Meta, OpenAI and Uber develop AI agents that decide when to communicate

Meta’s Muse, OpenAI’s Dots and Uber’s driver assistant are pioneering proactive AI agents that decide when to communicate with users. These systems shift the focus from what to say to when, how and through which channel to deliver messages. The challenge lies in balancing the value of the message against the cost of interrupting the user. High-value, time-sensitive messages are sent immediately, while lower-value ones are delayed or omitted. Decision models like TypeSafe’s Jev and Supersonic Labs’ Julia 1, which run on CPUs and take about 33 ms per decision, are being used to evaluate messages. These models assess context, but they cannot learn user-specific behaviors like ignoring mornings. Uber’s outcome tracking provides the necessary labels for these models. Meta is exploring commerce through Muse, OpenAI announced Dots alongside a $500 monthly tier, and Uber promotes delivery services. All three companies emphasize that messages must maintain user trust, as perceived sales pitches can damage credibility. The race for attention will determine success, with companies leveraging years of notification data gaining an edge.