Glossary
Auto-resolution
Definition
Auto-resolution is when an AI agent completes a support conversation end to end - correct answer given or action executed - without any human involvement.
A true auto-resolution has three parts: the agent understood the request (intent), it did the thing (a grounded answer from real data, or an executed action like a refund), and the conversation closed without a human stepping in or the customer bouncing back.
What earns the right to auto-resolve is infrastructure, not model quality alone: live data access, action guardrails (allow-lists, idempotency, approval thresholds), confidence scoring, and a clean escalation path for everything below the bar.
Why it matters
Auto-resolutions are the unit of value in support automation - they are the tickets your team never sees. They are also the unit you should be billed in, which keeps vendor incentives pointed at finishing tickets rather than replying to them.
How Resolvo HQ approaches it
Resolvo HQ auto-resolves only above a confidence threshold you control, never auto-executes destructive actions below it, bills only billable resolutions (never escalations or clarifications), and logs every action with an idempotency key so nothing runs twice.
Frequently asked
What share of e-commerce tickets can realistically auto-resolve?+
The transactional core - order status, tracking, simple returns and cancellations - which for most DTC stores is the majority of volume. The precise number depends on your mix, which is why replaying your own historical tickets beats any benchmark.
What stops an auto-resolving agent from doing something destructive?+
Guardrails: per-intent action allow-lists, no destructive actions below a confidence bar, idempotency keys so a retry cannot double-refund, and full audit logs. Ask any vendor to show you these before connecting your store.
See it on your own tickets
Replay real tickets in a sandbox and watch the agent resolve them before you pay.