AI Coding Tools Speed Up Work But Fuel ‘Performative Productivity’ Crisis: Report
Representational image (Getty Images)
New Delhi: Artificial intelligence (AI) is helping software developers deliver work faster, but the time saved through AI-assisted coding is not necessarily translating into higher productivity, as workers continue to spend hours reviewing machine-generated output and dealing with workplace bottlenecks, according to Australian software MNC Atlassian’s State of Developer Experience Report 2026.
The report, based on a survey of more than 1,500 software practitioners, found that 87 per cent of developers said AI helps them ship work faster, while 80 per cent said it enables them to deliver more without working longer hours. However, only 28 per cent reported experiencing noticeably less friction in their day-to-day workflow than before AI.
In Search of Lost Time
The gap between faster coding and actual productivity gains is reflected in the time developers say they save. The proportion reporting savings of at least 10 hours a week fell from 68 per cent in 2025 to 25 per cent in 2026. Meanwhile, 49 per cent said they spend four or more hours every week reviewing AI-generated output, correcting prompts and undertaking related tasks.
The findings also raise questions about workplace pressure surrounding AI adoption. While 64 per cent of leaders associated greater AI usage with better outcomes, 41 per cent of developers said they sometimes use AI more than necessary because they feel they are expected to do so. Only 17 per cent of developers and 23 per cent of leaders said their organisations had a clearly communicated and actively implemented AI strategy across the software development lifecycle.
Karnnika A Seth, legal advisor and cybersecurity expert to the Ministry of Electronics and Information Technology (MeitY) and Defence Research and Development Organisation (DRDO) under the Ministry of Defence, told ETV Bharat, “AI tools are being developed by some players just to remain in the news in the industry. It’s important to create good quality and responsible AI, and not with little or no innovation. It’s true many entities use external AI tools today, but a lot of manpower goes in correcting and sanitising prompt errors and building policies and systems that bar sensitive data or purge such information from AI tools.”
Ajay Sharma, an AI governance expert, told ETV Bharat that AI coding assistants can produce initial drafts, routine fixes, tests and documentation within seconds, but the output still requires human verification.
“AI coding assistants speed up the first draft: Boilerplate tests, documentation and routine fixes arrive in seconds, and developers can explore more approaches in less time. But the output is often only plausibly correct. It can invent APIs (Application Programming Interface), miss edge cases, introduce subtle security flaws, or ignore conventions of the existing codebase,” Sharma said.
Time Gained & Time Lost
He added that the time saved during coding can be lost during review, as checking code written by AI can be slower and more mentally demanding than writing it independently.
“The fix is to treat AI as a supervised collaborator, rather than an autonomous author,” Sharma said, emphasising the need for better context, automated testing, security checks and clear organisational policies on AI use.
Sharma also said wider AI adoption could create demand for code reviewers, verification engineers, AI governance and compliance specialists, prompt and context engineers, test automation experts and professionals who train teams to use AI effectively. As AI lowers the cost of software development, more businesses could undertake projects, increasing the need for people who can verify and take responsibility for the final output.
Drishyaa Duggal, an AI expert, described the pressure to demonstrate AI usage as “performative productivity”, arguing that employees may start prioritising the appearance of adoption over whether a tool actually helps them.
“When AI usage becomes a visible metric, developers stop asking ‘does this help me?’ and start asking ‘does this look like I’m adopting it?’ That’s a psychological shift from intrinsic problem-solving to impression management,” Duggal said.
She added that reviewing and correcting machine-generated output is cognitively demanding work that can quietly shift developers from creators to supervisors. When AI usage becomes part of how employees are evaluated, she said, they may become more concerned about appearing to use the technology than determining whether it improves their work.
The report identifies finding information as the biggest source of friction for developers, followed by meetings, context switching, technical debt and unclear requirements. It suggests that while AI can improve individual tasks, organisations may need to address these broader workflow problems to translate faster coding into meaningful productivity gains.
The findings underline a growing challenge for companies investing in AI: Measuring success through actual improvements in delivery, quality and efficiency, rather than simply how frequently employees use AI tools.
Also Read:
- CERT-In Warns of AI Agents Accessing Indian Firms’ Systems, Confidential Data
- AI Boom Could Trigger Sharp Equity Market Correction, Losses May ‘Spread Around The World’: IMF
- Anthropic Warns Of AI ‘Worms’ Spreading Across Smart Gadgets
- Amid Growing Concerns Over AI Consciousness, Anthropic Bans ‘Needlessly Cruel’ Behaviour Toward Claude
