AI 'Agents' Still Learning How to Browse the Web Properly

You've probably heard about clever new AI tools, often called 'agents', that promise to handle tasks online for you, like booking appointments or finding information across multiple websites. It sounds fantastic, a real time-saver for busy small business owners or anyone juggling a lot of online errands. The idea is that these AI agents act like your digital assistant, working independently in your web browser.

However, new research from a company called Decodo, shared by TechRadar, offers a bit of a reality check. They put these AI agents to the test with real-world web browsing tasks, and it turns out they're not quite as capable as some marketing might suggest. For instance, none of the agents they tested scored perfectly, and many struggled with simple things like managing multiple open tabs, which is a pretty common part of how we use the internet every day.

This doesn't mean AI agents are useless, but it highlights that the technology is still developing. Some performed better than others – for example, the Claude for Chrome tool did better than the ChatGPT Chrome extension in certain tests. The takeaway is that while these tools are certainly going to change how we work, they're not yet at the stage where you can just set them loose on complex online tasks and expect flawless results every time.

For everyday Australians, especially those running small businesses, this research offers a sensible perspective. It means it's wise to be excited about AI's potential, but also to approach new tools with a bit of caution. Think of it like a new employee still learning the ropes – great potential, but needs supervision for now. This insight helps manage expectations and encourages us to understand the current limits as well as the future possibilities of AI.