Review chat escalations after the fact
Updated September 28, 2026 · 3 min read
When the Customer Support Chat worker hands a conversation to your team, you can review the escalation later on its Results. See the reason it was escalated, then open the Agent trace to follow the worker's reasoning step by step.
Find escalated chats
- Go to Workers → Chat support and open the Results tab.
- Set the date range to the period you want to review.
- To see only handoffs, click Filters, set Escalated to Yes, and click Apply. You can also filter by Escalation type.
- Read the escalation columns on each row, described below.
Column | What it tells you |
|---|---|
Escalated | Yes when the worker handed the chat to a human instead of closing it. |
Escalation type | Why it escalated, for example Policy, Missing knowledge, Custom plan, or Unsupported attachment. |
Escalation reflection | A one-line AI review of what the worker was missing, written shortly after the handoff. It shows Pending until it's ready. |
Trace | Opens the Agent trace for that chat. |
Crisp | Opens the conversation in Crisp. |
Open the Agent trace
- On an escalated row, click Agent trace in the Trace column. The Agent trace panel opens on the right.
- Read the Triggering input card at the top. It shows the customer message the worker was answering.
- Check the summary under the interaction: how many Steps the worker took, the Outcome, and the Agent Skills it loaded.
- Scroll through Preparation work (how the worker classified the message and picked skills) and the numbered steps that follow.
Read the worker's reasoning
Each step shows the worker's Reasoning, when it wrote any, and the Actions it took, such as searching the help center or escalating to a human. Click View details on an action to see its Request and Response. On an escalation, the Request shows the reason and category the worker gave for the handoff.
Step back through earlier turns
A chat can have many turns. The Interaction card (for example, Interaction 8 of 8) has previous and next arrows so you can check how the worker handled each earlier message: what it planned, which help center articles it found, and whether a Guard stopped a draft. This often shows why the conversation ended in a handoff.
Fix the cause
Once you know why the chat escalated, fix it where it started. If the answer was missing, add it to your knowledge base. If the playbook steered the worker wrong, edit the Agent Skills. If the handoff was right, for example because the customer asked for a person, no change is needed.
FAQ
Where do I see why a chat was escalated?
On the Customer Support Chat Results tab. Escalation type gives the reason category, Escalation reflection gives a one-line review, and Agent trace shows each step the worker took.
Why does Escalation reflection say Pending?
The reflection is written in the background shortly after the handoff. Check back in a few minutes.
Why is there no Agent trace link on a row?
A dash means no trace was recorded for that result, which can happen on older chats.
Does the escalation review send anything to the customer or to Crisp?
No. The reflection and the trace are only visible to your team in Zentesimal.
Related articles
- Read and filter Results
Open the Results tab, filter by date or column, then read each row for the full outcome.
- Add a note on a Result
Leave an append-only note on any Worker Result so the next teammate who triages it has context.
- Edit Agent Skills for a worker
Open Knowledge → Agent skills to edit the playbook a worker reads at runtime.
- Add a prompt column on Results
Add a prompt column when you need a one-off label or summary on existing Results.