How we measure AI resolutions

In short

• A conversation only counts as AI-resolved when no human agent replied in it.
• A conversation the customer abandoned is closed by timeout, never counted.
• Every resolution carries a 3-day settling window.
• Reopened, escalated or rated badly inside that window: reversed and credited back.
• The ~70% we quote is what well-documented teams typically reach — not a promise.

01A resolution means zero human replies

A conversation only counts as AI-resolved when no human agent sent a single reply in it. We verify this against the actual author of every message in the conversation — if your team touched it, it is a human resolution and it is never billed as an AI one.

02Silence is not success

A conversation counts as AI-resolved when the AI answered the customer's question and no human agent had to step in. Conversations the customer never engaged with are closed by timeout without being counted. And if a resolution turns out wrong, the settling window reverses it (below).

03Resolutions can un-count themselves

Every resolution has a 3-day settling window. If within those 3 days the customer reopens the conversation, a human agent takes over, or the customer leaves a negative rating, the resolution is automatically reversed — removed from your usage and, if already invoiced, credited back.

04Where the ~70% figure comes from

The resolution share we quote reflects what teams with well-documented knowledge bases typically reach across routine ticket volume. Your number depends directly on how much of your ticket volume is answerable from documented knowledge: new workspaces with thin content start lower and climb as coverage grows.

05Why we bill this way

Per-resolution billing only works if the definition is honest — we only earn when the AI genuinely did the work alone, and every counted resolution is visible in your dashboard with its full conversation history, so you can audit any of them.