AWS Contact Center

Real-time customer insights using machine learning with Contact Lens for Amazon Connect

Contact Lens for Amazon Connect provides a set of machine learning (ML) capabilities integrated into Amazon Connect. These capabilities enable businesses to analyze call recordings for customer sentiment, trends, and compliance of conversations. Contact center supervisors want real-time insights into customer experience issues while calls are in progress to provide proactive assistance to their agents. Historically, they have had to listen to a random subset of live calls in order to identify these issues. Contact Lens now supports real-time call analytics capabilities, enabling you to detect customer experience issues during live calls and resolve them faster. In this blog, we will discuss how Contact Lens for Amazon Connect works and how you can get started.

Setup Contact Lens

To enable Contact Lens in your Amazon Connect contact center, follow these steps:

1. Add a Set recording and analytics behavior block to an Amazon Connect contact flow. The following is a sample contact flow shown for reference.

2. Open the Set recording and analytics behavior block and choose Agent and Customer to enable recording.

3. Choose Enable Contact Lens for speech analytics.

4. Contact Lens has two modes of operation:

a. Post-call analytics: In this mode, Contact Lens for Amazon Connect will evaluate call recordings after a call has completed. You can use this mode if you don’t have any real-time use cases and prefer the highest transcription accuracy for post-call analytics. To enable this option, choose Post-call analytics.

b. Real-time and post-call analytics: In this mode, Contact Lens for Amazon Connect provides real-time insights during the call along with the post-call analytics after the call has ended. To enable this option, choose Real-time and post-call analytics.

5. Choose the language variant of the calls that you want to analyze from the Language drop-down list. Contact Lens currently supports 18 post-call and 4 real-time language variants.

6. Enable redaction by selecting the Redact sensitive data check box. Enabling redaction automatically detects and redacts sensitive information from call transcripts and audio recording after a call disconnects. This information includes names, addresses, payment card information, and social security numbers.

7. Select Generate both redacted and original transcripts and audio to generate both transcripts after calls complete. Or, select Generate redacted transcript only and both redacted and original audio to just generate the redacted transcript.

That completes the setup required to start using Contact Lens. Next, we will specify customer experience issues that will be monitored.

Track customer experience issues with categories

With the automated contact categorization capability, Contact Lens allows you to define customer experience issues for triggering real-time alerts to supervisors. Specify keywords or phrases using categories on Amazon Connect’s Rules page to look for scenarios like escalation requests, account cancellations, or competitor mentions during calls.

While creating categories, you choose between exact matching, pattern matching, and ML-based semantic matching, based on your preference. With pattern matching, Contact Lens can categorize calls based on patterns related to the phrases spoken by the customer and/or the agent. For example, you can define a pattern to categorize calls by stipulating that the word “card” should not occur within one word of “credit.” This allows you to organize separately calls that are related to “credit” and those that are related to “credit card.” With semantic matching, Contact Lens allows you to categorize calls that contain words/phrases with similar meaning even if they are not mentioned explicitly during the definition of a particular contact category. This is achieved through an ML-based intent matching model that is trained to understand the context and meaning of the keywords specified in the category definition. For example, you could enter “Thanks for your help” or “I am grateful for your help” in the definition for a category called “Politeness.” The new ML-based categorization can understand that a call containing “Thank you for your assistance” is similar in meaning by using semantic matching, and will categorize the call with the label “Politeness.”

The following screenshot shows an example of a pattern matching category.

Monitor supervisor alerts

Once you have created real-time categories, Contact Lens analyzes live calls to detect when the rule criteria for a customer experience issue is met. It will then alert supervisors in Amazon Connect’s real-time metrics dashboard to notify them of the issue. From there, supervisors can listen in on the live call and provide guidance to an agent over chat to resolve the issue faster. If necessary, supervisors can have the call transferred from the agent. The following screenshot shows real-time alerts for ‘Escalation’ and ‘Angry customer’ categories on the real-time metrics dashboard.

Select a dashboard alert to view the real-time transcript and customer sentiment trend to get up to speed on the customer issue and assess the appropriate action. The transcript decreases the need for customers to repeat themselves during call transfers, since the context of the customer’s issue along with the customer’s sentiment is captured in the real-time contact summary. An example of the transcript and the sentiment trend is shown in the following screenshot.

Build customized solutions with the real-time API

The real-time capabilities of Contact Lens are available not only through the Amazon Connect UI, but also through a synchronous API. With the real-time API, you can build customized solutions for use cases like agent transfers. These custom solutions can pass the real-time transcript with conversational details like sentiment to the next agent or supervisor so that the customer doesn’t need to repeat themselves. You can also develop call summarization solutions that create a summary of the customer experience issues at the end of the call. This helps reduce the after-call work needed from agents, and enables them to serve the next customer looking for assistance much faster.

Get started

With these new real-time analytics capabilities, Contact Lens for Amazon Connect enables you to build and retain customer trust by solving issues while they are occurring, rather than after the end of the call. Contact Lens offers a free tier at 90 minutes for audio calls per month, the same as Amazon Connect free tier usage. Get started with Contact Lens in your Amazon Connect contact center instance today or reach out to your AWS sales partner for more information.