Artificial intelligence is creating a peculiar problem for business: the more efficiently a company works, the less it may earn from each unit of work. A factory that produces more than it can sell gains little. A bank that automates fraud detection but overwhelms investigators with false alarms saves little. An IT company that delivers software faster may simply give its clients grounds to demand lower prices. India’s companies cannot afford to ignore AI. Rivals that cut costs and deliver faster will win customers from slower competitors. Yet adoption is only the beginning.
The harder task is to convert faster work into lower total costs, better products and stronger returns. If every competitor uses the same technology to produce the same output more cheaply, competition will push prices down and erode the initial advantage. This is the economic test of AI productivity. Companies must move quickly enough to defend their market share, while changing how they organise work, price their services and capture the value they create. The danger is not that AI fails to make employees faster. It is that businesses become more efficient while the economic gains flow to their customers, their competitors or the suppliers of the technology.
Adoption is not productivity
India is adopting AI quickly. EY’s 2025 Work Reimagined Survey found that 88% of Indian employees used AI at work, while 37% used it daily. Yet these figures say little about the return on that investment. An employee can produce a report in half the time, but the company gains only if the report improves a decision, replaces other work or frees time for something more valuable. The distinction is economic, not semantic. A faster process can lower costs, but only if the expense of checking errors, preparing data and integrating systems does not consume the savings.
A customersupport team that answers more queries but fails to resolve them has increased throughput without necessarily improving service. The same applies to code that is written faster but takes longer to debug. The danger is that companies can measure AI adoption immediately while productivity gains require proof. Licences issued, employees trained and tasks automated are easy to count. Better decisions, lower total costs and higher margins require evidence over time. Confusing the first set of numbers with the second can make a disappointing investment look successful.
The value of getting it right
Banking offers a concrete test. AI can flag suspicious transactions and help investigators resolve disputed payments faster. But more alerts do not necessarily mean less fraud. If investigators spend hours pursuing false positives while genuine fraud slips through, the bank has automated detection without improving security. Block legitimate transactions and it also risks losing customers. The real return lies in fraud losses prevented and investigation costs saved, after accounting for false alarms, system expenses and disruption to genuine customers. Research shows why the distinction matters.
In a study published in the Quarterly Journal of Economics in 2025, Erik Brynjolfsson, Danielle Li and Lindsey Raymond examined AI assistance for more than 5,000 customer-support agents. They found that AI increased issues resolved per hour by 15% on average, with larger gains for less experienced workers. The tool helped spread the practices of stronger performers to colleagues who had less experience. This is a more useful measure of success than the volume of AI-generated text. The technology improved an outcome that mattered to customers. But the result also carries a warning: the gains varied across workers, and the most experienced agents saw much smaller improvements.
A productivity gain observed in one setting cannot simply be applied to every job. Management determines whether the gains endure. If an employee saves two hours and receives enough additional work to consume them, the organisation has increased output, but may not have improved the quality or value of what it delivers. If those hours are used to resolve difficult cases, prevent costly mistakes or serve customers better, the return can be greater. The difference depends on how managers redesign work and measure performance.
Indian it’s revenue problem
Indian IT services faces a more difficult test because AI threatens the way many firms earn money. Much of the industry has traditionally sold labour through billable hours or teams deployed on projects. AI allows developers to write and test code faster, automate routine maintenance and reduce the effort required to deliver software. That is good news for clients. For providers whose revenue depends on hours worked, it is more complicated. Suppose a project that once took ten weeks can now be delivered in six.
The client has a reasonable case for demanding a lower price. If the provider earns less because it supplies fewer hours, its productivity gain may come with lower revenue. If it continues charging for effort that AI has eliminated, it risks losing the client’s trust. The technology has made delivery cheaper, but the commercial contract may prevent the supplier from capturing the benefit. This is the central business problem that AI adoption figures miss. A company can become more efficient and still struggle to improve its profits if prices fall as quickly as costs. The solution is to change what it sells: reliable outcomes, faster delivery, specialist judgement and continuing business value rather than labour hours alone.
Firms that make this shift have a better chance of retaining some of the efficiency gains. Those that do not may find that their own technology erodes the economics of their business. The same principle applies beyond IT. Manufacturers must turn faster production into lower waste, better quality or higher margins. Banks must convert quicker investigations into fewer losses and lower costs. Across sectors, the question is whether AI changes the economics of the business or merely accelerates its existing processes.
The productivity dividend
EY estimates that generative AI could raise productivity in India’s organised sector by 2.61% by 2030, with the potential to transform 38 million jobs. These are projections, not gains already realised. A 2025 Crux study, covering 180 corporations across three industry verticals, projects a larger gain. Its precise estimate and methodology should be stated before the two projections are compared directly.
The larger economic prize will depend on whether companies change how they work, price their services and reward performance. AI may release hours, but those hours have no automatic value. They become productive when they reduce costs, improve decisions, create better services or allow employees to tackle work that was previously neglected. India’s companies can adopt AI faster than they can rewrite contracts, redesign jobs and change managerial incentives. That is the constraint.
The firms that treat AI as a cheaper substitute for human effort may see their prices and revenues fall together. Those that use it to deliver better outcomes have a chance to widen their margins. The risk for Indian business is not that AI fails to make workers faster. It is that companies become faster at doing work for which customers are paying less.
*Prof. Vikas Singh is a Professor at IIM Nagpur and Visiting Professor at the Indian School of Business (ISB). He teaches and writes on business, policy and India’s economic transformation.
(Views are personal)