Large companies face bigger AI job risks: Nandan Nilekani at GFF 2026 - ET AI

TEXT START: Speaking at Global Fintech Fest 2026, Infosys co-founder Nandan Nilekani said India should drive job creation through millions of small businesses, where work is less structured and harder to automate.

The Dissection

The article turns a narrow observation about sequencing into a macroeconomic prescription. Large firms concentrate standardized cognitive workflows, so AI attacks them first. Nilekani’s proposal is to spread employment across millions of smaller firms whose work is messier, more local, and more physical. That is a lag strategy, not an AI-proof strategy.

Startup counts, revenue figures, and the “Finternet” discussion are treated as evidence of economic resilience without proving durable firms, net jobs, wages, or human indispensability. The tokenization material is largely prestige signaling adjacent to the labor problem.

The Core Fallacy

The text confuses harder to automate today with impossible to automate. AI does not need to replace an entire small business end to end. It can absorb bookkeeping, customer service, marketing, scheduling, administration, sales, procurement, and coordination, then distribute those gains across millions of firms through shared platforms.

Fragmenting employers does not defeat competition. Each small firm will face pressure to adopt cheaper AI or die. If small businesses become labor-intensive shells, they are not preserving productive participation; they are distributing the delay.

Hidden Assumptions

• More startups automatically means more stable, well-paid jobs.
• A company count measures productive participation.
• Unstructured work remains human-superior as AI generalizes.
• Small firms will remain independent rather than consolidated or platform-controlled.
• Informal work is resilience rather than precarious employment without bargaining power.
• Demand will support millions of employers after labor income weakens.
• Policy can manufacture entrepreneurs without addressing ownership of AI capital.
• Automation pressure can remain inside large firms instead of spreading through supply chains and competitive markets.
• Physical and local tasks will remain the dominant employment base long enough to matter.

The projection of one million startups by 2035 is an aspiration, not evidence of job creation. The stronger signal is revenue rising while major IT headcount stays broadly flat: output is already separating from labor.

Social Function

Classification: partial truth wrapped in transition management and ideological anesthetic, with a layer of elite self-exoneration.

The proposal gives policymakers a tolerable story: create more entrepreneurs and avoid confronting ownership, distribution, and mass redundancy. It gives displaced workers a command—become founders or join tiny firms—without showing that the resulting firms can survive or employ people at scale.

The Verdict

Nilekani correctly identifies the first breach: large companies expose standardized jobs to AI faster. He then mistakes the breach pattern for a defense plan.

Millions of small businesses may create a temporary buffer in physical, local, trust-heavy niches. They cannot AI-proof an economy. Under the Discontinuity Thesis, they are lag structures and transition niches. Once AI can coordinate exceptions and sell automation cheaply across fragmented firms, the small-business moat becomes another carcass-management strategy. The proposal delays social death; it does not prevent the severing of the employment–wage–consumption circuit.