operating-model / loops/renter-loop/engagement.md
id: renter-loop.engagementtype: domainstatus: activeversion: 2.1loop: renter-loop

Domain · Conversational engagement

What it covers

Keeping each renter's conversation alive, on-context, and progressing — between the matching surfaces above and the conversion-path operations below. Includes proactive nurture, reactive question-handling, and recovery of conversations that have stalled. This is also where the renter profile is maintained as a live artefact across the conversation.

Activities

  • Maintaining the renter profile and matching state as the conversation evolves; capturing every preference signal back into the profile.
  • Conversational nurture — keeping the conversation warm with context-aware touches, prompts, and relevant content.
  • Personal handling of high-value, stuck, or complex conversations.
  • Answering renter questions in-conversation, with escalation to personal handling where confidence is low.
  • Handling objections and resolving blockers.
  • Diagnosing and acting on stalled conversations.

Operations

Conversational nurture operation

Signal: Active conversation in a stage longer than threshold without progress. Steps: Context-appropriate nurture touch generated (could be a question, a relevant unit, a piece of useful content, a timely prompt) → delivered into the conversation through whatever medium the renter last engaged on → engagement signals captured. Closed state: Nurture touch delivered; engagement captured. Cadence: Continuous. Different from broadcast email nurture: every touch is in-conversation and aware of state.

In-conversation Q&A operation

Signal: Renter asks a question in the conversation (any medium). Steps: Intent classified → answered automatically where the answer is known and confidence is high → escalated to personal handling otherwise → answer added to the conversation thread → response delivered through the renter's medium of preference. Closed state: Answer delivered. Cadence: Event-driven.

Stuck-conversation operation

Signal: Conversation sits in a stage longer than threshold. Steps: Flagged for personal review → diagnosis (renter went cold, blocker exists, automation missed something) → action taken (re-engage, re-match, escalate) or conversation archived with reason. Closed state: Conversation re-engaged, re-matched, escalated, or archived with reason. Cadence: Weekly.

Notes

Engagement is where the concierge promise gets delivered conversation-by-conversation. Every personally-handled conversation should produce a question: can this be folded into the automation next time? Personal handling is information about where the system is incomplete. Manual work that isn't learned from is just more work.

On partner (White Label) surfaces, these operations run identically but the outbound messages are co-branded. The loop mechanics do not change.

Comment on GitHub →