Asure builds post-call analytics with Amazon Bedrock and Amazon Q in QuickSight
Asure, a workforce management and HR software company, needed a scalable way to analyze thousands of customer support call transcripts after calls. The team converted call audio to transcripts, generated metadata such as summary, root cause, next steps, and callback or resolution indicators with Amazon Bedrock, and used Amazon Comprehend for additional sentiment and entity signals. Amazon Q in QuickSight enabled natural-language analysis across aggregated call data and individual-call insights so analysts could query trends and issues without writing SQL.
- Organization
- Asure
- Industry
- Professional Services
- Location
- United States
- Published
- March 2025
Reported outcomes
14 days
timeTime & speed
Strategic outcomes
Primary read
Use case focus
Showing 3 of 3
- 1Post-call analytics
- 2Contact center analytics
- 3Natural language BI
- Analyzing thousands of call transcripts manually took about 14 days.
- The existing approach did not scale well for identifying themes, root causes, and next steps.
- Slower insight generation delayed product and customer support improvements.
- Amazon Transcribe converted call audio into transcripts.
- Amazon Bedrock generated call metadata fields including summary, root cause, topic, next steps, and callback or resolution indicators.
- Amazon Comprehend provided additional sentiment and entity processing.
- AWS Step Functions orchestrated the workflow across the pipeline.
- Amazon S3 and Amazon Athena stored and queried transcript-derived data.
- Amazon Q in QuickSight enabled natural-language Q&A over call analytics for trend analysis and reporting.
- Asure reduced analysis time from about 14 days to minutes or even seconds.
- The team could identify trends and pain points earlier.
- The solution helped prioritize product development and customer support improvements.
Architecture
Call audio is transcribed with Amazon Transcribe. AWS Step Functions orchestrates downstream processing where Amazon Bedrock generates transcript metadata and Amazon Comprehend adds sentiment/entity analysis. The outputs are stored in Amazon S3 and queried in Amazon Athena, then surfaced in Amazon QuickSight with Amazon Q for natural-language analytics. The article also describes a human-in-the-loop evaluation UI and Bedrock-based evaluation metrics.
Implementation partners1
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.