Intact Financial Corporation is Canada’s leading provider of property and casualty insurance, a leading global provider of specialty insurance, and a commercial leader in the United Kingdom and Ireland. Intact faced the challenge of managing a vast network of customer support call centers and needed a viable solution within six months and a long-term solution within one year. With up to 20,000 calls per day, manual audit processes were inefficient and difficult to keep up with increased call traffic and customer service expectations. Quality control personnel had to manually select calls to audit, which was not a scalable solution. To address this, Intact turned to AI and speech-to-text technology to extract insights from calls and improve customer service. The company developed an automation solution called Call Quality (CQ) using Amazon Web Services (AWS) AI services. With CQ, Intact handles 1,500% more calls (15x more calls per auditor), reduces agent handle time by 10%, and generates valuable insights into agent behavior, helping customers This led to improved service.
Amazon Transcribe is a fully managed automatic speech recognition (ASR) service that helps developers add speech-to-text functionality to their applications. Convert speech to text quickly and accurately using deep learning. This post shows how CQ Solutions uses Amazon Transcribe and other AWS services to improve important KPIs with AI-powered contact center call auditing and analysis.
This allows Intact to accurately transcribe customer calls, train custom language models, simplify call auditing processes, and more efficiently extract valuable customer insights.
Solution overview
Intact aimed to develop a cost-effective and efficient call analytics platform for contact centers using speech-to-text and machine learning technology. The goal was to improve customer service scripts, provide coaching opportunities for agents, and improve the call handling process. In doing so, Intact wanted to improve agent efficiency, identify business opportunities, and analyze customer satisfaction, potential product issues, and training gaps. The following diagram shows the solution architecture. This will be explained in the following sections.
Intact chose Amazon Transcribe as its speech-to-text AI solution because it can accurately handle both English and Canadian French. This was a key factor in Intact’s decision, as they wanted a versatile platform that could adapt to a variety of business needs. Amazon Transcribe’s scalability to handle hundreds to tens of thousands of calls per day, as well as deep learning capabilities that can handle a wide range of voice and acoustic characteristics, also played a key role. . Additionally, Intact was impressed with Amazon Transcribe’s ability to adapt to a variety of post-call analytics use cases across the organization.
Call handling and model provision
Because Intact has an on-premises contact center and a cloud contact center, we built a call retrieval process to capture calls from both sources. This architecture includes a fully automated workflow powered by Amazon EventBridge that triggers an AWS Step Functions workflow when an audio file is uploaded to a specified Amazon Simple Storage Service (Amazon S3) bucket. Masu. This serverless processing pipeline is built around Amazon Transcribe, which processes call recordings and converts them from speech to text. Processed transcription notifications are sent to an Amazon Simple Queue Service (Amazon SQS) queue, which helps you isolate your architecture and restart Step Functions state machine workflows. This architecture uses AWS Lambda as the transcription processor to store processed transcriptions in Amazon OpenSearch Service tables.
The call processing workflow uses a custom machine learning (ML) model built by Intact running on Amazon Fargate and Amazon Elastic Compute Cloud (Amazon EC2). OpenSearch transcriptions are further enhanced with these custom ML models to perform component identification, named entity recognition, speaker role identification, sentiment analysis, personally identifiable information (PII) redaction, and more. provides valuable insights. Regular improvements to existing and new models add valuable insights to extract, including call reasons, script adherence, call outcomes, and sentiment analysis across different business sectors, from complaints to personal lines. This architecture uses Amazon DynamoDB to control queue limits. Call transcriptions are compressed from WAV to MP3 format to optimize storage costs on Amazon S3.
Machine learning operations (MLOps)
Intact also built an automated MLOps pipeline using Step Functions, Lambda, and Amazon S3. This pipeline provides self-service capabilities for data scientists to track ML experiments and push new models to an S3 bucket. It gives data scientists the flexibility to perform shadow deployments and capacity planning, and seamlessly switch between models for both production and experimental purposes. Additionally, the application provides a backend dashboard tailored to MLOps functionality, ensuring smooth monitoring and optimization of machine learning models.
Frontend and API
The CQ application provides a robust search interface created specifically for call quality agents and includes powerful auditing capabilities for call analysis. The backend of the application leverages Amazon OpenSearch Service for search functionality. This application also uses Amazon Cognito to provide single sign-on for secure access. Finally, use Lambda functions for orchestration to retrieve dynamic content from OpenSearch.
The application provides a trend dashboard customized to provide actionable business insights and helps agents identify critical areas to allocate their time. Intact uses data from sources like Amazon S3 and Snowflake to build comprehensive business intelligence dashboards that show key performance metrics like silent periods and call handling time. This feature allows call quality agents to drill deeper into call components to facilitate targeted agent coaching opportunities.
Call quality trend dashboard
The following image is an example of a call quality trends dashboard that shows the information available to agents. This includes the ability to filter by multiple criteria, such as date and language, average processing time by component and unit manager, and voice and silence time.
result
The new system has significantly increased efficiency and productivity. Audit speed increased by 1,500% and number of calls reviewed increased by 1,500%. Additionally, by building MLOps on AWS alongside the CQ solution, the team reduced delivery of new ML models to deliver analytics from days to just hours, increasing auditor efficiency by 65% I let it happen. This resulted in a 10% reduction in time per call and a 10% reduction in average hold time as agents received targeted coaching to improve their conversations with customers. This efficiency allows auditors to use their time more effectively to develop coaching strategies, improve scripts, and train agents.
Additionally, the solution provided intangible benefits such as extremely high availability with no significant downtime since 2020 and high cost predictability. The solution’s modular design also allows for robust deployment, significantly reducing new release time to less than an hour. This also contributes to the failure rate at the time of implementation being almost zero.
conclusion
In conclusion, Intact Financial Corporation revolutionized their approach to customer service by implementing CQ leveraging AWS AI services. This case study demonstrates the transformative power of AI and speech-to-text technology to improve customer service efficiency and effectiveness. The design and functionality of this solution will enable Intact to use generative AI in future transcription projects. As a next step, Intact will further leverage this technology by using Amazon Transcribe streaming for real-time transcription and deploying a virtual agent to provide relevant information and recommended responses to human agents. We plan to leverage this technology even further.
Intact Financial Corporation’s journey is an example of how implementing AI can significantly improve service delivery and customer satisfaction. Customers who want to jumpstart call analytics should consider Amazon Transcribe Call Analytics for live call analytics and agent assistance and post-call analytics.
About the author
Etienne Brouillard is the AWS AI Principal Architect at Intact Financial Corporation, Canada’s largest property and casualty insurance provider.
Amidani I’m a senior technical program manager at AWS with a focus on AI/ML services. During her career, she has focused on delivering innovative software development projects to the federal government and large corporations in a variety of industries, including advertising, entertainment, and finance. Ami has experience driving business growth, implementing innovative training programs, and successfully managing complex, high-impact projects.
Prabir Sekri I am a Senior Solutions Architect in Enterprise Financial Services at AWS. During his career, he has focused on digital transformation projects within large companies in a variety of industries, including the finance, multimedia, telecommunications, and energy and gas sectors. His background includes designing and building DevOps, security, and enterprise storage solutions. In addition to technology, Prabir has always had a passion for playing music. He is a pianist, composer, and arranger who leads a jazz ensemble in Montreal.
