Conversation Logs

Review individual conversations and AI behavior in detail

The conversation logs give you a detailed view of every interaction Cevro has with players. Use them to review how the AI handled specific tickets, identify areas for improvement, and override quality scores when needed.

Accessing Conversation Logs

Navigate to the Conversations page from the main sidebar. You’ll see a list of all conversations, filterable by:

• Status — Open, resolved, escalated
• AIP — Which AI Procedure handled the ticket
• Brand — Filter by specific brand
• Date range — Focus on a specific time period
• Agent — Filter by assigned agent or team

Click any conversation to open the full transcript.

What You Can See

Each conversation log shows:

• Full transcript — Every message between the player and the AI agent, including tool calls and responses
• Classification — Which AIP was selected and why
• Actions taken — Tools called, data retrieved, decisions made
• Escalation details — If the ticket was handed off, the reason and context provided to the human agent
• Quality score — Per-conversation quality rating with rule-by-rule breakdowns (shown when the QA Scoring feature is enabled)

The conversation log is the most direct way to understand why the AI behaved a certain way. Before adjusting AIP instructions, always review a few example conversations first.

Reviewing Conversations

When reviewing a conversation, look for:

Signal	What It Means	Action
AI gave wrong information	Instructions may be incomplete or misleading	Update the AIP instructions
AI escalated unnecessarily	Knowledge gap or overly cautious instructions	Add the missing context or adjust escalation rules
AI didn’t use available tools	Tool may not be attached to the AIP	Check tool configuration
Player was frustrated despite correct handling	Process issue, not an AI issue	Review your SOPs, not the AIP

Overriding Quality Scores

If your workspace has the QA Scoring feature enabled, reviewers can override AI-assigned scores — useful when the scorer is more stringent than your own QA standard, or when contextual knowledge changes the verdict. See the QA Scoring Manual Review section for the full flow: override overall or per-section scores, add comments, and calibrate the scorer over time.

Workspaces without QA Scoring don’t see quality scores in the conversation log. To enable it, contact your Customer Success Manager.

Searching Conversations

The search bar at the top of the conversation list accepts several types of input. It tries the most specific match first and stops as soon as it finds results.

What You’re Looking For	What to Search	Example
A player by email	Their full email address	[e-mail]
A player by name	First or last name	Hugo
A player by player ID	The player’s back-office ID	PLR-12345
A Zendesk ticket	The Zendesk ticket number	48291
A tagged player	The exact tag name	VIP
A ticket by phone	Phone number (or prefix)	+1555

External ticket ID search (e.g. Zendesk ticket numbers) is currently supported for Zendesk integrations. If your help desk is Intercom, LiveChat, Respond.io, or Zoho, search by player email, name, or player ID instead.

Search Tips

• Be specific. Searching a full email is faster than a partial name because it matches exactly instead of scanning all contacts.
• Use the Zendesk ticket number to quickly jump from your help desk to the Cevro conversation for that ticket.
• Tags are exact match — searching VIP finds contacts tagged exactly “VIP”, not partial matches.

Tips for Effective Reviews

• Start with escalated tickets — These are the highest-signal conversations for finding improvement opportunities.
• Sample automated tickets too — Don’t only review failures. Spot-checking successful resolutions ensures quality.
• Look for patterns — If multiple conversations escalate for the same reason, that’s a systemic gap worth fixing.
• Use filters — Narrow down by AIP or brand to focus your review on specific areas.

Related

• Feedback & Ticket Review — Leave feedback on messages, score conversations, and drive agent improvement
• Understanding Metrics — Automation rate, escalation signals, and dashboard columns
• Topic Monitoring — Track unresolved topics across conversations
• QA Scoring — Configurable quality evaluation (available on request)
• Delegation vs Escalation — Understand why tickets are handed off