Skip to Content

India Warily Eyes AI

Technology outsourcing has been India’s only reliable job creator in the past 30 years. Now artificial intelligence threatens to wipe out those gains.
Dadu Shin

Two days after K.S. Sunil Kumar received a promotion, Human Resources phoned him up and asked him to resign.

This happened in April, just as Kumar was beginning his ninth year at Tech Mahindra, one of the giants in India’s IT services industry. He worked in engineering services, where he designed components and tools for aerospace firms in North America and Europe. They’d send over specs—the materials available to construct a hinge, and the kind of load it had to bear, and the cost at which it had to be manufactured—and he mocked up options with the help of software. He was a foot soldier in the army of Indian engineers to whom work is outsourced from the West, so that it can be finished at a fraction of the expense. Sometimes he left his base, Tech Mahindra’s Bangalore campus, to serve stints at clients’ offices abroad: in Montreal, Belfast, or Stockholm.

When his employment was terminated, Kumar was earning close to $17,000 annually, a nice middle-class salary in India. Around the same time, Tech Mahindra announced profits for the previous financial year of $419 million, on revenue of $4.35 billion. (Tech Mahindra did not respond to a request for comment for this story.) IT services and related offerings in India record annual revenue of $154 billion and employ nearly four million people. The sector’s vigor has relied upon its ability to shave costs lower and lower—upon its ability to arbitrage the cheaply bought skills of workers like Sunil Kumar.

Bangalore is filled with IT professionals and engineers who resemble him. His curly hair is thin on top and gray at the temples; when we spoke he was wearing a faded checked Tommy Hilfiger shirt, a backpack, and a look of subdued anxiety. He grew up in a village a few hundred miles from Bangalore, where his father wove silk saris on a handloom. In 1995, when he was 15, he moved to Bangalore to study for a diploma in mechanical engineering; it was a step below a university degree, which he would gain later through a correspondence course.

Until he joined Tech Mahindra, in the summer of 2008, Kumar worked as a draftsman at an aerospace firm. The new job opened his life up the way the IT industry did for so many Indians, offering a chance to vault from a blue-collar background to a white-collar future. He got married, and the couple had a son; he took out a loan of $47,000 to buy a house, so that his parents and two brothers, who followed him to Bangalore, could stay with him. “I live a middle-class life,” he says. “I don’t want to showcase to people that I’m doing an IT job. Branded shirts, shoes—I don’t want them.”

The size and prestige of India’s IT industry are reflected in the design of Infosys campuses in Mysore (above) and Bangalore (below).
INDIAPICTURES/UIG VIA GETTY IMAGES

When Kumar lost his job, he became part of a wave of layoffs washing through the Indian IT industry—a term that includes, in its vastness, call centers, engineering services, business process outsourcing firms, and infrastructure management and software companies. The recent layoffs are part of the industry’s most significant period of churn since it began to boom two decades ago. Companies don’t necessarily attribute these layoffs directly to automation, but at the same time, they constantly identify automation as the spark for huge changes in the industry. Bots, machine learning, and algorithms that robotically execute processes are rendering old skills redundant, recasting the idea of work and making a smaller labor force seem likely.

An analysis by the business newspaper Mint reported that the top seven IT firms in India will lay off at least 56,000 employees this year. After its annual general meeting this summer, the $10 billion behemoth Infosys announced that 11,000 of its 200,000 employees had been “released” from repetitive tasks by automation and redeployed elsewhere within the company, the burden of their previous work now borne by algorithms. HfS Research, which studies the IT industry, predicted last year that automation will result in a net loss of 480,000 jobs in India by 2021. “If we sit still, there is absolutely no doubt that our jobs are going to be wiped out by AI,” Vishal Sikka said in March, when he was still the CEO of Infosys. (Sikka resigned in August.) “Sixty to 70 percent over the next 10 years—or maybe less than 10 years—of the jobs that we do today are going to be replaced by AI ... unless we continue to evolve ourselves.”

Automation could hit India particularly hard because much of its high-tech economy involves ­relatively routine work that is prime for computers to take over.

The fear that AI is pulverizing jobs is not unique to India, but automation could hit this country particularly hard because so much of its high-tech economy involves relatively routine work that is prime for computers to take over. In some cases, Indian IT services companies will automate the work themselves. In other cases, companies in the West will do it, so they no longer have to farm work out to humans in India.

Sunil Kumar wasn’t told, in any detail, why he was being let go; he insists that what he did at Tech Mahindra wasn’t automatable, and that he was fired as part of a drive to invigorate the bottom line. Devika Narayan, a sociologist at the University of Minnesota who is researching the subject for her doctoral thesis, thinks automation may indeed be getting too much blame for the loss of jobs like his. Companies might well be talking up automation to mask some of their own failures, or to distract from the ill effects of other factors beyond their control, she says. She points out that many IT giants are flabby and overstaffed, and that American companies are now wary, given the U.S. political climate, of sending work overseas. “The extent to which automation is being exaggerated is still unclear to me,” Narayan says. She suspects that Indian IT companies “want to leverage this automation narrative to undertake structural changes, particularly downsizing.’”

Where the truth lies is important for India. The IT industry may employ only a few million of India’s 1.3 billion people—but it has been a beacon for young men and women with aspirations. It motivated families to send their children to university, placed graduates in gleaming campuses, conferred independent urban lifestyles upon them, and provided stable incomes and access to the world outside India. Over the last 30 years, moreover, it has been the only industry in India to begin from seed and bloom to such success. India is otherwise struggling to create jobs: 12 million Indians enter the workforce every year, but only 135,000 jobs in the formal economy’s eight biggest sectors—including IT—were created in 2015. A dramatic contraction of the IT industry—a dimming of the beacon—would jolt the country’s economy and polity deeply.

Taking out the swivel chairs

Chetan Dube says he saw it coming. In 2005, Dube, the CEO of a company called IPsoft, was addressing a forum of IT companies in Mumbai. “If Indian industry does not wake up to the automation wave that is coming, then we will face an existential crisis,” he remembers telling the gathering. “I got chastised. The next day, we were having breakfast, and at the back of the Economic Times, an article said: ‘IPsoft CEO predicts the death of Indian outsourcing.’”

Bansal reckons that the IT sector hired 400,000 people annually until two or three years ago, and that the number has now shrunk to 140,000 to 160,000. Soon, he says, “net hiring will be barely above zero.”

Dube, a bow-tied and suspendered mathematician who once taught at New York University, founded IPsoft in 1998, but it was in 2014 that the company launched what is now its flagship product, Amelia. A conversational service agent, Amelia is designed to replace the humans who field customer queries in call centers and back offices. Amelia has been used to resolve suppliers’ questions for a large oil and gas firm; it runs the live chat service for SEB, a Swedish bank; it works in another bank’s team of mortgage brokers. For one client, Dube says, the average time to reach a conventional offshore agent used to be 55 seconds; an incarnation of Amelia made itself available in two seconds or less. An offshore agent needed 18.2 minutes, on average, to resolve a query successfully; for Amelia, it was 4.5 minutes. The customer service sphere is filling rapidly with products like Amelia—chatbots that, through text or voice, obviate the human presence altogether.

Only in a few cases has Amelia directly supplanted Indian workers, but Dube thinks that further change is inevitable. Call centers in India are already in flux: salaries have crept up, attrition has always been high, and firms like Infosys and Tata Consultancy Services have offshored some of their functions farther out to Manila, where labor costs are even lower than in India. (Three years ago, one official in an industry body named Assocham predicted gloomily that India would lose $30 billion in call center revenues to the Philippines over the following decade.) In the West, some companies are repatriating their voice service operations, while others are abandoning them altogether in favor of e-mail or chat help desks.

The prospect—or fear—of automation has thus become one more force reshaping the call center business. Voice recognition isn’t anywhere close to perfect yet, and even the sophisticated cognitive agents of the vaunted near future may not be able to parse rambling customers, complicated problems, or uncommonly thick accents. But most voice work is prosaic and repetitive. Given that humans in the first tier of this service calibrate their responses with the help of a script, their functions are among the simplest to transform into machine code.

Other fruit hangs similarly low elsewhere in the sector; as Dube says, “India is nothing but the blue-collar worker of IT,” so the lowest layer of work is plump with tasks that need diligence and stamina but not creativity or sharp technical skill.

At Genpact, a 20-year-old company that began in business process outsourcing before expanding into other services, there’s a lot of “swivel-chair work,” says Gianni Giacomelli, who leads the firm’s digital solutions business. The term captures the mechanical nature of these tasks. Until recently, a human has been required to deal with software systems that help with enterprise functions. Those systems are often disconnected from each other, so Genpact’s employees “are asked to process, very basically, things that come from one system and go into another system,” he says. “That toggling back and forth is a massive waste of time.” Since 2014, Genpact has been replacing workers in swivel chairs by ordering computers to take information from screens and servers and convey it into another system.

One level up is the kind of work that Giacomelli calls “reconciliation”: examining invoices and bills from a client’s various vendors and customers, with all their discrepancies and contradictions. It isn’t trivial work; it involves, right now, some grains of human judgment. “But once machines have seen enough of those things, they can do that kind of stuff,” he says.

Mayhem

For some of its clients, the IT colossus Infosys has been able to automate nearly all of the most routine chores of monitoring and maintaining their data infrastructure, says S. Ravi Kumar, the company’s deputy COO. Some intermediate work, such as triaging IT service requests, is now done by machines as well. At a still-higher level of service complexity—jobs that involve troubleshooting bugs deep within the code, or developing solutions to new problems—35 to 40 percent of tasks are performed by automated routines.

Overall, Somak Roy, an analyst at Forrester Research, estimates that only a quarter of the most easily automated work in India is being completed exclusively by machines. Companies are still enthusiastically dabbling in technologies that remain nascent. Nonetheless, Roy calls it a “distinct possibility” that IT will “cease to be a large-scale employment generator in India.”

One of the direst visions comes from Pankaj Bansal, the chief executive of PeopleStrong, a human-resources firm that frequently staffs IT companies with engineers. For IT services firms in the shape and form India has known them, Bansal says, “it will be mayhem.” He has been accused of fearmongering, but he holds fast to his assessment. Over the last two years, three or four out of every 10 jobs in the bottom layer of the pyramid of IT work have been “squashed” by automation, he says—and this has been manifested not in how many people have been laid off but in how sharply recruitment has dropped. Companies once gusted through the campuses of engineering colleges, picking them clean of fresh graduates. Bansal reckons that the IT sector hired 400,000 people annually until two or three years ago, and that the number has now shrunk to 140,000 to 160,000. Soon, he says, “net hiring will be barely above zero.”

There is a tension between the long arc of ­technological revolutions and the far shorter one of ­human lives.

Bansal’s prophesy of a deflating workforce may well come true for another reason. For years, IT firms hired inexpensive, hardworking youngsters wholesale—even if they were poorly skilled—because it made sense to staff projects heavily. The more warm bodies assigned to a task, the higher the bill that could be presented to the client. But the practice of calculating bills in this manner has waned; customers now pay for outcomes and impact. Meanwhile, the poorly skilled youngsters who stayed with their firms have received promotions and raises with clockwork regularity, until they’ve turned into mid-rung engineers who are now too costly, in their thousands, to sustain. Cue the purges.

Within the industry, Bansal’s grim views encounter profound disagreement, at least in public. Perhaps this is understandable: it has never been wise for companies to be effusive about the imminence of layoffs and workforce reductions. Sangeeta Gupta, a senior vice president at an industry body called the National Association of Software and Services Companies, predicts only a “decoupling” of revenues and head count over the next few years. If Indian IT required three million employees to touch $100 billion in annual revenue, she says, it will need only 1.2 million to two million additional people for its next $100 billion. By 2025, when revenues reach $350 billion, Gupta predicts, the sector will have added another 2.5 million to three million jobs to the four million it holds today.

Companies are eager to explain why automation won’t hollow out, and might even expand, their swarms of employees. For one thing, it isn’t as if machines can make people instantly redundant. “Jobs are not structured in such a clean way,” says Giacomelli, at Genpact. The architectures of modern work that have developed over decades all have human beings at their center; they rely upon people’s agility and their capacity to hold different things in their minds. “People do many things, so it’s not that easy to extricate one task or the other and make that happen through AI,” he says.

Companies also insist that they want to reskill the employees who risk being supplanted by automation. If an engineer’s work is best taken over by an algorithm, “it isn’t fair to then say ‘You don’t have a job,’” says K.M. Madhusudhan, the CTO of Mindtree, a services firm that employs more than 16,000 people. “Can we teach this engineer programming? Maybe not heavy lifting, but some scripting, which is not that difficult. For every role, we believe, there are adjacent, higher-level skills that can be acquired.” Madhusudhan calls this a “humane approach.” It will result in fewer job losses, although he acknowledges that firms like his will also be creating fewer jobs. “The numbers that were possible before will not be possible in the future,” he says. “That is the bigger concern for a country like India, because we still produce a lot of engineers, and not everyone will get a job.”

This is a familiar pattern in history: every technological stride forward has meant that the same amount of work can be done by fewer people. “Whenever there’s a revolution, there’s a worry about fewer jobs. It happened with the Industrial Revolution as well,” says Ravi Kumar, at Infosys. “The reality is, though, that there’s more consumption,” he adds. That eventually increases the need for new kinds of labor. At the moment, he says, enterprises spend 65 to 70 percent of their IT budgets “just to keep the lights on”—to pay for infrastructure and routine support. If that money is undammed, it may well pour into new—and as yet unimagined—streams of revenue and employment: “It would mean a much bigger canvas for us.”

But even if he is right, there is a tension between the long arc of these revolutions and the far shorter one of human lives. In the near term, people will lose their livelihoods. Sunil Kumar is still without a job.

In June, he filed a petition for wrongful dismissal with the office of the labor commissioner, a state body that resolves industrial disputes and enforces labor laws. Once, when he checked on its progress, an official advised him that his fight was likely to be a lengthy one, and now he suspects nothing will come of it. “Whatever confidence I had, I’m losing it,” he says. When he reads his newspapers, he stops just short of the business pages, which frustrate him. “There will be companies saying many things: ‘We’re hiring this many people, there are many opportunities.’ The CEOs keep saying it. I stopped reading all this,” he says. He knows he ought to start looking for a new job, but he hasn’t been able to pull himself together; it is as if his dismissal had stymied life itself. “I haven’t been able to concentrate on anything,” he says. “It’s very difficult now.”

Samanth Subramanian has written for such publications as Wired, the New York Times Magazine, and the New YorkerHis most recent book is This Divided Island: Life, Death, and the Sri Lankan War.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.