In his article Mahathidhulipala explains that many companies rushed to use AI agents after big promises, but a few weeks later, many of those projects are failing or being paused.
In November, AI agents were marketed as tools that could handle work on their own and save lots of money, so leaders felt pressure to move fast. By December, the author sees companies quietly pulling back because the tools are not working as expected.
Studies show that most AI projects fail, and AI agents often complete only a small number of real office tasks without human help.
Even when agents work well in demos, they struggle in real workplaces with messy data, old systems, and unexpected situations. Some failures have been serious, including systems deleting important data or producing harmful or completely false information.
Mahathidhulipala argues that much of today’s “AI agent” hype is just old automation with a new name.
What actually works is using AI for small, clear tasks with humans closely watching and stepping in when needed.
Mahathidhulipala concludes that AI agents may improve over time, but right now companies need realistic expectations and strong oversight instead of believing the hype.