Building Voice Agents with Amazon Bedrock: Enhance Customer Interaction with AI

Generative AI continues to revolutionize customer interactions, enabling natural and dynamic conversations across various channels. However, despite digital advancements, many customers still prefer phone interactions when seeking support or information. In this post, we’ll explore how businesses can effectively integrate voice interfaces with their existing Amazon Bedrock generative AI assistants, enabling seamless, voice-based customer engagements.

Voice AI Solution Overview

Using Workflow Studio for Amazon Web Services (AWS) Step Functions, we’ve designed a voice communication interface that integrates directly with the Amazon Nova Micro model on Amazon Bedrock. This setup leverages the model’s capability to handle open-ended customer queries. Depending on specific business needs, organizations can employ Amazon Bedrock Agents or Flows to customize interactions.

Figure 1: Step Functions workflow enabling voice interactions with generative AI.

How It Works:

1. Inbound call received
2. The system plays a welcome message
3. Prompts the caller to ask a question
4. Records voice input until silence detected
5. Parallel process:
   • Plays hold music
   • Transcribes the audio via Amazon Transcribe
   • Sends transcription to Amazon Nova Micro model
6. The AI-generated response is played back using text-to-speech
7. Loopback or terminate the call

Enhancing Your Voice Agent

Businesses can extend the capabilities of this AI voice agent in several innovative ways:

• Integrate other speech-to-text models, such as Whisper Large V3 Turbo via Amazon Bedrock Marketplace.
• Optimize audio recording parameters (e.g., silence duration, noise levels).
• Improve voice clarity with the PSTN VoiceFocus feature, enhancing customer call experiences.
• Utilize SIP trunking via Amazon Chime SDK Voice Connector for seamless integration with existing phone systems.
• Dynamically manage active calls using the UpdateSipMediaApplicationCall API.
• Parallel workflow states allow user-friendly handling of API service calls by playing music during processing
• PSTN audio service provides pay-per-minute rates with serverless, scalable telephony infrastructure

Deploying Your AI Voice Interface

Follow these simple deployment steps to launch your voice-enabled AI assistant using AWS services:

Prerequisites:

• AWS Management Console access
• Node.js and npm installed
• AWS Command Line Interface (AWS CLI) installed and configured
• Enable access to the Amazon Nova Micro model through the Amazon Bedrock console

Deployment Steps:

• Clone and navigate to the repository:

git clone https://github.com/aws-samples/sample-chime-sdk-bedrock-voice-interface
cd sample-chime-sdk-bedrock-voice-interface
npm install

• Bootstrap your AWS environment:

cdk bootstrap aws://<account-id>/<region>

• Deploy your stack:

cdk deploy --context phoneAreaCode=NPA

• Call your provisioned phone number to test your AI assistant.

Cleaning up resources:

cdk destroy

Origo’s Expertise: Enhancing and Customizing Your AI Solution

At Origo, we specialize in customizing and enhancing voice-enabled AI solutions to fit your unique business requirements. Our experienced team manages various AI model versions and architectures, removing the need for your business to develop deep technical expertise.

We focus on human-centric design principles, integrating a wide range of AI/ML models beyond Amazon Bedrock, tailored specifically to your industry and user needs. Our approach ensures that technology empowers people, promoting effective communication, efficiency, and user satisfaction.

Conclusion

Voice interfaces significantly extend the accessibility and engagement potential of generative AI assistants. Leveraging AWS Step Functions and Amazon Chime SDK PSTN audio services, this solution empowers businesses to deliver personalized, scalable, and cost-effective voice interactions through visual workflows.

From multi-language support to comprehensive knowledge bases and 24/7 customer service, voice agents built on Amazon Bedrock provide immense value by combining simplicity, efficiency, and advanced AI capabilities. With Origo, businesses can further enhance this solution to precisely match their operational needs, ensuring a seamless, user-focused experience.

Ready to explore how Origo can customize your AI voice solutions? Contact us today to start enhancing your customer interactions with innovative, human-centered AI.