When you first hear the term ‘agentic AI’, it might sound like we’re on the brink of machines taking over all our tasks. In reality, this buzzword merely repackages an old ambition without offering any new technical breakthrough. By painting a picture of imminently autonomous machines, it distracts us from the proven value of today’s generative and predictive AI.
There’s been a lot of chatter about the dream of one day achieving artificial general intelligence, where machines could do everything a human can. But if you’ve ever wrestled with the practical challenges of automation, you’ll recognise that long before machines ever go off on their own, they’re still very much a work in progress. Whether it’s machine learning, predictive modelling, or autonomous driving, these technologies represent targeted advances—not the sweeping leap hinted at by the term ‘agentic.’
Recent experiments, such as the Carnegie Mellon University and Amazon collaboration testing large language models (LLMs) for autonomous task management, have shown mixed results. For instance, if a model fails nearly half of its attempts at simple tasks, it’s clear that we have a long way to go before computers can truly operate without human oversight.
The temptation to attribute human qualities to machines is understandable, but it only adds to unrealistic expectations and unwarranted fears. Instead of chasing an elusive promise, it’s more worthwhile to focus on steady, incremental improvements that bring tangible benefits.
If you’re feeling bogged down by all the hype, take heart in knowing that practical, reliable advancements are where the real progress happens. By keeping our expectations grounded, we can better navigate the evolving landscape of artificial intelligence without getting sidetracked by lofty buzzwords.
