-to-market initiatives, executive briefings, and field events designed to educate government leaders on the practical applications of agentic AI for modernization.

The Broader Strategy: From Banking to Bureaucracy

Cognition AI’s push into the public sector is not an isolated event but part of a broader strategy targeting large, regulated industries grappling with decades of technical debt. This move follows a recently announced partnership with financial services giant Fiserv, where Devin is being deployed to modernize core banking platforms. The common thread is a focus on complex, high-stakes environments where the cost of failure is high and the need for precision is paramount.

The competitive landscape is heating up, with major players like OpenAI and Anthropic investing heavily in their own agent-like AI solutions, and a host of open-source alternatives gaining traction. Yet, Cognition AI's sharp focus on the complete software development lifecycle, particularly for legacy systems, gives it a distinct position in the market.

This partnership represents more than just a new product on a government contract list; it signals a new phase in the application of artificial intelligence. The focus is shifting from simple generation to complex problem-solving and from replacing discrete tasks to augmenting entire workflows. As AI agents like Devin become more integrated into enterprise and government engineering teams, they will undoubtedly reshape not only how software is built but also how organizations approach their most deep-seated strategic challenges.