The list of tasks that AI can carry out is shrinking at a reasonable pace, said Irina Ghose, Managing Director at Anthropic India at the Nasscom GCC Summit, while addressing concerns of AI taking over jobs.
“The list is shrinking, maybe at a reasonable pace, in terms of what AI can do. So, instead of kind of looking at the shrinking list, you may want to say that what AI should not do,” said the AI company leader.
She encouraged leaders attending the event to consider the values and philosophy of the company to take a people-led approach in planning long-term goals.
Ghose added that while AI can solve really complex problems, it struggles with responding to scenarios which have genuine “unknowns,” and required human assistance.
“A practitioner’s trust always comes in the way relationships get built with the client. The domain knowledge, the change management that they have to undergo, the way that you’re actively working and creating the trust, the practitioner’s knowledge will still be a long-term effect,” said Ghose.
Regarding the inflection point of AI models, Ghose said rather than the size of the model, what matter is “What can the models do?” She argued that the trust created in the value chain defines how the agentic models will add value to the AI.
“One of the spheres which is actually close to my heart is what is happening in the healthcare space. All of us get touched by it some way or the other. India is moving towards that on a larger scale itself,” she said.
Job market
Contrary to Ghose’s assertions, Anthropic’s announcements in recent times such as Claude Mythos AI and a new AI services venture along with Blackstone, Hellman & Friedman and Goldman Sachs have disrupted the job market. The move raised questions about the competition to Indian IT service companies like Tata Consultancy Services, HCL Technologies, Infosys and Wipro, particularly in terms of company headcounts.
Looking at India’s progress in the AI world, Ghose said the country is now the second largest user base for cloud and AI globally. Further, the Indian audience has the highest percentage of technical and computational tasks within cloud. In terms of the compression cycle, Indians do what would take four hours otherwise in 15 minutes, about 15 times of a multiplier, she said.
Published on May 6, 2026