What's New — August 16, 2026

What's New — August 16, 2026

This release expands your visibility into every corner of your contact center—from previously unanalyzed platforms to the true drivers of your AI spend. New granular controls give you the power to manage who sees what, ensuring the right data gets to the right people. These updates are paired with a powerful new way to measure first-contact resolution, giving you a more accurate understanding of your team's performance.

Executive Summary

Expanded Browser Recording — Capture and analyze conversations from 15 new contact center platforms, including Genesys Cloud, NICE CXone, Five9, and Talkdesk, directly from your browser.

True First-Contact Resolution Analytics — Isolate conversations that end the customer journey using a new filter that finds interactions with no follow-up call, giving you a precise way to measure FCR.

AI Token Usage Reporting — Go beyond dollar costs to see the exact token consumption of your AI workflows, making it easy to optimize prompts and validate provider invoices.

Granular Agent Permissions for Dashboards and AI Chat — Control which agents see team-level dashboards and have interactive AI chat access using new, dedicated roles for more precise data governance.

Highlights

Browser Recording for 15 More Platforms — Analyze Conversations Everywhere You Work

Your teams use a mix of leading contact center and meeting platforms, but you could only capture calls from a select few, leaving major gaps in your QA and analytics coverage.

▸ Expand your coverage instantly: Record and analyze calls from Genesys Cloud, NICE CXone, Five9, Talkdesk, Amazon Connect, and 10 other major platforms.

▸ Enable recording yourself: Activate new platforms in minutes from the WebRTC app settings, with no need to wait for a product update.

▸ Unify your analytics: Bring conversations from previously unsupported tools into the same scoring and analytics pipeline your other channels use.

• Before: Browser recording was limited to a few platforms, forcing you to use separate tools or accept blind spots for teams on systems like Genesys or Five9.
• After: You can enable recording for 15 additional CCaaS and meeting platforms with a simple settings change.
• In practice: An admin for a contact center on NICE CXone opens the WebRTC settings, enables the NICE CXone option, and agents can now have their calls recorded and analyzed in Ender Turing without changing their workflow.

To enable a new platform, open the WebRTC app settings and select the providers you want to record.

Next Conversation Interval Filter — Measure True First-Contact Resolution

Measuring First-Contact Resolution (FCR) was an approximation. You could see if a customer called before an interaction, but had no direct way to filter for conversations where the customer never called back.

▸ Identify resolved issues: Filter for conversations where the time to the next interaction is greater than a set period (e.g., 2 days), effectively isolating calls that truly solved the customer's problem.

▸ Investigate repeat contacts: Inversely, find conversations that triggered a quick callback to pinpoint training opportunities or process flaws.

▸ Simplify your workflow: A single, unified "Interval between conversations" filter now handles both previous and next intervals, with smart defaults to get you started faster.

• Before: To find "last-in-chain" calls, you had to manually trace customer journeys or export data to a spreadsheet.
• After: You can apply a single filter, like "Next interval is greater than 48 hours," to instantly create a cohort of resolved conversations.
• In practice: A QA Manager builds a scorecard for FCR quality. She uses the new filter to find all conversations with no follow-up call within 72 hours, creating a reliable sample for evaluation.

In the conversation list, open the filter panel and find the "Interval between conversations" filter. Choose "Next" and set your desired time range.

LLM Token Usage Reporting — Understand and Control Your AI Spend

While you could see the dollar cost of your AI usage, you couldn't see the underlying token consumption. This made it difficult to optimize prompts for efficiency or reconcile your Ender Turing usage with your LLM provider's bill.

▸ See your true consumption: View exact input, output, and total token counts for your AI workflows, broken down by prompt, model, and time period.

▸ Optimize for efficiency: Identify which prompts generate the most tokens and measure the impact of changes, helping you choose the most cost-effective configurations.

▸ Validate your invoices: Compare the token data in Ender Turing directly against your LLM provider's statements for easy reconciliation.

• Before: LLM usage was a black box reported only in USD, with no visibility into what drove the cost.
• After: The Gen AI Studio now shows detailed token metrics, splitting consumption into input, cached input, and output.
• In practice: A prompt engineer shortens a call summary prompt. She then checks the LLM Usage view and confirms a measurable drop in input tokens, validating that her change reduced AI costs.

Enable provider token reporting in Settings → Metrics and Variables → LLM configuration. Token data will then appear in the Gen AI Studio's LLM Usage view.

Granular Agent Permissions — Control Access to Dashboards and AI Chat

Giving agents access to team-level data was an all-or-nothing decision. A single setting controlled team dashboard access for every agent, and all agents had the ability to run their own AI chats, creating potential data governance and cost control challenges.

▸ Assign access with precision: Use the new "Agent with Team Dashboards" role to grant team-level analytics access only to specific agents, like seniors or mentors.

▸ Manage AI Chat usage: Agents now have read-only access to shared and scheduled AI chats by default. Interactive chat creation is a permission you can grant via roles.

▸ Simplify administration: All access is now managed through standard Roles and Permissions, removing confusing global settings and ensuring predictable behavior.

• Before: A single toggle controlled team dashboard visibility for all agents, and all agents could start their own AI conversations.
• After: Access is now role-based. A default agent sees only their own data and read-only chats, while a new dedicated role can grant expanded access.
• In practice: A team lead wants a senior agent to help mentor new hires. An admin assigns the "Agent with Team Dashboards" role to the senior agent, who can now see team performance funnels, while the new hires continue to see only their own metrics.

Navigate to Settings → Roles and Permissions to review the updated Agent role and assign the new "Agent with Team Dashboards" role to users.

Improvements

▸ Analyze internal agent-to-agent calls: You can now isolate, analyze, and report on internal calls as a distinct "Internal" direction. This allows you to exclude them from customer-facing analytics for cleaner metrics or review them separately to assess internal consultation quality.

▸ Improved AI reliability and diagnostics: AI-powered features are now more resilient. Failed analyses from invalid cached data will automatically retry with a fresh request, and any failures in EnderGPT Chat now display a clear reason, helping you understand if a retry will work.

▸ View full-time vs. part-time agent usage: The Monthly Usage report can now break down your active agents into "full-time" (11+ active days) and "part-time" (1-10 active days) groups. This gives admins a clearer picture of license consumption and helps justify capacity planning.

▸ Consistent language in AI-generated results: Auto QA evaluations and call summaries now reliably appear in the correct language. Projects using automatic language detection default to English, while Ukrainian bilingual projects correctly produce Ukrainian output.

Fixes

▸ Organization management pages are now more stable and no longer produce intermittent errors.

▸ The "Improve accuracy" feature for Topics is now clearly marked as Beta, with a cleaner UI for managing topic model versions.

▸ Resolved an issue where an entry in the corrections dictionary without alternative spellings could corrupt transcript text.