A Mumbai-based entrepreneur has now sparked a widespread online debate on the productivity advances introduced by AI in daily workflow. The Indian entrepreneur has pointed out a surprising downside of using AI tools in daily work, stating that some tasks that once took just 30 minutes now stretch to 2 hours, while others that used to take 2 hours are completed in 30 minutes.
Mustafa Yusuf, founder of Msquare Labs, shared his observation in a post on X (formerly Twitter), stating, “What used to take me 30 minutes now takes me two hours with AI. Also, what used to take me 2 hours now takes me 30 minutes with AI.” The candid post quickly went viral, resonating with many in the tech and startup community who have experienced similar mixed results from AI integration.
Many are calling his observation an “AI productivity paradox”, i.e., where AI accelerates well-defined, routine work but can slow down more complex tasks due to the extra effort required in prompting, context-building, and iteration.
AI slows down certain tasks
According to Yusuf and other commenters who shared similar experiences, the slowdown often occurs on tasks that lack a clear structure. Users end up spending significant time “arguing” with the AI model refining prompts, correcting outputs, or providing more context, only to realise manual completion would have been faster. One user noted, “The 30-minute tasks that now take two hours are always the ones where you spend 90 minutes arguing with the AI before realising you should’ve just done it yourself.”
Most users agreed that AI shines on short and clearly defined tasks because they require minimal setup. For longer and more complex projects, which you usually find common in creative, strategic, or exploratory work, the focus shifts from execution to preparation, such as crafting detailed prompts or verifying AI suggestions.
One commenter explained, “AI is great at the 30-minute tasks because they’re well-defined. The two-hour tasks were slow because they were ambiguous, and now you’re just arguing with the model instead of thinking.” Another added, “The stuff AI is fast at used to be the boring parts. Now I spend most of my time on context setup and prompt architecture instead of actually writing code… still net positive.”
AI at work requires more efforts to train
The debate on AI in workflow continues to bring up several drawbacks, with one of them being the need to put more effort into doing the job. While today’s large language models promise smartness and efficiency at work, they still need a major input from humans – directions from humans. While these tools are easier for straightforward tasks, they require being told how to do certain things in a certain way when complex tasks are involved.
Moreover, early adopters often face a learning curve and unexpected friction in most workflows. Despite these issues, many agree that the overall effect of AI at work remains positive in the long run, especially as AI handles repetitive elements quickly and frees up time for higher-value thinking once users master effective prompting.
Hence, incidents like these clarify that AI can only reshape the workflow to get the job done faster, but it still requires efforts and inputs from humans entirely.