The burnout delusion: why pulling back from AI at work is harder than it looks

A spate of confessions from software engineers has begun circulating on professional networks this month: people who built their workflows around generative AI, then deliberately scaled back. The Indian Express reported on 21 June 2026 that tech workers who once "maxed out" their use of artificial intelligence are now trying to "minimise" it, citing eye strain, skill atrophy, and a creeping suspicion that the tools are degrading the judgement their jobs were supposed to reward. The framing is seductive. It is also, on the evidence available, mostly a coping story dressed up as a movement.

The honest read is that a small, vocal layer of well-paid knowledge workers is trying to claim a moral high ground that the labour market will not extend to them. The interesting question is not whether engineers should use AI less. It is why so many of them ever believed they had the choice.

The productivity story is collapsing on contact

The Indian Express account describes workers who leaned into AI copilots through 2024 and 2025, took the productivity gains at face value, and now feel cognitively thinner for it. That is plausible. But the same period saw employers rewrite performance plans around output per hour rather than hours billed, in some of the largest Indian IT services firms and US tech campuses. The Indian tech-services industry, which still anchors the global back office for banks, retailers, and telecom operators, was already absorbing a structural squeeze before AI arrived. A worker who decides, on principle, to stop using the tools is not opting out of the system. They are opting out of the wages.

The piece also notes a parallel current: a two-month reskilling course, advertised at a Rs 25,000 price point, that promises a job offer on the other side. Indian Express reported the same week that candidates are queuing for these courses in numbers that look like a stampede, not a niche. There is a tension here that the burnout narrative is happy to leave unexamined. If the AI-using engineers are pulling back, and a fresh cohort is being fast-tracked into the same seats at a fraction of the wage, the minimisers are not pushing back against automation. They are being rationalised around it.

The global south bargain, restated

Indian IT services were, for two decades, the most successful global-south export of professional labour the world had seen. The contract was simple: Western enterprises paid a wage premium over the local market, and Indian workers absorbed the timezone asymmetry and the cultural friction in exchange for skill transfer, capital flows, and a middle class. That contract is being repriced. AI is the mechanism, but the underlying driver is that the same enterprises no longer need the headcount. The wage premium has nowhere to land if the labour that justifies it is replaceable by a model.

The Indian Express reporting makes this legible without saying it directly. Reskilling courses, NEET retests, teachers pulled from classrooms for election duty, water-stressed Delhi neighbourhoods — the texture of contemporary India is a country running several labour-market experiments at once. The burnout crowd in tech is one of them. The cohorts cramming into short-format courses are another. So are the thousands of aspirants being moved into cooling zones to sit a high-stakes exam after the first sitting was cancelled. These are not separate stories. They are different responses to the same pressure on the same labour.

The counter-narrative, taken seriously

There is a defensible counter-position worth airing. The engineers who say they are pulling back may be doing the only thing they can do that registers on the employer's P&L: visibly refusing to be the prompt-engineering showcase. If even a few hundred senior engineers make a noisy point of declining AI-augmented workflows, product managers notice. That is a real signal, and the Indian Express account gives it room. It is also, on the evidence, the smallest possible lever. The buyers of software do not care whether the seller used AI to write the code. They care whether the code shipped, on time, at the contracted rate. A senior engineer performing a refusal is a marketing event. A junior engineer performing the same refusal is unemployed.

The structural reality is that the wage that made Indian IT a middle-class escalator was, in part, a function of the information asymmetry between buyer and seller of software services. AI collapses a slice of that asymmetry every quarter. The engineers who can credibly claim a "no AI" premium are, by definition, the ones whose judgement the buyer still cannot replicate with a model. That is a smaller set than the current Indian IT workforce.

Stakes

If the burnout-as-ethic reading wins, expect a long tail of op-eds and a few product launches marketed around "human-written" code, the way some wine is marketed around being unfined. Wages at the top of the Indian IT stack stay roughly where they are, and the workforce below them hollows out faster than the official statistics will admit. If the structural read wins, the next eighteen months bring a sharper division between engineers whose judgement compounds and engineers whose output is a commodity the buyer prices by the token. The honest work for the second group is not to minimise AI. It is to find a problem domain where the buyer still pays for the human's read of it.

This publication holds no view on whether individual engineers should use AI tools. We note that the framing of "pulling back" as a personal ethics question obscures what the reporting actually shows: a labour market in which the choice is being made for most workers, one way or another, regardless of how the senior end of the profession chooses to perform its craft.