A recent experience challenged the notion that ‘AI doesn’t do anything useful yet’. While reviewing a task that required opening six different apps and copying-pasting information between them, a chatbot sat in a seventh tab providing summaries that still needed manual action. This is where the ‘AI is useless’ crowd has a point – the work still falls on the user.

However, switching to a desktop agent that could read Gmail, calendar, and Slack within the same task and take the next step itself, with a permission prompt before each action, changed the perspective. This shift from a chat window to a more integrated AI solution made all the difference.

This experience echoes the $500M-wasted-on-Claude thread, highlighting the issue of paying for AI tokens that generate paragraphs without leading to actual execution. The key to unlocking AI’s potential lies in its ability to finish tasks, not just describe them.

Photo by Tom Swinnen on Pexels
Photos provided by Pexels