End-to-end resolution

End-to-end resolution means an AI agent handles a customer query from first contact through to a final outcome, including any back-office steps, without handing the case to a human partway through.

It differs from partial automation, where an agent answers the easy part of a query and a human finishes the rest. End-to-end resolution covers the customer-facing reply and whatever happens behind it, such as updating a record or issuing a refund, so the case closes in one continuous interaction rather than reopening in another queue.

Frequently asked questions

How does end-to-end resolution differ from deflection?

Deflection measures whether a query stayed out of the human queue. End-to-end resolution measures whether the AI agent actually completed the case, back-office actions included. A query can be deflected without being resolved.

Does end-to-end resolution require back-office access?

Yes. Without access to the systems that record an outcome, such as a refund or an address change, an agent can only answer questions rather than close cases, which stops short of end-to-end resolution.

Why does end-to-end resolution matter in regulated industries?

A bank or lender needs an auditable record that a case was actually completed, not just answered. End-to-end resolution ties the customer-facing reply to the back-office action, which is what a regulator or auditor checks. Gradient Labs' guide on why resolution has to reach the back office sets out what separates a case that was closed from one that was finished.