Skip to main content
Test agent (also reachable as AI training from the Control center) opens the playground, a full simulation of a real conversation:
  • Pick the channel to simulate (a WhatsApp conversation formats differently than web chat), switch to a simulated time to test out-of-hours behavior, impersonate an existing contact to test with their real data, and choose which of your agents take part.
  • Chat as if you were the customer, or use the quick scenario chips (upset customer, pricing question, wants to book, hours question).
  • Every reply comes with a decision analysis: the intent it detected and the routing confidence, which agent stepped in and why, which knowledge bases and catalogs it consulted, and which transfer rules were evaluated, including why none fired.
  • Under each reply, Good response / Needs correction are the training loop. A correction lets you either improve the response in plain language or say it should transfer, and the agent learns it for every similar situation going forward. The same corrections flow powers the Evaluation center for replies already sent to real customers.
The playground: a simulated WhatsApp conversation with the decision analysis panel showing routing, sources consulted, and transfer rules evaluated

​
Reading the decision analysis

​
What a correction can and cannot teach

Needs correction opens two options, and they teach different things:
Corrections teach behavior. Facts go to the knowledge base, handoffs to transfer rules, tone to Basic configuration. A correction that tries to carry a price or a policy is caught and redirected. It never silently becomes agent behavior.
Every verdict also leaves a test behind: Good response protects that behavior, and a correction re-runs the flagged message to check that the fix holds. The Control center’s AI training card tracks how many replies you have rated.

​
One agent or every agent

When an agent wrote the reply, Improve response asks Apply to:
  • Only followed by the agent’s name (the default): only the agent that answered learns it. If you turn that agent off, the correction stops applying.
  • All agents: every agent learns it, including agents you add later.
The Add correction dialog with the Apply to choice: Only Agente calificador, or All agents Keep the default unless the correction is true for every conversation, like how to greet or how to close. A step that belongs to one product or one process (“on a renewal, ask for last month’s worker list”) saved for All agents is followed by every agent, including the ones that handle other products, and they all start answering the same way. If no agent answered (the first message, for example), the choice isn’t shown and the correction applies to all agents. Corrections made from the Evaluation center also apply to all agents.

​
Review and delete saved corrections

Saved corrections, in the playground header, lists everything your agents learned, grouped under All agents and then each agent with its own corrections. The playground header with the Saved corrections button between Simulate silence and Clear chat The same Coaching applied section appears in a real conversation’s View decision analysis in the Inbox, so you can trace a strange reply back to the correction behind it. You can also ask the assistant on your home page why the agent replied that way: it names the correction and offers to remove it. Use the trash icon to delete one that no longer applies, or one that was saved for the wrong scope. Agents stop following it from their next reply. Deleting can’t be undone, and you can’t edit a correction here: to change one, delete it and correct a reply again with the scope you want. A new correction never erases older ones, but there is room for a limited amount per scope (all agents, or one agent). When it is full, the correction is not saved and Ciarem says “Your saved corrections are full. Remove some in Saved corrections to add new ones.” Delete the ones that no longer apply and correct the reply again. The Saved corrections panel: two corrections for all agents, one only for Soporte Ciarem, and two only for Ventas Ciarem, each with a delete icon
Saved corrections are written in the language you corrected in. Your agents still reply in your business’s language.
If every agent suddenly answers the same way, or an agent you turned off still seems to be answering, open Saved corrections first. A product-specific correction under All agents is the usual cause.

​
Close the conversation

Close conversation ends the simulated chat the way a real close would. If satisfaction surveys are on for the channel, the survey is sent so you can answer it as a customer. After that the chat is locked until you restart it.

​
A routine that works

1

Write down your ten most common questions

In your customers’ words, misspellings included (“cuanto vale la limpieza?”, “do u do weekends”).
2

Run them and rate every reply

Do this before any change to agents, rules, or knowledge. Mark Good response on the ones that are right. Each becomes a test that protects that behavior.
3

Change one thing, run them again

Compare with the decision analysis: the same answers where you expected no change, better ones where you did.
4

Fix what failed in the right place

A wrong fact: the knowledge base. A wrong behavior: Needs correction, then Improve response. A moment that should have gone to a person: Should transfer.
Once you are live, the Evaluation center does the same for the replies your agent actually sent.