Testing Personal AI Agents on a WNBA Finals Trip Reveals Promise and Limits

As technology companies race to develop digital assistants capable of handling everything from everyday shopping to complex travel arrangements, consumers are left testing the boundary between helpful automation and personal control. To evaluate the current state of personal artificial intelligence, two newly released assistants—OpenAI’s Dots and Meta’s Muse—were recently put through a series of practical tasks using a fantasy trip to the WNBA Finals as a testing ground.
The experiment was designed to measure whether these digital tools could match human efficiency and judgment across a variety of assignments. The scenario involved planning a trip to Atlanta to watch the Golden State Valkyries, organizing a purple-themed brunch watch party, creating souvenir programs, and shopping for WNBA merchandise and basketball cards. The cost structures for the two tools vary significantly: Muse is free for most uses with optional monthly subscription tiers of $20 and $100, while Dots requires a ChatGPT Pro subscription starting at $100 per month or access via an enterprise or business plan.
When evaluated on research and multi-step problem solving, both agents delivered impressive results. Tasked with finding flights and tickets for Game 3 in Atlanta, both assistants successfully recognized that the game’s schedule had been moved up by a day and that general public ticket sales had not yet opened, noting that currently available seats were restricted to the resale market at potentially inflated prices. Furthermore, both located viable flight options. Dots demonstrated cross-application awareness by flagging a scheduling conflict with an enrollment detail discovered through an AI-assisted search regarding San Francisco high schools. Muse assisted by opening a browser window to help navigate a ticket presale sign-up process.
In creative assignments, both platforms converted broad instructions into detailed deliverables, though neither functioned without human oversight. For the themed brunch, Muse generated a 10-page recipe booklet featuring dishes like ube waffles and lavender-honey butter, alongside a shopping list that contained minor discrepancies, such as calling for 12 eggs while requiring 13 in the recipes. Dots produced a more practical eight-page guide focused on cooking instructions and dietary preferences. Both agents also drafted invitations that required language and aesthetic revisions before being ready for distribution.
Additional creative tests involved producing 1980s-style souvenir programs and digital photo keepsakes. Dots assembled a 38-page vintage program featuring player statistics, photos, and graphics in 47 seconds, though its statistics contained questionable data. Muse created a 35-page alternative that included fictional advertisements alongside an ad for the imaginary brunch, later sending an unsolicited update to correct several factual errors. For the photo keepsake project, Muse proposed an interesting layout highlighting multiple generations of Valkyries fans, though it required significantly more processing time than Dots.
When the tasks shifted from planning to execution, both systems encountered practical hurdles. Attempting to purchase WNBA trading cards from Target—chosen to utilize a store credit card providing free shipping and a 5% discount—revealed limitations in automated commerce. Muse attempted to complete the transaction directly but stalled at the retailer sign-in page. Dots opted to direct the user to the mobile application where account credentials and payment methods were already saved, proving faster despite both systems supporting purchases requiring user approval.
Ultimately, the experiment underscored a persistent trust gap regarding financial transactions and autonomous decision-making. While both assistants demonstrated strong utility in handling research, itinerary groundwork, and complex data organization, authorization for high-value purchases or consequential choices remains firmly in human hands. Ultimately, neither trip nor brunch occurred due to scheduling conflicts, though the exercise concluded with the acquisition of three boxes of WNBA basketball cards.

