Dashboard
Operational intelligence across call quality
30-Day Quality Trend
Team Performance by department
Top Coaching Opportunities
Failure-Rate Heat Map
Recent Critical Flags
Sentiment
Category Breakdown
Resolution Rate
Reports
Search and filter call analyses
| Date | Agent | Customer | Category | Score | Sentiment | Resolution | Summary |
|---|
Agent Analytics
Performance across all agents
Create New Agent
Flagged Calls
Calls carrying system flags. Only rows marked Needs review need human action — Phase 2 also records compliance and coaching flags on calls it auto-finalized.
| Date | Agent | Score | Status | Flag Reasons | Category | Summary |
|---|
QA Review Queue
Human QA reviews for AI-human alignment measurement
| Date | Agent | AI Score | AI Recommendation | Review Status | Calibration |
|---|
AI-Human Alignment
POC success criteria tracking
What is this?
This dashboard measures whether the AI evaluation system is accurate enough to trust. Before going to full production, the AI must meet the POC success thresholds defined in the 360 Payments QA Evaluation Framework (Section 10).
- The AI (Haiku/Sonnet) evaluates a call across 17 metrics using observation labels
- A human QA reviewer independently evaluates the same call using the review form on the call detail page
- This dashboard compares both evaluations and calculates agreement rates
Recommended: Review 20-30 calls (mark as calibration batch) before scaling to full volume. If metrics are green, the AI is ready. If red, prompts and thresholds need adjustment.
Analyze Calls
Upload audio files or paste a transcript to run QA analysis
Transcript
Paste a call transcript or upload a text file — no audio needed.
Drag & drop a text file here, or
Supports .txt, .csv, .log, .srt, .vtt
Single Call
Drag & drop an audio file here, or
Supports MP3, WAV, M4A, FLAC
Batch Upload
Drag & drop multiple audio files here, or
Each file is analyzed as its own job — progress appears in the job cards below
Recent Jobs
| Submitted | Type | File | Rubric | Status | Result |
|---|
Rubric Profiles
Manage scoring rubrics used for call quality analysis
Create Rubric
Org Settings
Manage the call categories the AI classifies against, the department roster your reporting is bucketed by, and (admins) the org's Zoom connection
Call Categories
The AI picks one category per call from this list. Edit labels and descriptions to match your team's vocabulary — the descriptions are read by the AI, so the more specific the better.
| Label | Key | Description | Order | Active |
|---|
Departments
Who took the call. Every employee is assigned one on the Agents page; the dashboard's Team Performance and heat map are bucketed by it. Unlike categories, this list is never read by the AI. Deactivating a department keeps its history — assignments are preserved and it simply stops getting a column.
| Label | Key | Order | Active |
|---|
Zoom Connection
Server-to-Server OAuth credentials for this org's Zoom Contact Center account, from the app in the Zoom App Marketplace. The client secret is stored in AWS Secrets Manager, never in this app's database, and is never shown again after you save it. Test connection signs in to Zoom and reads one engagement to confirm the app has the access ingestion needs.
Ingestion
Each night the ingest job reads the previous UTC day's Contact Center engagements, applies the filters below, and picks up to the daily cap. Dry run records what it would ingest without downloading, transcribing or analysing anything.
Preview selection
Lists one UTC day's engagements from Zoom and applies the saved settings (save first to preview a change). Nothing is downloaded, analysed or recorded.
Activity
Zoom agents
The people Zoom names on the org's engagements. Map each one to their roster card so calls ingested from Zoom are attributed by Zoom identity, not by name. A suggested card is only preselected: nothing is mapped until you click Map. When ingestion is on, the nightly run adds the agents it sees; Refresh also looks up their email.
Admin Settings
Manage organizations and user access