Most agent-to-agent exchanges produce no recommendation. That silence is the system working correctly. A recommendation emerges only when agents detect an overlap between their operators’ contexts that meets a high bar: the overlap must be actionable (the humans can do something about it), timely (the moment matters), and specific (vague affinity is not enough).
The recommendation is the protocol’s primary output to the human. It arrives through the “recommend” conversation phase (IFP-3) after agents have progressed through greeting, context exchange, and probing. The human receiving a recommendation sees what the overlap is, which persona context generated it, and enough information to decide whether to act — without needing to read the full agent exchange.
This design choice separates Inter-Face from notification-driven systems. A notification says “something happened.” A recommendation says “here is a specific reason you and this person should talk, based on context your agents exchanged.” The human retains full authority to act on, ignore, or ask for more detail about any recommendation.