> ## Documentation Index
> Fetch the complete documentation index at: https://docs.keloa.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Resolving conversations

> Close conversations cleanly, collect CSAT, and know when to reopen.

Resolving a conversation closes it out — removes it from the open list, stops the SLA timer, and optionally triggers a CSAT survey.

## Resolve a single conversation

Press <kbd>e</kbd> or click **Resolve** in the conversation header.

You can add an optional **resolution type**:

* **Resolved** — the normal case
* **No action needed** — spam, mis-directed, nothing to do
* **Duplicate** — same as another conversation
* **Spam** — filter and never show again

## Bulk resolve

Select conversations in the list, then **Resolve** in the bulk bar. Useful after triage.

## Reopen

A resolved conversation reopens automatically when the contact sends a new message. You can also reopen manually:

* Open the conversation (use the **Resolved** filter or search) and press <kbd>r</kbd>.
* Or click **Reopen** in the header.

## CSAT (customer satisfaction)

If CSAT is enabled on a flow's **Resolve** node, a one-tap survey is sent after resolution — 👍 / 😐 / 👎 — and the result appears on the conversation and in analytics.

Enable CSAT per flow, or globally by asking support to turn it on for every resolved conversation.

<Tip>
  Don't CSAT everything. Send it on human-resolved conversations, skip it for "no action needed" and spam.
</Tip>

## Resolution time

The clock starts when the first inbound message lands and stops when you resolve. Shown in **Analytics → Response times**, and broken down by channel, tag, and assignee.

## When to resolve

* Customer confirmed the issue is sorted ("thanks, that worked!").
* You've given a final answer and there's no expected follow-up.
* The conversation is spam or misrouted.

## When *not* to resolve

* You're waiting on the customer. Use the tag `waiting-on-customer` instead, and keep it open.
* The AI promised a follow-up that hasn't happened. Let the flow auto-follow-up, then resolve.