What challenges did Intact Financial Corporation face before AI implementation?

Manual call auditing could not scale with up to 20,000 calls per day. Quality teams needed a workable solution within 6 months and a long-term solution within 1 year. The company wanted better insight into customer service scripts, coaching opportunities, and call outcomes across departments.

What was the impact of the AI solution?

1,500% increase in auditing speed and number of calls reviewed. 65% more efficient auditors due to faster ML model delivery. 10% reduction in agent handling time. 10% reduction in average hold time. No major downtime since 2020 and near-zero deployment failure rate.

Source

ExpandedEvidence: Low35/100

Intact Financial: Call Quality (CQ) for contact-center auditing with Amazon Transcribe

Use case typeCustomer experience analyticsUpdated Jun 13, 2026

Intact Financial Corporation built an automated Call Quality (CQ) solution to audit up to 20,000 contact-center calls per day across on-premises and cloud systems. The workflow uses Amazon Transcribe, AWS Step Functions, Amazon S3, Amazon SQS, Amazon OpenSearch Service, AWS Lambda, Amazon EC2, and custom ML models for entity extraction, speaker identification, sentiment analysis, PII redaction, script adherence, and call outcome analytics. The company also built an MLOps pipeline to speed model delivery from days to hours, provide dashboards and coaching insights, and improve agent handling and hold times.

Organization: Intact Financial Corporation
Industry: Insurance
Location: Canada
Published: October 2024

Reported outcomes

Speed: +1.5%

Time & speed

Time: −65%Time: −10%

Catalog median for time & speed deployments: +59% across 138 reported metrics. Compare benchmarks →

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Quantified impact: 1.5% increase

AWS BlogOct 16, 2024Blog postInferred claimLow evidence strength

1,500% increase in auditing speed and number of calls reviewed.

Normalized claim

Time: 65% decrease

AWS BlogOct 16, 2024Blog postInferred claimLow evidence strength

65% more efficient auditors due to faster ML model delivery.

Normalized claim

Time: 10% decrease

AWS BlogOct 16, 2024Blog postInferred claimLow evidence strength

10% reduction in agent handling time.

Normalized claim

Time: 10% decrease

AWS BlogOct 16, 2024Blog postInferred claimLow evidence strength

10% reduction in average hold time.

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Intact Financial Corporation
Provider: AWS
Maturity: Unknown
Linked source: AWS Blog

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 3

1Call Quality
2Speech Analytics
3Contact Center Analytics

Ingest calls from both on-premises and cloud contact centers into Amazon S3.
Trigger AWS Step Functions workflows for transcription and enrichment.
Use Amazon Transcribe to convert audio to text and store transcripts in Amazon OpenSearch Service.
Apply custom ML models on Amazon EC2 for named entity recognition, speaker role identification, sentiment analysis, PII redaction, reason-for-call detection, script adherence, and call outcome analytics.
Run an automated MLOps pipeline with Step Functions, Lambda, and S3 to support experiment tracking, shadow deployments, and model switching.

Technologies

Amazon Transcribe AWS Step Functions Amazon S3 Amazon SQS Amazon OpenSearch Service AWS Lambda Amazon EC2

1,500% increase in auditing speed and number of calls reviewed.
65% more efficient auditors due to faster ML model delivery.
10% reduction in agent handling time.
10% reduction in average hold time.
No major downtime since 2020 and near-zero deployment failure rate.

Architecture

Call acquisition from on-premises and cloud contact centers feeds an AWS Step Functions workflow triggered by Amazon S3 uploads. Amazon Transcribe converts audio to text, transcripts are stored in Amazon OpenSearch Service, and custom ML models running on Amazon EC2 enrich transcripts with entity recognition, speaker role identification, sentiment, PII redaction, script adherence, and call outcome analytics. An automated MLOps pipeline uses Step Functions, Lambda, and S3 to manage experiment tracking, shadow deployments, and model switching.

Sources & evidence1

Evidence: Low35/100Evidence strength