Integrate your AI-powered IVR/IVA for seamless customer interactions with Amazon Connect

Introduction

If your business has a contact center, you may be looking to leverage the power of generative artificial intelligence (AI) to enhance user experience and agent productivity. Features such as agent assist or intelligent bots have been gaining traction as a result of this push towards contact center AI-assisted modernization.

Many of our customers are already using Interactive Voice Response (IVR) or Intelligent Virtual Assistants (IVAs) as a primary customer support channel to reduce time to resolution and optimize operational efficiency. And, a growing number of them are seeking to create a seamless integration between AI-driven customer interactions and human agent-driven interactions. This can provide a powerful customer care solutions, striking the right balance between the speed of automation and the personalized experience of a human agent. An increasingly common use case involves integrating Amazon Connect, a cloud contact center solution by Amazon Web Services (AWS), with an existing IVA or IVR system for AI-human handoff.

In this blog post, we’ll explore how businesses can further enhance their customers’ experience by seamlessly integrating their AI-powered IVR systems and IVAs with Amazon Connect. We’ll dive into the key benefits of such integration, as well as the architectural patterns that enable a seamless handoff between AI-powered assistants and human agents. Whether you’re a third-party provider looking to offer more integration options to your customers or you are seeking to modernize your existing customer service operations, this post will provide you with the insights and strategies to enhance the power of AI-human collaboration in your contact center.

Integrating AI-powered IVR/IVA with Amazon Connect

Below are two common patterns you can use while integrating intelligent assistants with Amazon Connect:

Enabling an AI-based assistant to seamlessly hand off voice calls to human agents when needed. This allows customers to interact with an AI virtual assistant for basic inquiries, while providing a smooth transition to a human agent, complete with full context and customer information gathered from the AI assistant, ensuring a smooth transition and avoids repetition, further expediting the resolution process.
Integrating third-party applications and tools into the Amazon Connect agent workspace. This is useful when you want to seamlessly integrate custom features and insights, whether from third-party sources or custom-built, into the agent workspace, providing additional functionality and information. It enables integration with various applications such as CRM systems, knowledge bases, and order management platforms into a unified interface, enabling agents to work more efficiently without toggling between multiple systems.

Architecture patterns

Let’s dive into some of the main architecture patterns involved in a seamless integration between your AI-powered IVR/IVA and Amazon Connect.

Pattern 1: Third-party assistant to Amazon Connect interaction transition

a. Key Capabilities

One of the primary integration patterns involves facilitating a smooth transition from an AI-powered self-service to a human agent across both voice and chat. When a caller requests to “speak with an agent” they expect a seamless and continuous interaction. It is important that the agent who takes over the conversation is provided timely and actionable information to best assist the customer. Here are some key capabilities required to ensure an effective hand-off:

Initiate an Amazon Connect contact from the system hosting the virtual assistant
The agent workspace should display customer information, including name and account data. The assistant should provide identification details during the transition. Preferably, the agent should have access to the same level of customer information as when the interaction originates from Amazon Connect channels. The workspace should also provide insight into pre-transition interaction. At the minimum, this should include a conversation transcript, metadata (date, time, duration), and any meaningful data extracted through pre-processing. This could encompass conversation tone, customer’s issue description, suggested solutions, enabling agents to quickly identify next steps and enhance the caller experience.

b. Architecture overview

Figure: Third-party IVA/IVR to Amazon Connect transition – Solution Architecture

Below is a description of the architecture and flow of information:

The customer interacts with the third-party IVA/IVR app. When the customer requests to speak with an agent, the request is sent to the Amazon API Gateway. The API Gateway routes requests to separate APIs for storing the conversation transcript (“/store”), processing the transcript to extract relevant information (“/process”), and initiating a new contact in Amazon Connect (“/start-contact”).
The “/store” API endpoint receives the conversation transcript and stores it in an Amazon S3 bucket using an AWS Lambda function.
The “/process” API endpoint triggers another Lambda function that processes the conversation transcript stored in Amazon S3. This AWS Lambda function may utilize AI services like Amazon Bedrock, Amazon Transcribe, and Amazon Comprehend to extract relevant information from the transcript.
The extracted information is stored in an Amazon DynamoDB. Depending on the use case other data store types may be used
Once the conversation insight data has been prepared, the third-party app invokes the “/start-contact” API endpoint which invokes the Amazon Connect APIs to initiate a live agent interaction with the customer though Amazon Connect instance. More details on this step in the following sections.
The Amazon Connect instance receives the request to start a new contact (text or voice).
As the agent acknowledges the support case and assigns it to themselves, they get access to all the information relevant to the customer’s request. Using the Amazon Connect agent workspace’s flexible integration capabilities, customers can display critical data such as chat or call summaries and derived insights.
The agent can review the conversation details and relevant information extracted from the transcript, then pick up the support case from the queue. Depending on the initial channel, the interaction could be a new incoming chat or a new voice call.

While the overall architectures for voice and chat are comparable, each channel has specific nuances and leverages distinct Amazon Connect features.

c. Voice channel

For voice-enabled assistants, a simple transition strategy is scheduling a callback. The third-party application can collect callback details from the caller, acknowledge the request, and initiate the interaction transition flow afterwards. A callback provides multiple benefits:

Minimize customer wait time during the transition from the third-party assistant to Amazon Connect, as generating and processing voice transcripts can be time-consuming, potentially exceeding the caller’s patience on the phone. Ensure that the agent assigned to the case has enough time to review the information available prior to interacting with the customer

To achieve this, Amazon Connect offers a number of Action APIs to build an automated callback flow such as CreateCallbackContact. Feel free to consult the Caller Schedule Callback blog for a sample callback solution.

d. Chat channel

For chat assistants, the strategy highly depends on the chatbot’s underlying solution.

Using a custom-built AI-powered chatbot

If your IVA is built on a custom chat platform, you can integrate it with Amazon Connect chat capabilities using the available APIs as described in the solution above. In this case, one of the critical components of the solution is the StartChatContact API that allows to initiate a flow to start a new chat for the customer. You can also pass custom attributes that can be accessed in flows. For instance, you could attach a unique hand-off identifier giving access to customer information and chat transcript data.

Feel free to consult the Amazon Connect Chat UI Examples repository for sample code and technical architectures.

Using Amazon Lex chat UI

Alternatively, if your IVA is built on Amazon Lex, you can leverage the native integration between Amazon Lex and Amazon Connect to create a unified chat experience. This approach allows you to take advantage of the conversational capabilities of Amazon Lex while seamlessly transitioning the chat to a human agent when necessary. The default intent capability of Amazon Lex also allows you to augment the chat experience with generative AI, adding another layer of automation before involving human agents.

The QnABot provides a good example of such a solution.

Pattern 2: Integrating third-party applications (3P apps)

You can also enrich the agent’s experience by integrating third-party applications (3P apps) as well as your own custom-built generative AI-powered solutions into the Amazon Connect agent workspace.

Integrating third-party applications (3P apps) into the agent workspace is a native capability of Amazon Connect and a powerful way to enhance agent productivity and customer experience. By consolidating critical business applications, data, and functionality within a single interface, agents can access all the information they need without switching between multiple systems. This streamlined access leads to faster issue resolution, improved first contact resolution, and better customer experiences.

There are several approaches to integrating 3P apps with Amazon Connect. AWS Marketplace offers a selection of pre-built 3P app integrations that can be easily deployed and configured. Alternatively, businesses that would rather build a custom integrations can leverage the platform’s robust APIs to programmatically integrate external applications and surface their functionality within the agent interface. For instance, they can use iframes to embed third-party web applications directly into the agent workspace, creating a seamless visual integration.

Common examples of 3P apps integrated with Amazon Connect include CRM systems, knowledge bases, order management platforms, and custom in-house applications. By consolidating these critical tools and data sources, businesses can create a streamlined agent workflow where all necessary information and actions are accessible in a single workspace. This can significantly impact key metrics like average handle time and first contact resolution.

Beyond pre-built integrations, businesses can also leverage the flexibility of Amazon Connect to build and integrate their own custom applications and services, including proprietary tools and AI-powered assistants. This allows for a truly tailored agent experience that aligns with unique business needs and workflows, unlocking new levels of productivity and customer service excellence.

Figure: Accessing a third-party application from the Amazon Connect agent workspaces

Conclusion and call to action:

The integration of AI-powered virtual assistants with Amazon Connect offers a compelling solution to elevate customer service operations. By seamlessly transitioning customers from an AI-driven interaction to a live agent while transferring full context, businesses can deliver exceptional experiences and boost agent efficiency. This approach enables higher resolution rates, improved satisfaction as agents receive relevant information upfront, and increased productivity by consolidating third-party applications and custom AI services within the agent workspace. For organizations optimizing contact center operations, this AI-human collaboration model presents a strategic opportunity to harness AI’s speed and scalability combined with live agent expertise.

To learn more and get started, please refer to the following resources:

Ready to transform your customer service experience with Amazon Connect? Contact us.

About the Authors

Aarushi Karandikar is a Solutions Architect at Amazon Web Services (AWS), responsible for providing Enterprise ISV customers with technical guidance on their cloud journey. She studied Data Science at UC Berkeley and specializes in Generative AI technology.

Guy Bachar is a Senior Solutions Architect at AWS based in New York, he specializes in assisting Capital Markets customers with their cloud transformation journeys. His expertise encompasses identity management, security, and unified communication.

Narcisse Zekpa is a Sr. Solutions Architect based in Boston. He helps customers in the Northeast U.S. accelerate their business transformation through innovative, and scalable solutions, on the AWS Cloud. He is passionate about enabling organizations to transform transform their business, using advanced analytics and AI. When Narcisse is not building, he enjoys spending time with his family, traveling, running, cooking and playing basketball.

Sarah Patrick is a Solutions Architect at Amazon Web Services (AWS), where she supports SMB Engaged customers in leveraging cloud computing services. Sarah studied Information Science and Business Analytics at the University of Maryland.She now guides customers through the early stages of implementing Amazon Connect for their contact center needs.

Agnel Joseph is a Professional Services Consultant at Amazon Web Services. He focuses on helping customers deploy scalable contact center solutions with Amazon Connect. He is a technologist and student who loves learning and creating new products.

AWS Contact Center