Version 5.28.0 Release Notes
Changes in System Behavior
None
Added/Updated Features
AI Agent in Voice and Chat Scenarios
The new AI Agent block allows you to automate complex customer interactions with the help of generative AI language models from OpenAI and Google. Just tell the block what it should try to achieve, and it will conduct natural dialogs (voice or chat) with customers. You can easily provide customized instructions to the AI for each interaction, allowing the AI to hold conversations that take each customer’s unique circumstance into account. This could be as simple as having the AI greet the customer by name, or including more complex background information such as order information or interaction history from your CRM.
Beyond simply having natural, personalized, and engaging conversations, the AI Agent block guides customers towards providing the information to reach any number of desired outcomes. Once customers have provided the necessary data for a requested outcome, the block cleans and structures the data for you to use elsewhere in the scenario.
AI Provider Integration Accounts
The AI Agent scenario block relies on generative AI models provided by third parties. The AI Provider integration makes it easy to connect your Google Vertex AI or OpenAI accounts to Bright Pattern. Setting up an integration account is straightforward, and only requires an API key or service account credentials from your preferred provider. The simplicity of account configuration allows you to focus on designing engaging and effective customer interactions.
Google Vertex AI
Google Vertex AI integration accounts allow you to use Gemini 1 Pro, Gemini 1.5 Flash, and Gemini 1.5 Pro language models to power the AI Agent block in your scenarios, as well as offering options for data-privacy.
OpenAI
OpenAI integration accounts give access to many of the latest models, including GPT-4 Turbo, GPT-4o, and GPT-4o mini. Other models may be available to you, depending on the tier of your OpenAI account.
Improved Management of Speech to Text Services
You now have more control over how Speech to Text (STT) is used in your contact center. The new STT Configuration section in the administrator application provides a granular overview of which transcription engines you are using for historical, realtime, and in-scenario speech to text services. For realtime and in-scenario transcription, you have the choice of using the default engines offered by your service provider, or using one of your own STT integration accounts. Engines for historical transcriptions are determined by your service provider.
If your service provider has configured default transcription engines and enabled STT Configuration for you, switching away from your old STT integrations is easy and offers a number of advantages:
- Automatically get the best transcription engine: By choosing to use the default engines from your service provider, you can be confident that the best engine will be selected automatically, regardless of the language being transcribed and whether you need historical transcriptions from call recordings, live transcription to assist agents, or in-scenario transcription for IVR.
- Access to the latest engines with fewer accounts to manage: You no longer need to manage configurations, usage, or billing for multiple third-party transcription providers, while still getting access to the latest STT options.
- Expanded data-protection options: Contact your service provider for transcription options that reduce the risk of data exposure to third-parties.
Services
You can choose whether historical transcriptions are generated by default for all your services, and override these global settings for individual services. If you have a large number of call recordings, you can set a specific date from which to start transcribing calls. Realtime transcriptions can be enabled per-service, with the option to give agents access to a live transcript of each call.
Keywords
Maybe some of your customers use jargon or pronunciations that tend to get missed by the transcriber? Beyond the choice of transcription engine, the new speech keywords option helps you improve transcription accuracy for both historical and realtime transcriptions by providing lists of words or phrases for the transcriber to give extra attention to. In-scenario transcription accuracy can also be improved by defining keywords for each AI Agent and Play-Listen scenario block that you use.
Whisper, Provided by Bright Pattern
This release introduces Bright Pattern's own transcription service, powered by the open-source Whisper model from OpenAI. Our Whisper service offers a robust multi-language solution for both realtime and historical transcription that keeps your data securely within the Bright Pattern infrastructure.
Bright Pattern Transcription Service is currently enabled on a per-request basis, please contact your service provider for more information.