Cognizant and Infosys Deploy Agentic AI Crews to Automate Enterprise Software Maintenance

India's IT giants are transitioning from simple co-pilots to autonomous Agentic AI crews, automating up to 70% of legacy codebase migrations and maintenance tasks.

India's major information technology services firms, including Cognizant and Infosys, have announced the deployment of autonomous Agentic AI crews to manage high-volume software maintenance and system migration contracts.

The shift marks a transition from human-assistant coding bots (like GitHub Copilot) to fully autonomous systems that plan, write, test, and deploy code updates with minimal human oversight.

#

From Autocomplete to Autonomous Crews

Unlike standard generative AI models that require prompt-by-prompt guidance, agentic crews coordinate multiple specialized agents (e.g., an analyst agent, a coder agent, and a tester agent) to resolve complex, multi-step engineering tickets: 1. Ticket Analysis: An agent analyzes incoming bug reports or feature requests, mapping dependencies across the codebase. 2. Implementation: A coding agent generates the diffs needed to address the issue. 3. Self-Correction: A testing agent runs test suites, catches compilation or runtime failures, and sends error logs back to the coding agent for self-correction. 4. Human-in-the-Loop Approval: Once tests pass, the proposed pull request is presented to a senior software architect for final review and deployment.

#

Reshaping the IT Talent Landscape

Tech leads report that these agentic workflows are reducing the time required for legacy Java-to-Kotlin migrations and Cobol-to-Cloud rewrites by up to 70%. While this automation is dramatically cutting costs for enterprise customers, it is also forcing IT services companies to upskill their entry-level workforces. The focus is shifting rapidly from pure coding syntax towards systems architecture, security auditing, and prompt engineering, redefining the profile of the software engineer in India's tech corridor.

Planning your startup's tech architecture? Use the Tech Stack Decider to pick the right backend for your scale.