AI Tools Are Adding Work for Developers, Not Removing It

Software developers at major tech companies, including one working at a FAANG firm, told India Today Tech in August 2026 that AI coding tools like Claude Code and ChatGPT Codex have not reduced their working hours. Instead, faster output has prompted managers to pile on more tasks and compress timelines, with at least one developer reporting 15 to 16 hour days for three consecutive months. The gap between what AI was sold as and how it plays out on the ground is now a concrete, measurable problem for people building software every day.

What happened

Detail	Fact
Story published	August 18, 2026
AI tools cited	Claude Code, ChatGPT Codex
Token spend cap reported	$250 per month (approx. Rs 24,000)
Hours worked (one developer)	15-16 hours daily for 3 months
Manager deadline expectation	Tasks completed in 30-50% of previous time

India Today Tech spoke to multiple software developers across the Indian tech industry to document what actually happens when AI coding assistants are rolled out inside companies. The headline finding: AI does make individual tasks faster, but that speed is immediately absorbed by managers who assign more work or cut estimated delivery times in half.

Shubham, who works at a FAANG company (Facebook, Amazon, Apple, Nvidia, or Google) and declined to give his surname, put it plainly: “We are able to work quickly but the work has increased. So at the end we’re working for the same amount of time.”

Why managers are the main variable here

Ankita, a senior software developer in Bengaluru, says the pressure is explicit. Her managers now assume that any task with AI assistance takes half the time, sometimes 30 percent of the time. When output is faster, the response is not fewer tasks. It is more tasks scheduled into the same window.

Rajat, another senior developer, makes the structural point that free time at work is not really a category that exists. “If you have to work for 9 hours a day, you have to work for 9 hours,” he told India Today Tech. The idea that AI-created efficiency turns into personal leisure is, in his view, a misreading of how employment works.

Beyond task volume, staying current with AI tooling itself adds hours. Ankita describes attending in-office AI training sessions and workshops as a requirement, with regular project work starting only after those sessions finish.

What token caps mean in practice

There is a specific constraint that does not get enough attention in the broader AI productivity debate: usage caps. Ankita’s employer has set a hard spending limit of $250 worth of AI tokens per month. For intensive work, including proofs of concept, that ceiling is reached quickly, and the developer must either slow down or absorb the cost personally. Leadership, meanwhile, still expects AI-speed output regardless of whether the token budget has run out.

This creates a trap. Management sees AI as a fixed cost that justifies faster delivery commitments. Developers see a capped resource that runs dry mid-project. When results suffer, the blame lands on the employee rather than the policy. Businesses building AI integration strategies for their teams need to account for this: token budget and productivity expectations must be set together, not independently.

Is this unique to software developers?

No. Diya Jana, a graphic designer at Oddly Studios, describes the same pattern in a creative role. AI speeds up parts of her process, but the output requires prompt refinement and manual correction cycles. “It doesn’t necessarily reduce the overall workload,” she told India Today Tech. The efficiency gain is real but partial, and the remaining manual work still takes time.

This matches what we see in other sectors too. Our earlier coverage of how ChatGPT has shifted user behaviour shows AI changing where effort goes, not eliminating it entirely.

Our take

The Bill Gates “three days a week” vision from November 2023 was always a long-run structural argument, not a product promise. What got lost is that individual speed gains at the task level do not translate into free time inside a firm. They translate into higher throughput expectations from the same headcount. That is not a technology problem. It is a management and incentive problem.

For business owners thinking about rolling out AI tools to their teams, the honest question is: what will you actually do with the time saved? If the answer is “assign more work,” your people will burn out faster, not slower. The token cap issue is also underappreciated. A $250/month cap sounds reasonable until a developer burns through it in week two running experiments. Set the budget to match the actual workflow, not an idealized one.

If you are evaluating how AI tooling fits into your operations or workflows, it is worth getting a clear picture of the realistic capacity and cost tradeoffs before rolling anything out. Our team works through exactly these questions with clients across AI integration projects and broader workflow automation builds.

What to do about it

1. Audit how much time AI tools actually save before adjusting team capacity or deadlines. Track it for at least four weeks.
2. Set token or API budgets based on observed usage, not estimated usage. Build in a buffer for exploratory work.
3. Align management expectations with real AI output quality. Not every task completes cleanly without human review.
4. Account for upskilling time in workload planning. Training sessions and prompt engineering are working hours, not overhead.
5. Review whether speed gains are being captured as reduced cost or just absorbed as increased scope. Only one of those benefits the team.

The honest takeaway: AI tools shift where effort goes, not how much of it your team needs to put in.

Source: Bing News · Claude AI