The rise of AI agents in software engineering is no longer just an experiment but an industrial bet. That’s the message from the announcement of Infosys and Cognition, who have announced a strategic partnership to scale Devin—presented as “the first AI-powered software engineer”—within the Infosys development ecosystem and, gradually, in projects for clients worldwide.
The agreement, announced on January 7, 2026, between San Francisco and Bengaluru, places a major tech consulting firm on a clear path: integrating agents into development not just as assistants, but as operational components within their delivery models. At the core of the partnership is the combination of Devin and Infosys Topaz Fabric, a “layer” designed by Infosys to unify infrastructure, models, data, applications, and workflows within a composable, agent-ready environment.
From internal pilot to large-scale adoption
Infosys states that it has been using Devin for six months and, after observing “significant” improvements in quality and efficiency, has decided to integrate it into internal teams, incorporate it into client delivery models, and enable deployments within the engineering environments of its users.
This is a key point because it addresses a core aspect of digital transformation: system modernization doesn’t happen in a “test lab” but in real-world scenarios, with accumulated technical debt, historical dependencies, and operational risks. In this context, Infosys and Cognition aim to automate “brownfield” work (projects built on existing bases), reduce technical debt, and shorten modernization timelines; they even mention creating “virtual engineers” capable of tackling complex production incidents and maintenance tasks that currently consume many hours from senior teams.
To scale adoption, both companies plan to work on shared frameworks and training programs that bring combined capabilities to engineers across various industries. The strategy also includes the joint development of industry-specific solutions, AI-native modernization blueprints, and scalable engineering frameworks, supported by co-innovation labs.
Topaz Fabric: the “fabric” aiming to organize enterprise AI agents
Infosys’s approach is not just about adding an agent and hoping for magic. The announcement positions Infosys Topaz Fabric as a structural component: a suite of layered agent services designed to bring coherence to an increasingly fragmented enterprise landscape—covering models, data, controls, and workflows.
Practically, the promise is that companies can integrate agents within a modular and secure architecture, preventing each team from adopting isolated, hard-to-govern tools. This aligns with a current challenge: the real difficulty isn’t just testing agents but governing them (permissions, traceability, security, compliance) as they become part of the software lifecycle.
Devin: from “copilot” to autonomous agent
Cognition defines Devin as an AI team companion aimed at expanding engineering capacity: an coding agent that can collaborate with humans and undertake development tasks with greater autonomy. The partnership with Infosys seeks precisely this: to move Devin from individual or small-team use to complex enterprise environments, where software is built via shared processes, controls, and responsibilities.
In the announcement, Scott Wu, founder and CEO of Cognition, frames the partnership as a step toward more autonomous and “agentic” engineering in large organizations, emphasizing that Infosys would be the first major digital services and consulting firm to deploy such tools at this scale. Similarly, Salil Parekh, CEO and Managing Director of Infosys, describes the agreement as a catalyst allowing organizations to realize value from AI, leveraging Cognition’s capabilities combined with Infosys’s industrial reach.
Initial use cases: banking, payments, markets, and insurance
The press release hints at where deployment will begin: Infosys’s Financial Services practice is already using Devin to transform engineering delivery in areas like banking, payments, capital markets, insurance, and wealth management. The choice seems deliberate: these sectors rely on critical legacy systems, face regulatory pressures, and constantly need to evolve products without disrupting operations.
The opportunity… and the reality check
On paper, the goal is clear: accelerate time-to-market, increase productivity, and shorten modernization cycles. However, recent history in software automation shows that outcomes depend heavily on how these initiatives are implemented: which tasks are delegated, how quality is controlled, who is accountable when an agent acts, and what metrics are used to measure real impact.
Therefore, the announcement emphasizes an “enterprise-grade” approach: sector-specific solutions, scalable frameworks, and enablement programs. The challenge isn’t just technical; it’s organizational as well. For many companies, the critical question isn’t whether an agent can write code but whether it can do so within a governance system that ensures security, auditability, and traceability without slowing delivery velocity.
Infosys, which positions itself as a global player with over 320,000 employees across 59 countries, is betting here on more than just headlines: proving that AI agents can be integrated into the machinery of consulting and industrial software delivery without becoming operational risks.
Frequently Asked Questions (FAQ)
What is Devin and how does it serve a company with legacy systems?
Devin is an AI programming agent focused on software engineering tasks. In environments with legacy systems, it aims to support modernization, reduce technical debt, and assist with maintenance within a controlled and validated framework.
What does Infosys Topaz Fabric add compared to using standalone AI agents by teams?
Topaz Fabric acts as a layer to unify infrastructure, models, data, and workflows within an environment designed for agents, making adoption manageable and scalable in large organizations.
What is “brownfield engineering” and why is it key to modernization?
“Brownfield” refers to working on existing systems (with dependencies, legacy code, and critical processes). It’s crucial because most enterprise software exists in such setups, and modernizing these is often more complex than creating new projects from scratch.
What risks should companies watch when deploying autonomous coding agents?
Primary risks include security (access to repositories and data), change control, traceability, and technical responsibility. An “enterprise-grade” approach requires permissions, audits, validation, and clear review processes.
via: infosys