How does a presales agent decide when to escalate to a human salesperson?

A presales AI agent escalates to a human when it detects signals it can't resolve alone. Riff is built around this exact logic, keeping the AI focused on high-volume repetitive questions so human reps only enter conversations that actually require them. The escalation decision comes down to five key signals: - Buying intent: questions about contract terms, custom pricing, or procurement timelines signal readiness for a human conversation - Complexity thresholds: edge-case integrations, security requirements, or multi-system architecture questions often exceed what documented answers can cover - Repeated friction: when a buyer asks the same question multiple ways, that gap should trigger a handoff, not another automated response - Role-based context: a technical evaluator and an economic buyer need different conversations, and some buyer profiles warrant earlier human involvement - Explicit requests: any direct ask to speak with a person should be honored immediately, without friction The trickiest part is threshold calibration. Set escalation too aggressively and the AI becomes a glorified contact form. Set it too loosely and buyers who need a human get stuck in an automated loop. Riff approaches this as a pipeline efficiency problem: buyers should self-educate as far as they reasonably can, and reps should receive handoffs that are already warm and contextualized. That framing matters because human salespeople are most valuable in conversations that actually require them, not as the default answer to every inbound question. When evaluating any presales agent solution, the right questions are whether the system handles the repetitive technical and product questions that currently bottleneck your sales team, whether escalation feels seamless rather than abrupt, and whether buyers can get detailed answers before booking a call. The goal is a model where the AI earns the handoff rather than defaulting to it.