Call center ai · Production

MaestroQA integrates Amazon Bedrock to enable open-ended conversation analytics at enterprise scale

The problem

MaestroQA's enterprise clients needed to analyze high volumes of unstructured customer interaction data—call recordings, chat messages, and emails—at enterprise scale, but the existing keyword-based rules engine could not handle open-ended questions where users phrase the same intent in many different ways, and clients handle anywhere from thousands to millions of customer engagements monthly.

First attempt

The keyword-based rules engine broke down on semantically equivalent phrases that share no keywords, and manual compliance monitoring at a lending client was error-prone, causing many compliance risks to be missed.

Workflow diagram · grounded in source
1
Call transcription
ai_action
“When customers receive incoming calls at their call centers, MaestroQA employs its proprietary transcription technology, built by enhancing open source transcription models, to transcribe the conversations”
2
Sentiment analysis
ai_action
“MaestroQA offers sentiment analysis for customers to identify the sentiment of their end customer during the support interaction”
3
Open-ended query submission
trigger
“When a customer submits an analysis request through MaestroQA's web application”
4
Transcript retrieval and Bedrock dispatch
ai_action
“an ECS cluster retrieves the relevant transcripts from Amazon S3, cleans and formats the prompt, sends them to Amazon Bedrock for analysis using the customer's selected FM”
5
Cross-region inference routing
routing
“Cross-Region inference dynamically routes traffic across multiple Regions, providing optimal availability for each request and smoother performance during these high-usage periods”
6
Results stored and surfaced
output
“stores the results in a database hosted in Amazon Elastic Compute Cloud (Amazon EC2), where they can be retrieved by MaestroQA's frontend web application”
Reported outcome

MaestroQA can now run open-ended queries across millions of transcripts; a lending company detects compliance risks with almost 100% accuracy on 100% of conversations; an education company increased sentiment-scoring coverage from 15% to 100% of conversations; and cross-Region inference delivers twice the throughput compared to single-Region inference.

Reported metrics
Client monthly customer engagement volumeanywhere from thousands to millions of customer engagements monthly
Compliance risk detection accuracy (lending company)almost 100% accuracy
Conversations analyzed for compliance (lending company)100%
Sentiment scoring sample size increase (education company)increased their sample size from 15% to 100% of conversations
Show all 5 reported metrics
client monthly customer engagement volumeanywhere from thousands to millions of customer engagements monthly
compliance risk detection accuracy (lending company)almost 100% accuracy
conversations analyzed for compliance (lending company)100%
sentiment scoring sample size increase (education company)increased their sample size from 15% to 100% of conversations
throughput with cross-Region inference vs single-Regiontwice the throughput compared to single-Region inference
Reported stack
Amazon BedrockAmazon ECSAmazon S3Amazon EC2Amazon ComprehendAmazon CloudWatchIAMAWS SDKClaude 3.5 SonnetClaude 3 HaikuMistral 7b/8x7bCommand RLlama 3.1
Source
https://aws.amazon.com/blogs/machine-learning/revolutionizing-customer-service-maestroqas-integration-with-amazon-bedrock-for-actionable-insight?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

MaestroQA can now run open-ended queries across millions of transcripts; a lending company detects compliance risks with almost 100% accuracy on 100% of conversations; an education company increased sentiment-scoring…

What tools did this team use?

Amazon Bedrock, Amazon ECS, Amazon S3, Amazon EC2, Amazon Comprehend, Amazon CloudWatch, IAM, AWS SDK, Claude 3.5 Sonnet, Claude 3 Haiku.

What results were reported?

Client monthly customer engagement volume: anywhere from thousands to millions of customer engagements monthly; Compliance risk detection accuracy (lending company): almost 100% accuracy; Conversations analyzed for compliance (lending company): 100%; Sentiment scoring sample size increase (education company): increased their sample size from 15% to 100% of conversations (source-reported, not independently verified).

What failed first in this deployment?

The keyword-based rules engine broke down on semantically equivalent phrases that share no keywords, and manual compliance monitoring at a lending client was error-prone, causing many compliance risks to be missed.

How is this call center ai AI workflow structured?

Call transcription → Sentiment analysis → Open-ended query submission → Transcript retrieval and Bedrock dispatch → Cross-region inference routing → Results stored and surfaced.

Related call center ai cases