Introduction

Generative AI, or Gen AI, has gained widespread popularity due to its ability to foster creativity, automate tasks, and drive innovation across various domains. Its versatility and practical applications in content creation and summarisation, virtual assistants, and personalised recommendations contribute to its widespread appeal.

Amazon is among the leading Gen AI providers for application developers offering a fully managed service called Amazon Bedrock, providing a variety of top-tier foundation models (FMs) developed by leading AI companies through a single API.

Implemented alongside Amazon Connect, Bedrock succinctly distils lengthy customer conversations into clear, concise, and contextually rich summaries. This helps supervisors to improve the customer experience by surfacing insights during contact reviews, simplifying quality and compliance evaluations, and promptly pinpointing areas for enhancing agent performance. Ultimately, it improves call centre efficiency and customer satisfaction.

‍

Amazon Connect Contact Lens Post-Contact Summaries

On March 25^th 2024, Amazon Connect Contact Lens post-contact summaries, powered by Amazon Bedrock, was made generally available in the US West (Oregon), and US East (Northern Virginia) regions. Once enabled conversation summaries are available in the Contact details section of Contact Lens and included in the raw Contact Lens output JSON file.

‍

Setup Post-Contact Summaries

• Add a “Set recording and analytics behaviour” block to a Contact Flow
• Set Analytics to On
• Select either Enable speech analytics, Enable chat analytics, or both
• Select Post-contact summary

‍

Security Profile Permissions

The following security profile permissions need to be applied on the Analytics and Optimisation section of the Edit security profile page to users who need access to this feature:

·      Contact Lens–post-contact summary, View

·      Recorded conversations (redacted), View orRecorded conversations (redacted), View

‍

Viewing Post-Contact Summaries

After successful setup and after a call has ended and closed from the CCP the post-contact summary is displayed in the Contact details page.

‍

‍

The raw Contact Lens output JSON file containing the post-contact summary is stored in an S3 bucket, in a similar location to this:  

/connect-instance- bucket/Analysis/Voice/2024/04/09/contact's_ID_analysis_2024-04-09T16:24:16Z.json  

‍

‍

Error Messages

If summaries aren’t being generated, an error message will appear on both the Contact details and Contact search pages. The ReasonCode for the error will be shown in the ContactSummary object within the Contact Lens output JSON file:

‍

‍

‍

AWS VPC and Services Setup for Power BI Refreshes

Creating Amazon Athena tables and views from the Contact Lens data stored in S3 allows for the execution of SQL queries against this data. An Amazon Glue ETL job can be implemented to use Athena as a source to populate additional data stores, such as Amazon RDS or Amazon Redshift.

To facilitate data refreshes in the service a Microsoft on-premises data gateway is installed on an Amazon EC2 instance within the same private subnet as the data source. Routing requests to the internet from this subnet is done via an Amazon VPC Network Address Translation (NAT) gateway situated in a public subnet. Utilising a NAT gateway facilitates connectivity for instances within a private subnet to access the internet or other AWS services whilst ensuring inbound connections from the internet are blocked.

‍

‍

Power BI Reporting Utilising Post-Contact Summaries

After creating a Power BI report that consumes the Amazon Bedrock Gen AI post-contact summary data and published to the service, scheduled refreshes can import the data from Athena using the Amazon Athena Power BI connector and the latest Amazon Athena ODBC driver installed on the EC2 instance.  Alternatively, from an RDS SQL Database using a SQL connection or a Redshift connection. All connections import data via the Microsoft on-premises data gateway installed on the EC2 instance.

‍

‍

Conclusion

This was an overview of some of the steps required to setup and configure AWS services for conversational summaries in Amazon Connect using Microsoft Power BI for visualisation.  

Combining the feature-rich functionality of Amazon Connect and the Gen AI capabilities of Amazon Bedrock makes voice and chat post-contact summaries a valuable addition to the Contact Centre.  

Paired with the power, performance and flexibility of Power BI, positioned as a leader in the 2023 Gartner Magic Quadrant for Analytics and Business Intelligence Platforms for sixteen consecutive years, completes the picture for a very capable all-round, modern Gen AI enabled Contact Centre.

Power BI offers extensive customisation options and supplements the default reporting features available in Connect, providing a holistic solution for advanced analytics and reporting needs.

Rush Nour

Data and Business Intelligence Architect at CloudInteract