Build AI or use it better

Students are experimenting with AI tools, older folks are trying to keep up, and the promise of AI-powered careers often feels both exciting and overwhelming. B Ravindran, head of the Wadhwani School of Data Science and AI at IIT Madras, offers two ways to approach the AI-enabled world.

“One is to ask, how am I going to build the next generation of AI?” he says. “The other is to ask, how do I use AI to do what I already do better?” AI is not just for those who want to become machine learning experts. It is already reshaping how work gets done across fields. “Whether it is drug discovery, designing new materials, understanding history, psychology, accounting – it doesn’t matter,” Ravindran explains. “AI is allowing you to do things more efficiently and also in ways you didn’t imagine possible.”

But whichever path a student chooses, Ravindran is clear shortcuts will not work. “You need strong fundamentals.” For those wishing to push the boundaries of AI, that means mathematics – linear algebra, calculus, probability, statistics – along with computing basics and an understanding of how systems run.

The ease of using AI tools, he says, can create the illusion of expertise. Ravindran warns against that. “If you jump straight into AI algorithms, you are basically doing things by rote or becoming a tool user,” he says. “That is not enough.”

For those not interested in building AI systems, the advice is different but equally demanding. Depth in one’s own field matters more than ever. “You should have good expertise in your domain and know how to use AI properly in that domain,” he says. “People with domain expertise who can use AI effectively will be more valuable than people who just run AI tools.”
It is also difficult to separate useful suggestions from nonsense without some domain expertise.

That distinction is becoming sharper as the technology evolves. Tasks that once required dedicated engineers are increasingly automated. “A year ago, I would have said people could become AI engineers who just implement things,” he says. “But that layer is rapidly being taken over.” Essentially, simply knowing how to operate AI tools is unlikely to be a stable career strategy.

At the same time, Ravindran pushes back against the fear that AI will wipe out jobs entirely. Instead, he sees a shift in how work is structured. Smaller teams, supported by AI, will be able to build more with fewer resources.

“A small set of high-quality programmers working with an AI workforce can build solutions at a lower cost,” he explains. This could even unlock new opportunities – such as highly specialised software built for small groups of users that were previously ignored because they were not profitable enough.