Tech firms embrace agentic AI, drives faster code transformation

Tech companies are evaluating the benefits of deploying AI agents to augment

developer productivity

while broadening their GenAI implementation. These intelligent systems are being utilised across diverse functions, spanning from programming to systems transfer.Anupam Mishra, director of Developer Programs, AWS India and South Asia, said that these agents are successfully executing moderate-complexity coding assignments. Their functionalities extend to test case creation, documentation generation, security flaw detection, and solution development. At the recent AWS Summit Bengaluru 2025, Mishra informed TOI that his AWS teams achieved four times faster migration from .NET (pronounced dot net ) to Linux while working with clients. The cloud services company's product demonstrated an 83% quicker conversion of older Java scripts to newer versions for their client Persistent Systems. "We have a case study where several teams converted their Java code from older versions to newer versions. Amazon developers saved 4,500 years of manual work and $260 million annually from performance improvements by using our Amazon Q Developer Agent for code transformation to migrate over 30,000 applications from Java 8 or 11 to Java 17," he said.Mishra said the average time to upgrade an application to Java 17 plummeted from typically 50 developer days to just a few hours. "In under six months, Amazon upgraded more than 50% of its production Java systems to modernised Java versions at a fraction of the usual time and effort. Amazon developers shipped 79% of the auto-generated code reviews without any additional changes. As we use AI extensively, mundane tasks like writing test scripts will be taken away from the coder," Mishra said.Arun Parameswaran, EVP & MD of sales at Salesforce South Asia, said that agents generate 30% of code, expected to reach 50% by year-end. Their AI product for creating autonomous AI agents achieved 93% accuracy for a European airline client. For Salesforce's helpline, agents handle 84% of queries, with escalations reduced by half. Their CodeGenie model has processed over 7 million code lines, addressed 500,000 developer queries, and saved 30,000 monthly hours, enhancing developer efficiency and reducing costs.

Cognizant earlier reported that agents write 20% of code, with plans to increase this proportion whilst passing AI-derived cost benefits to customers. US-based Constellation Research's principal analyst Ray Wang indicates that with

agentic AI

, code can be written using 20% of developers, achieving 50% higher productivity, while testing requires 40% fewer personnel. "We are seeing fully automated software development life cycles. In a world of exponential efficiency, IT services startups are delivering $1 million revenue per employee today and expect $5 million in 3 years," Wang said.IBM, in its proxy filing, said it achieved over $3.5 billion in annual productivity run-rate savings since 2022. Speaking at the IBM Think 2025 event in the US last week, its CEO Arvind Krishna said that in the long run, 'productivity is everything' and there is an 'insane need to go fast'. That is where agents come into play. According to Krishna, IBM’s AI products Spyre and Telum, when used to detect fraud in financial transactions, could save $190 billion.Nasscom's recent GenAI report highlighted India's tech industry leading in agentic AI readiness, from vertical solutions in legal or pharmacovigilance processes to multiple platform-based use cases.