|
Release Name |
CX5.14.0 |
|---|---|
|
Release Date |
|
|
Release Summary |
Improves the reliability and accuracy of AI-powered quality evaluation, working consistently across any supported AI model and across SMS, voice, and WhatsApp conversations. Adds live call transcription to Agent Desk, showing text on screen while a voice call is still in progress. Strengthens outbound campaign dialing with new agent-capacity controls, and compliance checks on voice campaign dials. Migrates tenant configuration into isolated per-tenant databases and enforces authentication on the CX Tenant API, improving tenant data isolation and security. Completes a scheduled security hardening review across CX platform components, including Quality Management and the PII-Masker service. This release also includes all fixes shipped in the CX5.13.1 patch release. Rounding out the release, a range of reporting, communication, and agent-experience fixes resolve issues reported from production environments. |
🚀 Deployment & Upgrade
|
Scenario |
Guide |
|---|---|
|
Deployment Guide |
|
|
Upgrade from CX5.13.0 |
IMPORTANT — Read Before Upgrading to CX5.14.0
Authentication is now enforced on the CX Tenant API. Starting with this release, every request to the Tenant API requires a valid Authorization token. Any integration or caller that does not send one will begin failing with 401 Unauthorized after the upgrade. Review your integrations against the API Authentication – Integration Guide beforehand.
New in CX5.14.0
This is a minor release on top of CX5.13.0. It focuses on improving AI-powered quality-evaluation accuracy, strengthening outbound campaign dialing controls and compliance, and isolating tenant configuration into per-tenant databases, alongside a range of reporting, communication, and agent-experience fixes. This release also includes all fixes shipped in the CX5.13.1 patch release.
Quality Management Enhancements
|
Feature |
Description |
|---|---|
|
AI-powered quality evaluation now works reliably with any supported AI model, produces more accurate results across SMS, voice, and WhatsApp conversations, and no longer limits conversation length before evaluation. Every review now completes, even when the AI service is briefly unavailable, and scoring excludes questions that do not apply to a conversation, so quality teams see fewer incorrect or missing results across real conversation types. |
Outbound Campaigns
|
Feature |
Description |
|---|---|
|
Allow inbound tasks to preempt agents reserved for outbound campaigns |
CX can now preempt an agent reserved for outbound campaign dialing to handle an incoming inbound task. This keeps inbound customers from waiting when all available agents are tied up with outbound reservations, improving overall queue responsiveness. |
|
Reserve agent capacity for inbound voice calls during outbound dialing |
When the outbound dialer reserves an agent for a predictive campaign call, CX now applies a configurable limit on how many times it can do so before stopping. This lets contact centers control how many agents outbound campaigns can reserve, so inbound voice queues always keep agents available. |
|
Check Do Not Contact and opt-out status before CX-Voice campaign dials |
CX-Voice outbound campaigns now check each contact against Do Not Contact and opt-out status before dialling. This prevents outbound campaigns from connecting agents to numbers that should not be contacted, supporting compliance with customer-consent requirements. |
Agent Experience Improvements
|
Feature |
Description |
|---|---|
|
CX now transcribes voice calls live and shows the text on Agent Desk while the call is still in progress, displaying caller and agent speech separately. This gives agents a readable, real-time record of the conversation without waiting for the call to end. |
Platform & Security Improvements
|
Feature(s) |
Description |
|---|---|
|
Global configuration for channels, queues, forms, and core services now migrates into segregated, per-tenant databases instead of shared global storage. This improves tenant data isolation, reduces the risk of cross-tenant data leakage, and supports scaling CX to more tenants on a single deployment. |
|
|
The CX Tenant API now enforces authentication on all requests. Callers must include a valid access token; requests without one are rejected. |
|
|
Complete a scheduled security hardening review across CX platform components |
Completes a comprehensive security review across CX platform components, including Quality Management and the PII-Masker service, resolving every item identified during the assessment. This also ensures CX consistently masks sensitive data, such as passwords, in system logs. See the CX Security Assessment Report for details. |
Bug Fixes
|
Feature |
Description |
|---|---|
|
Fix incorrect "Initiated Since" duration in Ongoing Queue |
Issue: When a supervisor joined or left a live chat monitoring session, the "Initiated Since" duration shown in the Ongoing Queue reset instead of reflecting the chat's true elapsed time. Resolution: CX now preserves the original chat start time when a supervisor joins or leaves monitoring, so the Ongoing Queue always shows the correct cumulative duration. |
|
Issue: The system sometimes failed to release an agent back into the pool after a predictive campaign call, so it couldn't be reused; an inbound call to an unavailable agent also had no fallback. Resolution: CX now correctly releases and reuses agents after a campaign call, and administrators can route an inbound call to voicemail when the target agent is unavailable. |
|
|
Fix error returned by realtime reports when querying wait-time statistics |
Issue: Calling the Estimated Wait Time (EWT) data alongside certain realtime reports queue statistics returned an error instead of the expected data. Resolution: Realtime reports now return results reliably when combined with wait-time data. |
|
Fix embedded Agent Desk conversation data not reaching CRM parent window |
Issue: When Unified Agent was embedded inside a CRM such as Salesforce on a different domain, the browser's cross-origin security checks could block conversation data events. Resolution: Unified Agent now sends these events in a way that works safely across domains, so embedded CRM integrations receive conversation data as expected. |
|
Issue: PII masking could consume significantly more memory than expected during normal operation, risking resource exhaustion. Resolution: CX now uses memory more efficiently for PII masking; updated sizing guidance is available for administrators. |
|
|
Improve PII masking behaviour under heavy load |
Issue: When PII masking was at full capacity, callers could wait until their request timed out instead of receiving a clear response. Resolution: PII masking now replies immediately with a busy response when at capacity, and allows more time for large requests to complete. |
|
Fix reports showing no data or the wrong queue name |
Issue: Several reports could show no data on first load due to paused data pipelines or a queue-name filter that didn't handle missing values correctly, and the Agent Productivity By Queue report's queue filter showed internal channel names instead of the actual queue name. Resolution: CX now loads these reports correctly without needing a page reload, and the Agent Productivity By Queue report shows the correct queue name. |
|
Fix missing email response details in integrated tool and transcripts |
Issue: Email responses with larger content could show missing details in the integrated tool and in transcripts. Resolution: CX now supports larger email response bodies across Unified Agent and the email connectors, so full content comes through correctly. |
|
Fix campaign channel reverting to the first channel after a Redis flush |
Issue: After a Redis flush and a campaign service restart, a campaign could revert to using the first available channel instead of its own configured channel. Resolution: CX now retains each campaign flow's configured channel after a Redis flush and service restart. |
|
Reduce Redis write load from Conversation Studio |
Issue: Conversation Studio periodically wrote the full in-flight conversation context to Redis, adding unnecessary write load during normal operation. Resolution: Conversation Studio now persists only steps that are still running after a short delay, reducing Redis write load without changing conversation behavior. |
✅ Compatible with
This release has been thoroughly tested with various environments and tools.