Axios
Follow
I let agents plan a WNBA finals trip (but kept my wallet to myself)
The author tested two new personal AI agents, OpenAI's Dots and Meta's Muse, by having them plan a trip to the WNBA Finals, organize a brunch, and create souvenir programs. These AI agents are designed to assist with various tasks, but current capabilities still fall short of a human's speed and quality for certain actions. Both agents demonstrated impressive ability to handle complex, multi-step assignments like booking travel and tickets. They were also creative, producing elaborate content such as recipe booklets and invitations, though these required human editing. Dots and Muse generated detailed WNBA souvenir programs, with Dots' being more visually appealing. However, both struggled with directly executing purchases, with Muse getting stuck on a sign-in page. Dots' less automated approach for shopping proved faster by directing the user to a pre-configured app. A significant hurdle is user trust; the author was not comfortable granting the AI agents full access to financial accounts for purchases or consequential decisions. Ultimately, the author finds the AI agents most useful for research and legwork, with the human making the final choices. Despite not completing the planned events, the author did purchase WNBA basketball cards for continued research.