Response checks · Autessa Evaluations
Testing checks and reading results
The Test checks panel lets you try response checks on a sample reply before you publish, and every checked reply in history carries a Response checks card that shows what happened. You see each check result and reason, what matched, the decision and the exact text a customer would receive. Results are for your team only; they are never shown to the customer.
The problem
A guardrail you cannot see is a guardrail you cannot trust
Teams hesitate to turn on blocking because they do not know what it will catch. Autessa makes checks inspectable at both ends: test them on your own samples first, then read exactly what they did on live conversations, including the original draft whenever the customer received something different.
What you get
- Paste a sample reply and, optionally, sample sources: the tool results, knowledge or customer messages the reply should be based on.
- The panel shows each check result and reason, what it matched, the decision such as Redacted or Blocked, and the exact text the customer would receive.
- If the decision is a retry, the panel says the reply would be retried.
- AI review checks really run in the test panel.
- In the agent Test window, the final message shows a checks chip with the decision, and a retried draft never appears in the stream.
- In Agent History, every checked reply gets a Response checks card with one row per check and a badge: Passed, Flagged, Redacted, Blocked, Retried or Sent after retries.
- Check results are never shown to the customer and are not returned by the client conversation API.
How it works
How does testing checks and reading results work in Autessa Evaluations?
Open Test checks
In the Response checks card of the agent, open the Test checks panel.
Paste a sample
Add a sample reply and, if you like, the sources it should be based on.
Click Test
Read each result, the decision and the text the customer would receive.
Try it live
Use the agent Test window for a real conversation; the final message shows the checks chip.
Review in history
After publishing, open any conversation in Agent History and read its Response checks cards.
Every option
What you see
| Option | What it does |
|---|---|
| Sample reply | The text you want to check. |
| Sample sources | Optional tool results, knowledge or customer messages for the facts check to compare against. |
| Per-check result | Pass or fail, the reason and what matched. |
| Decision | For example Redacted or Blocked, or a note that the reply would be retried. |
| Customer text | The exact text the customer would receive. |
| History badge | Passed, Flagged, Redacted, Blocked, Retried or Sent after retries. |
| Original draft | A collapsible section shown when the customer received something different. |
Permissions
Who can do what?
| Who | Can |
|---|---|
| Agent editor (WRITE) | Use the Test checks panel on the draft. |
| Viewer (READ) | Test-run the agent in the Test window. |
| Agent Admins and company Admins | See conversations started through the API or an app, with their Response checks cards. |
Connected
How it connects to the rest of Evaluations
EVALUATIONS FEATURE
Conversation logs
Response checks cards sit alongside the transcript and every tool call in Agent History.
EVALUATIONS FEATURE
What your checks caught
A chart of replies the response checks caught each day, by what happened to the reply.
EVALUATIONS FEATURE
Start on Flag only
Run new checks on Flag only for a while, review what they catch, then switch to remove, block or retry.
FAQ
Questions about testing checks and reading results
Can customers see why a reply was changed?
No. Check results are for your team, never shown to the customer and not returned by the client conversation API.
Does the test panel run AI review for real?
Yes. AI review checks really run in the test panel, so the verdict you see is a real one.
How do I see what the agent originally wrote?
Open the conversation in Agent History. When the customer received something different, the Response checks card includes the original draft in a collapsible section.
100% customizable, code and all
Make Autessa Evaluations fully your own.
Use it out of the box, then change anything. Ask Autobot, Autessa’s AI builder, for a new field, screen, report, workflow or agent and it builds it into your copy of the app. Or open the code and edit it yourself. Nothing is locked: the app, its data and its code are yours.
Start free and make it yours