Skip to content

Analytics & Dashboards

Monitor agent performance, conversation quality, and operational metrics.

Overview

The fifthelement.ai platform provides reporting and analytics capabilities through Agent Reports, Conversation Monitoring, and Hub-Integrated Reports. Track usage, quality, handoff performance, and knowledge repository insights.

Agent Reports

Access reports from the My Agents page by clicking on an agent's Reports option. Reports provide insights into:

  • Usage metrics — conversation counts, message volumes, active users
  • Performance metrics — response times, token usage, session costs
  • Conversation quality — ratings, feedback, flagged conversations
  • Human agent metrics — Unpicked Conversations, Picked Conversations, Requests for Human Agents, average rating

For details, see Agent Reports.

Conversation Monitoring

The Conversations page for each agent provides real-time monitoring with:

  • Conversation filters — Flagged, Flagged + Resolved, Unresolved, Like, Dislike, Dislike+Feedback, and custom post-processor filters
  • Human agent ratings — filter by specific agents or rating levels
  • Conversation details — view full transcripts, flags, feedback

For details, see Conversations.

Knowledge Base Analytics

Reports can be generated directly from a Knowledge Base data source (in addition to standard agent-based reports):

  • Dynamic metrics — available measures update based on the specific Knowledge Base and its extraction field configurations
  • Custom reporting — create reports based on derived values from Knowledge Base data

Hub-Integrated Reports

Reports can be configured for Hubs, making analytics visible directly to hub users:

  • Admins configure reports by visiting the Hubs page and clicking the Report icon next to a hub
  • Stakeholders can view reports without needing direct agent access

KPI Formatting

  • Readable duration formatting — time-based metrics automatically scale to human-readable format (e.g., 5h 20m, 54s, 16m)
  • Preserved precision — raw values available via tooltips for detailed analysis

Choose the Right Operational View

Question View to inspect What to check
Which conversations need review? Agent Conversations Filters, message details and analysis results.
Which tools or processors have failures? Agent Statistics Success/failure counters; inspect a specific case before diagnosing the cause.
What happened in a particular invocation? Agent debug logs Date, message/tool details and available error information.
Is asynchronous processing waiting or failing? Batch Monitoring from the account area Conversation, message and document processing records.
How does activity compare over time? Reports Measure, date range, filters and underlying retained data.

An existing failure counter may include intentional test cases. Correlate it with a reproducible invocation before treating it as a platform defect. Redact conversation text, credentials and identifying metadata from shared debug evidence.