Bots are calling and closing loans at Bajaj Finance. Xiaomi produces smartphones in a lights out plant with no workers on site. AI agents at LimeChat and Haptik handle 95 percent of customer queries, cutting staffing by four-fifths. Hyundai plans more than 25,000 Boston Dynamics Atlas robots across its plants by 2030. TCS, which has been shedding workforce, has set up a dark factory lab in Pune. Meituan’s Keeta Drone makes food deliveries in Dubai. Salesforce has trimmed over 4,000 jobs as it has built what it calls an “Agentforce” of customer service bots. At IBM, an AI agent called AskHR now handles about 94 percent of routine HR tasks. AI has withered the hiring process, with candidates now applying to and being interviewed by AI as the first filter of recruitment.
In Kazuo Ishiguro's Klara and the Sun, the machine studies the humans it serves more closely than they study themselves. AI is retrenching human interface across sectors. Last month I asked Claude, Grok, ChatGPT, DeepSeek, Perplexity, Kimi and HY4 to map an atlas of jobs at risk. They were given the same brief: map, occupation by occupation, which jobs AI eliminates, compresses, transforms, augments, spares or creates, and analyse the investment and business models behind it. The bots didn’t need research assistants to produce over 160 pages. The coming compression of the human footprint in the job market is stark. Jobs at greatest risk include the data-entry clerk, telemarketer, call-centre agent, bookkeeper, paralegal, junior coder and claims processor.
On the safe-for-now list are the electrician, plumber, bedside nurse, carer, firefighter and those roles that need a human to own the call. AI reads the scan, but the radiologist stays because regulation requires a human sign-off. And all seven were silent on two things. None put a figure on the money driving the disruption. None estimated how many new jobs AI would create. The jobs AI removes were scored to the decimal. The machines also agreed on the sequence, and the sequence explains why the damage is invisible until it is done.
AI does not take jobs. It takes over tasks first, and jobs follow. The machine learns to read—invoice, claim, complaint. It drafts replies, codes and summaries. AI-plus human becomes the norm. Ten customer service agents are shrunk to two-plus AI. Five bookkeepers become one controller. Hiring freezes first, followed by headcount cuts. Often the ones on the roster are longer doing the work but catching the machine when it errs.
The hierarchy of risk shows up in data at the entry level. Stanford Digital Economy Lab’s research titled, Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence, found employment of 22-to-25-year-olds in AI-exposed occupations running “about 19 percent below where it would have been. Across the G7, youth unemployment averaged 10 percent in July: 20.7 percent in France, 18.9 percent in Italy, 16.4 percent in Britain. In India, the jobs most at risk are the ones a generation was told to chase: the BPO seat, testing desk, back-office ledger, junior coder's cubicle. TCS has pared its campus intake, and its chairman expects “half a million AI agents” to sit alongside half a million employees. In India, the pink slip is mostly an offer letter never sent.
Dario Amodei, who runs Anthropic, issued two warnings. In May 2025, he said AI could “wipe out half of all entry-level white-collar jobs” and push unemployment to 10-20 percent within five years. Last month he told the UN Security Council that, “if managed poorly”, AI “could be a risk to humanity as a whole”. Governments heard the second warning—it even got a Security Council session. The warning about job loss manifesting across economies got a shrug. It is easier to legislate against an apocalypse than to budget for a 22-year-old without a job.
Investments signal the magnitude of disruption. Goldman Sachs Research expects global AI investment to top $1 trillion this year. Capital at scale does not wait for linear evolution of business models. The investment cycle turns tech capability into headcount decisions. What AI can do decides little. What it can do profitably, at scale, decides everything. Escape and relief are rare. In April, a court in Hangzhou ruled that a fintech firm had unlawfully dismissed a worker replaced by AI and ordered it to pay 260,000 yuan.
The potential for retrenchment of human interface is real. MIT/Oak Ridge’s Iceberg Index estimates 11.7 percent of US workforce tasks ($1.2 trillion in wages) as already automatable. The IMF estimates AI will affect roughly 40 percent of jobs globally and 60 percent in advanced economies; the ILO finds that about one in four workers worldwide is exposed and vulnerable to generative AI.
The impact of pink slips is still unfolding, and the consequences are grave. Job losses hit incomes, shrink revenues, dent spending and weaken growth. As the tax base thins, the welfare bill is bound to rise—already 81 crore people in India draw free grain. The room to pay for it is thin. In the G7, public debt runs at 126 percent of GDP in the US, 138 percent in Italy, 118 percent in France and 204 percent in Japan; India's general government debt is near 80 percent. A state that borrows to pay interest cannot easily borrow to pay the unemployed, and no one has worked out how to tax the software that replaced the taxpayer.
India's Parliament spent its monsoon session debating how many members the Lok Sabha should have. The politicians are counting seats. So are the machines—the ones at the desk. The machines have decided whose jobs go first. The voters will decide who goes next. The 21-plus cohort denied the first job is bound to identify the politicians who did not notice. Power in democracies rests with the masses.
Read all columns by Shankkar Aiyar
Shankkar Aiyar
THE THIRD EYE | Author of The Gated Republic, Aadhaar: A Biometric History of India’s 12 Digit Revolution, and Accidental India