Overview - Studio

PolyAI releases monthly Studio updates. Explore the update cards for release summaries:

July 2025

Click the bolded titles for a breakdown of what’s new, or go to the notes tab for full details.

Smart Analyst enhancements: Keyword filtering, SQL-style aggregation, and integrated charts now make LLM-powered QA review faster and more insightful.
New homepage: A redesigned dashboard shows key agent metrics, best/worst recent calls, and quick links to support tools.
Rich text refactor: Input fields now support slash-command rich text, allowing inline references to flows, functions, and more.
Add Amazon Polly voices: Add new voices with language/accent tags via the improved Amazon Polly integration.
Adjust voice speed: Control TTS speed between 0.5× and 1.5× with a precision slider for pacing and accessibility.
Claude integration for Agent Studio functions: Use Claude to generate summaries, text, or structured outputs inside functions.
Channel filter and call metadata improvements: Conversation Review now shows clearer call direction tags (Inbound, Outbound, Agent chat) and improved filtering.
Outbound call delivery tracking: Delivery outcomes like Busy, Invalid Number, and Declined are now visible in Conversation Review filters and columns for outbound calls.

June 2025

Click the bolded titles for a breakdown of what’s new, or go to the notes tab for full details.

Agent configuration: You can now define settings like opening hours or routing rules in a JSON config form, without editing flows.
Version comparison enhancements: The diff version structure introduced in the 25.02 release is now supported across functions and flows.
Action name search and autofill: Type /name in any rich text field to trigger action insertion.
Cached disclaimer messages: Disclaimer messages are now stored in the Cache under audio management, reducing latency and improving reliability across multiple flows.
Smart analyst (Beta): This LLM-powered QA tool can rate calls, summarize agent behavior, and flag weak containment. Contact your PolyAI rep to participate in the Smart Analyst beta.
Agent memory: Retrieve persistent values from previous conversations with the conv.memory object.
Test suite: Save real conversations as test cases and re-run them against draft or sandbox versions to check for changes in behavior.
PolyScore: New automatic scoring for conversation quality using defined behavioral metrics.

May 2025

Click the bolded titles for a breakdown of what’s new, or go to the notes tab for full details.

DTMF support: Configure tone-based input (DTMF) during specific steps to collect keypad responses from callers with support for timeouts, digit length, and early listening.
Chat with variants: You can now test how different variants respond in the chat panel—choose any draft, sandbox, or variant on demand.
Duplicate and set default variants: From the ellipsis menu on the variant page, duplicate an existing variant or assign a new default with a simple toggle.
Build KBs with URL and PDF upload: Create knowledge base topics directly from a website or uploaded PDF. Real-time crawl status and error reporting help troubleshoot issues when importing.
Chat-to-review shortcut: Quickly jump from a chat session to the conversation review page with one click, streamlining your debugging process.
Handoff reason and utterance: You can now include reason and utterance fields when writing a handoff implementation in a KB topic, flow, or function call.
Entities and Intents in Conversation Review: View detected intents and extracted entities directly inside the Conversation Review page.
Improved multi-language ASR accuracy: Speech recognition updates improve accuracy for numbers, dates, and accented speakers across several languages.
Conversation review: session duration filter: Filter conversations by session length to focus QA on very short or long calls.

April 2025

Click the bolded titles for a breakdown of what’s new, or go to the notes tab for full details.

Information architecture updates: Tools have been regrouped by function. “Analyze” is now “Manage”, and sections like Voice and Environments & Versions have moved or been renamed.
Metrics: Use the new Assistant Analysis tab to run custom metrics across batches of calls and generate insight-rich evaluations.
Live collaboration support: Agent Studio now shows who else is editing a draft, alerts you when updates are saved, and lets you review edit history to avoid accidental overwrites.
Call categorization: Use an LLM to rate calls for tone, intent, or compliance. Categories are defined per project. Contact your PolyAI rep to configure.
Workspace filter in Conversations: Filter conversation review by workspace to focus on your team’s calls.
Utility function – extract_address: Extracts a structured address from a user message. Returns a typed object, or a clear error. Opt-in only.
Conversation history access – conv.history: Use conv.history to access earlier turns from inside a utility function. Supports custom logic and debugging.
Import and export variants: Export all variant data to CSV, edit it, then re-import across agents. Includes overwrite protection.
Variant ID added to conversations: Each call now shows the variant ID used—helpful for QA and tracking.
Multi-voice agents: Assign multiple TTS voices to an agent and control their distribution. Great for A/B testing or simulating real-world voice teams.

March 2025

Click the bolded titles for a detailed breakdown of the released feature, or open the notes tab on the sidebar for the full release notes.

Overview dashboard enhancements: Additions to the dashboard include call volumes, durations, containment rate, SMS and function stats—refreshed hourly.
Safety dashboard enhancements: A refreshed version of the safety dashboard to match the new overview, with faster load times and a clearer layout.
Open matched topics in a new tab: In conversation review, clicking a matched topic opens it in a new browser tab—letting you keep your place while checking the KB.
Version tracking in conversation review: Each conversation now shows which version of the agent it ran on.
Add voices directly in Agent Studio: Voices can now be added or managed without leaving the platform.
Filter voices by language, accent, or gender: The voices page now supports helpful filters to speed up selection.
Send SMS immediately during function calls: SMS messages now send in real time as part of the function, so agents won’t falsely confirm delivery if the message fails.
Improved KB search for actions and tags: You can now search for SMS, handoff, and function names in the knowledge base using the main search bar.

February 2025

Click the bolded titles for a detailed breakdown of the released feature, or open the notes tab on the sidebar for the full release notes.

View knowledge base differences between versions: Easily compare changes across different versions of the knowledge base for better content tracking and updates.
Utterances for high latency functions: Manage function latency effectively by configuring filler utterances and dynamic delay timing.
Upload custom audio to replace cached items: Upload pre-recorded audio directly from the Agent Studio audio management resource.
Add variants to knowledge base items: Add variant attributes to knowledge base content dynamically.
Manage transition functions in a flow: Improve transition function management within flows, preventing unintended deletions.
New conversation review annotations: Add annotations and comments to conversations for better review and collaboration.
A new version of the overview dashboard: A new standardized dashboard displaying key operational data like total calls, duration, and handover rates.

January 2025

Click the bolded titles for a detailed breakdown of the released feature, or open the notes tab on the sidebar for the full release notes.

Enterprise safety dashboard: A new dashboard focused on flagged calls and safety filter trigger analysis.
Custom dashboards: Track project-specific success metrics with tailored dashboards.
In-app calling: Test voice interactions directly in-studio without telephony setup, enabling faster and simplified workflows.
Audio management: Delete or regenerate cached TTS audio for improved flexibility.
A new knowledge base structure: Separate questions and content for better organization.
End-of-conversation functions calls: Trigger automatic function calls at the conclusion of interactions for seamless integration with external systems.
API enhancements: Variant information now included in the Conversations API.

December 2024

Click the bolded titles for detailed for a detailed breakdown of the released feature, or open the notes tab on the sidebar for the full release notes.

Multi-site config: Manage multiple site-specific Knowledge Bases (KB) in a single agent, with location-specific variants for tailored responses.
Global ASR corrections, biasing, and keyword boosting: Customize Automatic Speech Recognition (ASR) corrections to refine transcribed text, improving LLM input accuracy.
Audio management: Browse, delete, regenerate, or upload cached TTS audio for better quality control.
Customize latency response delay: Set dynamic response delays to fit conversational tone and audience preferences.
Copy and paste flows: Transfer flows or nodes between projects to streamline builds and enable consistent configurations.
Call recordings download: Download call recordings directly from the conversation detail page, with required permissions.

November 2024

Click the bolded titles for detailed for a detailed breakdown of the released feature, or open the notes tab on the sidebar for the full release notes.

In-app function debugging: View and copy input, output, and error messages for function calls directly in the Conversation Review and agent Chat Panel.
Enhanced SMS: agents can send SMS messages to alternative numbers specified by callers and upgrade SMS templates to advanced functions with added logic and configurations.
Knowledge base export: Export knowledge base topics to a CSV file for version control and reuse.
Barge-in feature: Enable agents to be interrupted by callers, improving natural conversation flow and reducing latency.
Stop keywords: Configure specific words or phrases to automatically halt the agent’s response and trigger a function.

October 2024

Click the bolded titles for a detailed breakdown of the released feature, or open the notes tab on the sidebar for the full release notes.

Separate voice disclaimer with ringing tone: Customize greetings and dial tones to enhance user interaction.
ASR biasing for flows: Configure speech recognition biasing for specific input types, like alphanumerics, times, or precise dates, and improve ASR performance and accuracy.
UK and EU regions: Regional expansion for improved performance and compliance.
View versions: Access and manage historical and live versions of agents for better development and maintenance.
Content safety filters: Full control over content filter severity and configurations for project-specific needs.

Updates will be added to this page sequential with product releases.

Releases