Calm before AI storm: A moment to prepare

In The Driver in the Driverless Car, I warned of a coming storm: A jobless future driven by automation and Artificial Intelligence (AI). That future is no longer on the horizon — it has begun. But unlike in the West, where entire white-collar industries are already being restructured by AI, India has a bit more time. Not because it is more insulated, but because so much of its infrastructure still needs to be built — smart cities, modern digital systems, and scalable platforms for a billion-plus people.

That breathing room is a gift. But treating it as a licence for complacency would be a costly mistake.

Let’s be clear: The foundational work done by India’s IT services industry — code maintenance, software patching, integration testing, report generation, and process documentation — is squarely in the crosshairs of the next wave of AI. That wave is now forming.

At a recent Google event, DeepMind CEO and co-founder Demis Hassabis laid out what’s coming. He warned that within a decade, we may reach artificial general intelligence — the holy grail of AI, where machines exhibit human-level cognition. But the disruption won’t wait that long. It is already underway in the form of agentic AI — software systems that don’t just answer questions, but break down problems, reason through them, and take action. These systems are becoming capable of handling complex workflows with minimal human oversight. That’s a polite way of saying: They can already do the kind of work that much of India’s IT sector depends on.

Google’s new Agentspace platform, rolling out this year, integrates tools like Gemini with enterprise data and provides company-branded agents that can search, decide, summarise, and execute — on command. Tools like NotebookLM can instantly synthesise vast troves of documents, emails, or manuals and surface insights that teams would take weeks to extract. These aren’t passive chatbots. They’re fully operational digital workers — reliable, tireless, and improving by the day.

Already, tools like GitHub Copilot are writing production-grade code. Salesforce’s Einstein AI is autonomously handling customer support and sales optimisation tasks. Within a few years, technologies like these will mature to the point where they can replace much of the repetitive, back-end work done by hundreds of thousands of professionals in IT.

The DeepMind CEO acknowledged that current AI models still suffer from what he calls “compounding error” — a 1% error rate multiplied over thousands of steps renders outcomes nearly random. Games like Go or chess, with fixed rules, allow algorithms to plan and reason reliably. The real world, with its ambiguity and complexity, doesn’t. That gap offers a brief window of opportunity.

India should be using this time to prepare.

The IT sector emerged not by building transformative technologies, but by filling the world’s need for outsourced back-office work — digitising insurance forms, fixing Y2K bugs, maintaining legacy code. But those business models won’t survive the AI transition.

The key question now is whether the country will remain a consumer of foreign-built AI systems — or emerge as a developer and exporter of its own.

This disruption presents massive opportunity. India has one of the world’s youngest and most tech-savvy populations, broad English proficiency, and a long history of leapfrogging traditional development pathways. It has already demonstrated what’s possible with Aadhaar, UPI, and ONDC — building platforms at population scale that solve real problems. But this next leap requires a mindset shift.

Agentic AI will need new interfaces, governance structures, fine-tuning tools, and orchestration platforms. It will need ethical frameworks, transparency layers, and sector-specific customisation. These are all areas where India can lead.

Smart cities, for example, could be designed from the ground up with autonomous agents managing traffic, energy usage, and public services in real time. The very AI that threatens to obsolete legacy IT jobs could drive a wave of hyper-efficient, AI-first urban development.

IT needs to pivot from outsourcing and services to intellectual property creation, developing AI co-pilots for health care, construction, law enforcement, and agriculture, and investing in open-source models, multimodal agent platforms, and domain-specific AI stacks.

Startups that train AI on India’s tax codes, regional languages, SME supply chains, and labour markets can outperform western competitors in local contexts.

The government can use its digital infrastructure as a testbed for next-gen AI systems: Predictive disease surveillance, AI-based procurement, real-time policy simulations, and automated dispute resolution.

The old ways won’t work. Engineering graduates can no longer coast on routine Java or testing jobs. Curricula need a ground-up redesign focused on prompt engineering, cognitive architectures, model tuning, and ethical AI. Students must train in real-world labs, not just on Zoom calls with remote teams.

India would benefit from launching a national initiative — akin to its semiconductor or space programmes — dedicated to building domestic capability in agentic AI. That means public-private partnerships, indigenous foundation models, and massive upskilling of the existing workforce. A five-year crash programme to prepare the next 10 million knowledge workers for a radically different tech industry.

If this transition is managed wisely, India could leapfrog into global leadership in applied AI — just as it did with fintech and telecom. But if it’s mismanaged, the result could be a hollowing out of the very industry that lifted millions into the middle class.

This is the calm before the storm — and the moment to prepare.

Vivek Wadhwa is CEO, Vionix Biosciences. The views expressed are personal