Reviewing AI conversations
On Configuration > AI > AI conversations you find every conversation the AI agent has had — through the chat widget, through WhatsApp and on the phone — with what each one used and cost. The tabs at the top filter on All, Phone, Chat or WhatsApp.
The columns
| Column | What it shows |
|---|---|
| Started | The moment the conversation began. |
| Channel | Phone, Chat, WhatsApp, or "Tools only" when tools were called but no conversation took place. |
| Caller | Who the conversation was with, as far as known. |
| Duration | How long the conversation lasted. |
| Messages | The number of messages exchanged. |
| Tool calls | How often the agent used a tool. If calls failed, that is marked. |
| Charged | The credits charged for this conversation. For a phone call this may temporarily read "not reconciled": the final cost only arrives after the call has been reconciled with the telephony platform. |
When credits are off
If credits are switched off for this licence, the screen says so at the top. No cost per conversation is recorded then — a column of zeros would suggest the conversations cost nothing, and that is different from not measuring.
Tool calls per conversation
Behind a conversation sits a Tool calls button. It opens a panel under the table showing which tools were used in that conversation: time, name, through which channel, whether it was a read or a write, how long it took and the result.
This detail view is reserved for Teqa support. Tool names reveal how an environment is wired technically, while the licensee's own question — did the agent do the right thing? — is already answered by the conversation itself.
What you use this for
- Checking what the agent said to a specific customer.
- Seeing whether tool calls fail, and whether a setting is the cause.
- Explaining where a month's usage came from: many short chats, or a few long phone calls.
