Customer support · Production

Verisk builds PAAS AI on AWS to deliver 96–98% reduction in premium audit research time

The problem

Premium auditors using Verisk's PAAS platform struggled to find accurate information across a large repository of documents, with manual searches being time-consuming, slow, and yielding inconsistent or incomplete results.

Workflow diagram · grounded in source
1
User submits audit query
trigger
“answering questions and quickly retrieving and summarizing multiple PAAS documents like class guides, bulletins, rating cards, etc.”
2
Keyword extraction by LLM
ai_action
“Keywords are extracted from user questions and previous conversations to be used for creating the new summarized prompt and to be input to Verisk's knowledge base retrievers to perform vector similarity search”
3
Hybrid document retrieval
integration
“a solution was implemented to combine a sparse bm25 search in combination with the dense vector search to create a hybrid search approach, which yielded much better context retrieval results”
4
Claude response generation
ai_action
“For scenarios where latency was more important and less reasoning was required, Anthropic's Claude Haiku was the perfect solution. For other scenarios such as question answering using provided contexts where it was more important for the…”
5
Custom evaluation API check
validation
“This custom API evaluates the answers based on three major metrics”
6
Customer feedback loop
feedback_loop
“The team actively collects and analyzes feedback from customers to identify potential data issues or problems with the generative AI responses. This analysis helps pinpoint specific areas that need improvement.”
Reported outcome

PAAS AI reduced processing time per specialist by 96–98%, transforming hours of manual review into minutes, and enabling Verisk's subject matter experts to focus on more strategic initiatives.

Reported metrics
Processing time reduction per specialist96–98%
Research time acceleration vs traditional methods98%
Manual review timehours reduced to minutes
Reported stack
ElastiCacheAmazon BedrockOpenSearchAnthropic's ClaudeClaude HaikuClaude SonnetSnowflakeJira
Source
https://aws.amazon.com/blogs/machine-learning/turbocharging-premium-audit-capabilities-with-the-power-of-generative-ai-verisks-journey-toward-a-sophisticated-conversational-chat-platform-to-enhance-customer-support?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

PAAS AI reduced processing time per specialist by 96–98%, transforming hours of manual review into minutes, and enabling Verisk's subject matter experts to focus on more strategic initiatives.

What tools did this team use?

ElastiCache, Amazon Bedrock, OpenSearch, Anthropic's Claude, Claude Haiku, Claude Sonnet, Snowflake, Jira.

What results were reported?

Processing time reduction per specialist: 96–98%; Research time acceleration vs traditional methods: 98%; Manual review time: hours reduced to minutes (source-reported, not independently verified).

How is this customer support AI workflow structured?

User submits audit query → Keyword extraction by LLM → Hybrid document retrieval → Claude response generation → Custom evaluation API check → Customer feedback loop.

Related customer support cases