Skip to main content

Ender Turing Updates — 2026-08-20 to 2026-09-02

Important changes

Automatic correction of names in call transcripts

Category: feature
Importance: high
What changed: Agent, company, and brand names that are misheard during speech recognition are now automatically corrected in call transcripts.
Customer benefit: This improves the accuracy of Auto QA checks for agent greetings and brand mentions, reducing false negatives.
Highlights:

  • Corrections are only applied when a misrecognized word sounds similar to a known name, and it never inserts words that were not said.
  • The original, uncorrected text is preserved, and all correction attempts are recorded in the conversation's activity log.
  • Auto QA scorecards for greeting compliance are now more reliable, especially for agents with unique or non-standard names.

Explicit control over topic accuracy evaluation

Category: improvement
Importance: medium
What changed: You can now manually trigger accuracy evaluation for topic review datasets by clicking a 'Finish review' button, instead of it running automatically.
Customer benefit: This gives you full control over when to measure topic accuracy, allowing you to review and revise labeled data without triggering unintended calculations.
Highlights:

  • Labeling samples now only saves your work, without automatically starting an evaluation.
  • The 'Finish review' action re-measures the topic's accuracy and suggests improvements if enough data is available.
  • If an evaluation cannot be started, you receive immediate feedback instead of a false confirmation.

Warnings for outdated topic accuracy scores

Category: improvement
Importance: medium
What changed: The platform now displays a warning on topic accuracy scores that were measured before the underlying review dataset was last changed.
Customer benefit: This helps you immediately identify which accuracy scores are outdated, so you can avoid making decisions based on stale data.
Highlights:

  • A warning icon and tooltip explain why the score is outdated and how to re-run the evaluation.
  • The warning helps ensure fair comparisons between different versions of a topic definition.
  • This check is applied retroactively to all existing topics and their accuracy histories.

Faster loading for the Agent Metrics dashboard

Category: fix
Importance: medium
What changed: The Agent Metrics dashboard now loads faster because it no longer calculates CSAT scores for columns that are hidden by default.
Customer benefit: This allows the main agent performance table to load immediately, avoiding long delays on accounts with large call volumes.
Highlights:

  • CSAT scores are now calculated on-demand when you show a CSAT column or sort by it.
  • The on-demand calculation runs independently and does not freeze the rest of the metrics table.
  • Data exports to CSV or XLSX still contain complete CSAT values as before.

View conversations behind category evaluation scores

Category: improvement
Importance: medium
What changed: You can now click on the result counts in a category's evaluation report to see the list of conversations for that metric.
Customer benefit: This allows you to quickly find specific examples of correct or incorrect classifications, helping you understand and improve the category's performance.
Highlights:

  • In the drill-down view, you can see the expected label, the AI's result, and its confidence score for each conversation.
  • Quickly identify misclassified conversations without having to manually search for them.
  • The list of conversations respects access permissions, so users only see conversations they are allowed to view.

Other changes

Improved reliability for Binotel and Aircall integrations

Category: fix
Importance: medium
What changed: Calls from Binotel and Aircall integrations now process correctly, even when they do not include extra metadata.
Customer benefit: This prevents calls from these sources from getting stuck during import, ensuring all recorded conversations are available for analysis and review.
Highlights:

  • Previously stuck calls from these integrations will be automatically recovered and processed.
  • This fix helps ensure your reports and dashboards reflect complete call volumes.

More reliable import of Fireflies meetings

Category: fix
Importance: medium
What changed: Meetings from the Fireflies integration now sync reliably, even when participant attendance data is incomplete.
Customer benefit: This ensures that all recorded meetings are imported into Ender Turing, preventing gaps in your conversation history.
Highlights:

  • This change prevents a single meeting with bad data from halting the entire sync process.
  • Meetings that failed to import previously due to this issue will be included in the next sync.

Consistent results for 'not a repeat call' filters

Category: fix
Importance: medium
What changed: Filters for 'not a repeat call' now return consistent conversation counts across the Conversations list, dashboards, and detailed reports.
Customer benefit: This ensures that reporting on first-contact resolution or non-repeat callers is reliable and consistent everywhere in the platform.
Highlights:

  • Conversations with no prior call history are now correctly included when filtering for non-repeat callers.
  • Automations based on repeat-call conditions now behave consistently with search and reporting logic.
Product Release