Some Hard Truths About AI & India’s IT Future

Recently, I was discussing the prospects of Indian IT companies in the wake of the on-going AI revolution, with Sajid Malik — my friend & the co-founder of Ventura. He recommended that I watch a particular episode of dAIlogues by Stellaris featuring Dr. Vishal Sikka

𝗦𝗼𝗺𝗲 𝗲𝘅𝗰𝗲𝗿𝗽𝘁𝘀 𝗳𝗿𝗼𝗺 𝘁𝗵𝗶𝘀 𝗶𝗻𝘀𝗶𝗴𝗵𝘁𝗳𝘂𝗹 𝗶𝗻𝘁𝗲𝗿𝘃𝗶𝗲𝘄…

𝗢𝗻 𝘁𝗿𝗮𝗻𝘀𝗶𝘁𝗶𝗼𝗻𝗶𝗻𝗴 𝗳𝗿𝗼𝗺 𝘁𝗵𝗲 𝗼𝗹𝗱 𝗺𝗼𝗱𝗲𝗹
“The traditional model of IT companies involved deploying large teams on-site or off-site. Now, projects that once required 50 people can be completed by 5-10 people, with AI doing the rest. With this shift Indian IT firms must own their AI platforms because if the underlying AI comes from elsewhere, the value migrates there too. Building an ‘AI practice’ in-house could come at the cost of large existing projects getting disrupted. The first signs of this are already visible in margin compression across the sector.”

𝗢𝗻 𝗔𝗜’𝘀 𝗶𝗻𝗵𝗲𝗿𝗲𝗻𝘁 𝘂𝗻𝗿𝗲𝗹𝗶𝗮𝗯𝗶𝗹𝗶𝘁𝘆 & 𝗻𝗲𝗲𝗱 𝗳𝗼𝗿 𝗮 '𝗹𝗲𝗮𝘀𝗵’
“LLMs are not deterministic models. Typically, randomness is baked into them to keep things interesting. So when you use LLMs in enterprise systems, you inherit unreliability at the foundation. You have to build guardrails around AI so the system knows when it doesn’t know the answer.”

𝗢𝗻 𝗯𝗲𝗶𝗻𝗴 𝘂𝘀𝗲𝗱/𝗰𝗼𝗻𝘀𝘂𝗺𝗲𝗱 𝗯𝘆 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹𝘀
“Every time we interact with chat GPT or with Claude, we are enriching it with our own interaction. Rather than being the consumer, we are actually the consumed.”

𝗢𝗻 𝘀𝗲𝗲𝗶𝗻𝗴 𝗔𝗜 𝗳𝗼𝗿 𝘄𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀
“To say these systems are super-intelligent is actually worse than nonsense; it’s a disservice to what the thing is. LLMs are basically powerful lookup machines. Technology exceeded human capabilities a long time ago. But we’re still here; still in charge.”

𝗔𝗻𝗱 𝗳𝗶𝗻𝗮𝗹𝗹𝘆, 𝗼𝗻 𝘁𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 & 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝘀𝘂𝗰𝗰𝗲𝘀𝘀
“We get enamoured with technology — including AI — but finally, business success is about solving a problem that people care about & are willing to pay for.”

These excerpts are not completely verbatim. I recommend watching the entire interview to get a better sense of the context.

𝘋𝘪𝘴𝘤𝘭𝘢𝘪𝘮𝘦𝘳: 𝘛𝘩𝘦 𝘷𝘪𝘦𝘸𝘴 𝘦𝘹𝘱𝘳𝘦𝘴𝘴𝘦𝘥 𝘪𝘯 𝘲𝘶𝘰𝘵𝘦𝘴 𝘣𝘦𝘭𝘰𝘯𝘨 𝘵𝘰 𝘵𝘩𝘦 𝘪𝘯𝘵𝘦𝘳𝘷𝘪𝘦𝘸𝘦𝘦. 𝘕𝘰 𝘤𝘭𝘢𝘪𝘮 𝘰𝘧 𝘰𝘸𝘯𝘦𝘳𝘴𝘩𝘪𝘱 𝘰𝘳 𝘢𝘧𝘧𝘪𝘭𝘪𝘢𝘵𝘪𝘰𝘯 𝘸𝘪𝘵𝘩 𝘵𝘩𝘦 𝘰𝘳𝘪𝘨𝘪𝘯𝘢𝘭 𝘸𝘰𝘳𝘬 𝘪𝘴 𝘪𝘯𝘵𝘦𝘯𝘥𝘦𝘥 𝘰𝘳 𝘪𝘮𝘱𝘭𝘪𝘦𝘥. 𝘛𝘩𝘦𝘳𝘦 𝘪𝘴 𝘯𝘰 𝘶𝘯𝘦𝘹𝘱𝘳𝘦𝘴𝘴𝘦𝘥 𝘳𝘦𝘢𝘴𝘰𝘯 𝘵𝘰 𝘱𝘳𝘰𝘮𝘰𝘵𝘦 𝘵𝘩𝘦 𝘷𝘪𝘦𝘸𝘴 𝘰𝘧 𝘵𝘩𝘦 𝘲𝘶𝘰𝘵𝘦𝘥 𝘴𝘰𝘶𝘳𝘤𝘦 & 𝘯𝘰 𝘷𝘦𝘴𝘵𝘦𝘥 𝘪𝘯𝘵𝘦𝘳𝘦𝘴𝘵 𝘪𝘯 𝘥𝘰𝘪𝘯𝘨 𝘴𝘰, 𝘣𝘦𝘺𝘰𝘯𝘥 𝘧𝘢𝘤𝘪𝘭𝘪𝘵𝘢𝘵𝘪𝘯𝘨 𝘢 𝘵𝘩𝘰𝘶𝘨𝘩𝘵𝘧𝘶𝘭 𝘥𝘪𝘴𝘤𝘶𝘴𝘴𝘪𝘰𝘯 & 𝘦𝘹𝘤𝘩𝘢𝘯𝘨𝘦 𝘰𝘧 𝘪𝘥𝘦𝘢𝘴.