Normalized claim
Time: 75% decrease
Reported ~75% reduction in time spent on research tasks.
TP ICAP – ClientIQ CRM assistant using Amazon Bedrock Knowledge Bases (RAG)
Use case typeConversational assistantsUpdated Jun 13, 2026
TP ICAP's Innovation Lab built ClientIQ, a production-ready CRM assistant for searching and summarizing tens of thousands of Salesforce vendor meeting notes. The solution uses Amazon Bedrock Knowledge Bases for RAG, Amazon Bedrock Evaluations for automated quality testing, and a text-to-SQL path for structured queries. It includes permission-aware retrieval with Okta group claims, source attribution, custom ingestion from Salesforce to Amazon S3, OpenSearch Serverless vector search, and CI/CD evaluation. The article reports that after launch with 20 users, research time fell by about 75% and insight quality improved.
- Organization
- TP ICAP
- Industry
- Finance
- Location
- United Kingdom
- Published
- October 2025
Reported outcomes
−75%
timeTime & speed
Strategic outcomes
New product / capabilityBuilt a CRM assistant for notes search and summarizationBetter decisions & insightImproved insight quality and comprehensivenessSpeed & agilityReduced research effort for usersRisk & complianceAdded permission-aware retrieval and traceability
Why do we believe this?Outcome claims, sources, and evidence checks
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- TP ICAP
- Provider
- AWS
- Maturity
- Exploring
- Linked source
- AWS Machine Learning Blog
It includes permission-aware retrieval with Okta group claims, source attribution, custom ingestion from Salesforce to Amazon S3, OpenSearch Serverless vector search, and CI/CD evaluation
Customer identity supportedSource describes one deploymentMaturity supported
Primary read
Use case focus
Showing 3 of 4
- 1Conversational analytics
- 2Knowledge management
- 3CRM assistant
- Tens of thousands of vendor meeting notes in Salesforce were underutilized.
- Business users spent hours manually searching and summarizing qualitative CRM data.
- The solution needed accurate answers, traceability, and strict permission controls.
- Built ClientIQ as a natural-language assistant that routes queries to either RAG over Amazon Bedrock Knowledge Bases or SQL generation over structured data.
- Used custom Salesforce ingestion, custom chunking, metadata tagging, OpenSearch Serverless, Athena, and permission-aware filtering.
- Integrated Amazon Bedrock Evaluations into development and CI/CD to compare models, chunking, embeddings, and prompt parameters.
- Initial launch had 20 users.
- Reported ~75% reduction in time spent on research tasks.
- Stakeholders reported improved insight quality and comprehensiveness.
Architecture
Frontend React app hosted in Amazon S3 behind an internal-only Application Load Balancer; API Gateway and AWS Lambda orchestrate a WebSocket-based conversational workflow; requests flow through Amazon SNS and Amazon SQS to downstream processing; structured queries use Amazon Athena over AWS Glue Catalog; unstructured retrieval uses Amazon Bedrock Knowledge Bases with Amazon Titan embeddings in Amazon OpenSearch Serverless; security is enforced with Okta group claims mapped to CRM metadata; Amazon Bedrock Evaluations is integrated into the CI/CD pipeline for quality testing.
Sources & evidence1
Evidence: Medium50/100Evidence strength
- Customer explicitly identified
- Deployment status explicitly supported
- Quantified outcome available
- Technical implementation details available
Type: Blog PostPublished: Oct 17, 2025Publisher: AWSEvidence: VendorConfidence: Medium
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Customer evidence
Provider evidence
AI patterns
Browse the catalog
Was this useful?
Community
Comments
Loading comments...