5 phone jobs I still trust Gemini and Claude with after 8 months of agentic AI

Over the past eight months, I've been testing the capabilities of Gemini and Claude on my phone, assigning them various tasks and observing their performance. One simple task I consistently rely on them for is reordering my usual meal from Starbucks. By having them handle this routine, I can avoid the need to open the app and tap through various menus. They successfully complete this task, albeit with a final payment confirmation step that requires human input.

However, when it comes to handling more complex tasks that require a deeper understanding of context and decision-making, such as interpreting urgent emails or managing my calendar, I still prefer to have a human take the lead. Gemini and Claude are excellent at specific tasks, but they can make mistakes or overlook important details when faced with open-ended or nuanced requests.

One of the most interesting aspects I observed was Gemini's ability to create automations on its own after I stopped micromanaging it. By providing it with a goal, it can determine triggers, conditions, and actions needed to achieve that goal. While this proved useful in automating certain home tasks, it required a Google Home Premium subscription to fully utilize.

Personal Intelligence, Gemini's feature that allows it to reason across connected Google apps, is touted as a powerful tool for finding specific information across files and photos. While it can locate relevant data, it sometimes produces inaccurate results or is overly confident in its findings. For document-related tasks, I found Claude to be more reliable and efficient in organizing my Google Drive, moving files to appropriate folders and identifying duplicates.

Ultimately, the effectiveness of Gemini and Claude in handling various tasks depends on the nature of the job. Simple, routine tasks like reordering food are well-suited for these AI agents, while more complex or context-dependent tasks are better left to human oversight. I encourage readers to experiment with giving different chores to Gemini and Claude, observing their performance over time, and determining which tasks they handle best.