AI will create more software, transform developers' roles: IBM's Raghavan

IBM's Sriram Raghavan says AI agents will accelerate software creation and reshape developers' roles, while enterprises must strengthen cybersecurity, governance and skills

As many as 2 billion artificial intelligence (AI) agents could eventually be deployed globally, with AI writing half the code, according to Sriram Raghavan, general manager of IBM Software India and Software Innovation Lab.

 

He said that rather than making software development redundant, the proliferation of AI agents would lead to the creation of far more software, fundamentally changing the role of developers and how enterprises build technology.

 

“If anything, you’re going to see way more software,” Raghavan told Business Standard on the sidelines of its flagship event, IBM Think, which recently concluded in Mumbai. He did not specify a timeline for the deployment of these agents.

 

According to Raghavan, AI has already brought about changes in software development, including how teams are organised, workflows are managed and software is built.

 

The role of developers is also undergoing a fundamental shift, from writing code to monitoring, reviewing, and evaluating AI-generated output.

 

“The role of the developer has now changed from doing to monitoring, reviewing, and critiquing. That is a skill change. You have to train your developers differently,” he said.

 

He drew a parallel with electronic design automation in the semiconductor industry. Raghavan said automation had enabled engineers to design increasingly complex chips without eliminating the need for hardware engineers. “We are doing more of semiconductors and chips everywhere in the world,” he added.

 

He expects a similar transformation in software development, with automation allowing enterprises to develop more applications while requiring developers to acquire different skills. “The expertise is now checking, verifying, setting strategy, guardrails, guiding the AI to do the right thing. That is where I think software engineering is going,” he said.

 

In fact, Raghavan believes that over the next two to three years, more software will be produced than ever before.

 

“I think we are going to see more software produced than we have ever seen. Because I think the opportunity to innovate has never been easier. So now the premium is on ideas and deployment. Premium is not on creating software,” he said. However, he cautioned that the growing use of AI would not diminish the importance of fundamental engineering skills. “If anything, judgement, core skills become more important than less important,” he added.

 

Asked about recent security concerns surrounding the release of new frontier AI models and whether enterprises were becoming more cautious about their AI road maps, Raghavan said AI adoption was fundamentally changing the cybersecurity landscape. “Is there a different way that we have to think about cybersecurity because of AI? The answer is absolutely yes,” he said.

 

He said organisations needed to adopt a proactive, continuous approach to cybersecurity, using AI to integrate best practices, strengthen technology, manage vulnerabilities, and prioritise risks.

 

“What we are telling our clients, and our offerings reflect that, you have to think about AI infusion into defence as much as we all worry that AI will be used by the bad guys,” Raghavan said. “Waiting for every fix for some human being to sit and check and click a button, you are going to be behind," he said.

 

On concerns surrounding AI agents taking unintended actions, Raghavan said enterprises must impose stricter controls on agent identities, permissions, and access to internal systems. He said agents should be granted only the minimum permissions required to perform a specific task, rather than broad access to enterprise systems.

 

“The agent needs a certificate and a token to take an action for the next 30 seconds on your behalf. That’s all that it needs to do,” he said.

 

He also emphasised the importance of continuous monitoring, as AI systems, unlike conventional software, do not always behave predictably.

 

India’s AI race: Few at the finish line

 

Citing an IBM study conducted with the Ministry of Electronics and Information Technology earlier this year, Raghavan said only about 15 per cent of Indian enterprises had reached the stage of scaling AI deployments, while the remaining 85 per cent were still struggling to move beyond initial adoption.

 

Companies scaling AI have moved beyond individual use cases and are deploying the technology across their organisations. “About 15 per cent of enterprises are truly scaling AI, which means they have reached a stage where a use case has been deployed across the organisation and they are perhaps moving on to use case No. 2 or No. 3,” he said.

 

For the remaining companies, the biggest obstacles are data management, governance, and skills.