Enhancements and Solutions at Sincronismo with AWS Generative AI

The Solution

Sincronismo partnered with CloudDog, an AWS Advanced Tier Services partner, to find the ideal solution. Together, the companies implemented a serverless architecture utilizing Amazon Transcribe for automatic call transcription and Amazon Bedrock for generative analysis. This combination enabled Sincronismo to offer new services, such as call auditing and consulting, creating more value for its clients.

Thanks to AWS generative AI, Meet Insights was developed—a service that uses AI prompts to evaluate service quality, identify issues, and summarize inbound calls. The architecture was designed to ensure that these insights were generated securely and at scale, allowing Sincronismo to quickly access the most relevant information for its clients.

Architecture

The architecture implemented by CloudDog for Sincronismo is composed of specific layers to ensure security, efficiency, and compliance:

• Amazon S3: Central repository for storing multimedia files sent by legacy systems (audio/video) as well as input and output data.

• Amazon SQS (Multimedia): Queue responsible for managing messages related to multimedia file processing, ensuring asynchronous and scalable workflows.

• AWS Lambda (Audio & Video Process): Performs initial validations and transformations on multimedia files sent to the SQS queue.

• Amazon Transcribe: Automates the conversion of audio files into text, transcribing data stored in Amazon S3.

• Amazon DynamoDB: NoSQL database storing transcription results and associated metadata, offering high scalability and low latency.

• Amazon DynamoDB Streams: Tracks real-time changes in DynamoDB and triggers Lambda functions for additional processing.

• AWS Lambda (Process Stream): Processes data from DynamoDB Streams and sends notifications or events to other systems.

• AWS SNS: Notifies the legacy system (via webhook) about processing completions or results.

• Amazon Bedrock: Tool for advanced analytics, such as classification or semantic insights applied to transcribed data.

• AWS Lambda (Insights): Executes specific logic for advanced analysis of transcribed data before sending it to Amazon Bedrock or other queues.

• Amazon CloudWatch: Monitors and collects metrics across all services involved in the workflow, ensuring visibility and problem detection.

• AWS CloudTrail: Provides detailed auditing of API activities and actions on the AWS services used.

• AWS Config: Ensures compliance and traceability of AWS resource configurations.

• AWS Budgets: Monitors costs associated with the services in the workflow and alerts about predefined financial limits.

This modular structure allows Sincronismo to efficiently and securely manage the growing demand for its communication services.

Results

The solution implemented by CloudDog brought several significant improvements to Sincronismo:

• Cost Reduction: The adoption of Amazon Bedrock and generative AI resulted in a significant reduction in operational costs for the call analysis process.

• High Accuracy in Analysis: The developed solution achieved 99% accuracy in transcription analysis in Portuguese, ensuring higher service quality.

• Process Automation: Between July and September 2024, 20% of call audits were automated, freeing human resources to focus on more strategic issues.

• Regulatory Compliance: AI integration contributed to enhanced compliance with regulations, safeguarding sensitive data during automated responses.

• Real-Time Analysis: Sincronismo can now perform real-time call analysis, using prompts to evaluate service quality, operator performance, and customer satisfaction.

These improvements resulted in enhanced operational efficiency, increased customer satisfaction, and a significant reduction in operational costs, allowing Sincronismo to continue expanding its operations and meeting the needs of Brazil’s corporate communication market.