What's New — July 16, 2026
This release is about giving you precise control over your AI analytics. Instead of one-size-fits-all settings, you can now fine-tune the cost, speed, and quality of AI for each specific workflow, from conversation summaries to AutoQA. These new controls, combined with greater visibility and more reliable data pipelines, let you manage your contact center analytics with more confidence and precision than ever before.
Executive Summary
Per-Feature LLM Controls — Fine-tune the cost, performance, and quality of AI for each analytics feature, from conversation summaries to AutoQA, directly from the settings UI.
Highlights
Per-Feature LLM Controls — Tune AI Cost and Performance for Every Workflow
Previously, AI reasoning settings were a global, hidden control, forcing a difficult trade-off between cost and accuracy across all features and sometimes causing chat failures on specific cloud platforms.
▸ Optimize by workflow: Set a low AI effort for fast, cheap summaries while keeping AutoQA on high effort for maximum accuracy.
▸ Unlock full compatibility: EnderGPT Chat now works reliably on all Azure OpenAI deployments by letting you match the reasoning level your model requires.
▸ Manage directly in the UI: Adjust AI effort or "thinking budget" for each feature in your settings, with no need to edit configuration files or restart services.
- Before: A single, hidden setting controlled AI reasoning for all features, forcing a one-size-fits-all compromise and blocking EnderGPT Chat on some Azure deployments.
- After: You now have a visible control next to each feature's model in the UI, allowing you to set a specific reasoning level (e.g., low, medium, high) for each workflow independently.
- In practice: Alex, an admin, sets "Summary" to use 'low' effort for speed and cost savings, but keeps "AutoQA" on 'high' effort to ensure scoring accuracy remains top-notch.
Go to Analytics → LLM Settings. Next to each feature's model selection, a new dropdown lets you configure its reasoning effort.
Improvements
▸ Gain visibility into AI costs. If your organization has cost tracking enabled, you can now see the LLM costs associated with each EnderGPT chat and AutoQA prompt directly in the interface. This gives admins and QA managers with automation permissions the data they need to manage AI spend effectively.
▸ Fast-track urgent recordings for analysis. You can now mark critical recordings for high-priority processing by including a specific tag in the filename during upload. This allows escalations and time-sensitive reviews to jump to the front of the analysis queue, ensuring you get results in minutes, not hours.
▸ Filter automations by Campaign Name. Build more precise workflows by scoping automation triggers to specific campaigns. This allows you to create targeted rules for QA, alerts, or tagging that apply only to conversations from a new product launch or a seasonal marketing effort, for example.
▸ Get more reliable and measurable AI call categorization. Our call categorization engine is now more robust, reducing errors and ensuring that every category version gets an accuracy score, even brand-new ones. This provides a clearer, more trustworthy picture of category performance from day one.
▸ Instantly revert or benchmark topic definitions. When you edit an LLM topic, the system now automatically saves a "baseline" version based on the original description. This lets you instantly revert if a change doesn't perform well or directly compare a complex new definition against the simple original to measure its impact.
▸ Save costs and reduce noise by skipping AI analysis on empty calls. The platform now intelligently bypasses AI-powered actions like AutoQA, summarization, and categorization for short or silent calls with no transcript. This prevents wasted processing and keeps your dashboards clean, while still allowing metadata-based automations like tagging or notifications to run.
Fixes
▸ @-mentions in conversation comments are working again, allowing you to tag colleagues for collaboration.
▸ Clicking a data point in a weekly chart now correctly filters the conversation list to that specific week.
▸ Editing a scheduled Ender Chat's frequency, time, or timezone now correctly updates the "Next run" time.
▸ Media retention policies are now reliably enforced across all storage types, ensuring old recordings are properly deleted on schedule from Azure Blob and other backends.
▸ Failed scheduled chat runs now show clear, user-friendly error messages without exposing internal system details.
▸ The main dashboard no longer crashes when using the channel filter to view only Email sessions.
