Intelligence Engine Version 1.5 Release Notes
Enhancements
Unanswered Email and Service Transfer Processing Improvements
The Intelligence Engine features improved processing for unanswered emails (such as spam or no-action items) and emails transferred between services.
Improved AI Summary Setup
Historical AI Summaries and Real-Time AI Summaries are now managed on separate screens within the Intelligence Engine. This change provides separate workspaces for configuring real-time and historical summarization, making it clear which settings apply to ongoing interactions versus past data. This separation allows you to apply distinct strategies to each.
From the new Historical AI Summary screen, you can immediately identify the active status of your services. The dashboard highlights services that are running correctly, stopped, or misconfigured. For any misconfigured services, a tooltip provides the specific reason, such as missing transcriptions or an incomplete AI provider setup. The dashboard also displays interaction metrics over the last 7 days, including successfully summarized interactions, items delayed by AI processing limits, and failures. You can download a CSV file of any configuration errors to assist with troubleshooting.
You can manage summary templates across your services directly from this view. The interface supports bulk operations, making it possible to assign, change, or reset templates for multiple services at once. If you need to halt historical summarization for operational or technical reasons, you can pause and resume the process for selected services without losing your configuration.
Use the new Selective Past Processing wizard to summarize interactions created before ongoing summarization was configured. Past processing runs in the background, allowing you to continue working while monitoring success and error rates.
Control Suggestion Frequency
The new Suggestions Frequency setting controls how often a suggestion is presented to an agent during an interaction. Use this to prevent repetitive notifications when a customer repeats phrases or intents that have already triggered guidance.
You can choose from three frequency options:
- Once per interaction: The suggestion appears only the first time the intent is detected.
- Every time intent is extracted: The suggestion appears every time the system detects the intent.
- Not more often than N minutes: The suggestion reappears only after a specified interval (between 1 and 60 minutes) has passed.
While these settings control what the agent sees in real-time, the interaction review continues to log every instance an intent is extracted. This provides a full record of the conversation for reporting and quality management without cluttering the agent's view.
Existing configuration sets are set to Every time intent is extracted during migration to maintain current behavior. New sets default to Once per interaction.
New Grammar Intent Variables to Detect Holds, Mutes, and Silences
The Intelligence Engine now supports Grammar Intent variables that indicate hold, mute, and silence states without requiring AI-based processing. The following variables are available to trigger intents:
AGENT: HOLD: Identify the frequency or duration of hold segments.
AGENT: MUTE: Identify the frequency or duration of the agent muting their microphone.
SILENCE > N: Detect call-wide "dead air" where no party is speaking for more than N seconds.
CUSTOMER: NO_SPEECH > N: Detect when the customer is silent for more than N seconds.
|
These new variables require Bright Pattern Contact Center 5.36 and Historical Transcriber 1.2. | |
Assign Different Autoscoring Forms for Inbound and Outbound Campaigns
You can now assign autoscoring forms based on whether an interaction is Inbound or Outbound. This new "Direction" filter joins the existing Talk Time and AM filters within the Intelligence Engine, and it functions across voice, chat, and email channels. The Direction filter makes it possible to assign multiple forms to a single service based on interaction direction.
All forms assigned to a service are now visible on the relevant product page. At runtime, the system reviews the assignment filters against the interaction. If one assignment matches, that form is applied. If multiple assignments overlap and match the interaction, the system applies the first form in the defined order.
Analyze and Score Self-Service Interactions in the Intelligence Engine
Scenarios that include the Self-Service Provided block are now treated as a distinct type of agent within the Intelligence Engine. This change allows you to search, analyze, and score your automated interactions using the same tools you use to evaluate human agents, giving you direct visibility into the performance of your automated flows.
When using Conversation Search or Intent Testing, the Agents/Teams filter now includes a "Self Services" group. You can filter by specific scenario names or select "Any" to view all self-service interactions. Autoscoring dashboards now feature a "Self Services" pseudo-team, so you can compare the quality of your automated scenarios to your human teams' performance.