About the Customer
About Client
The customer is a U.S.-based industrial manufacturer whose inside sales and distribution teams support detailed product selection and technical questions. Employees rely on product catalogs, FAQs, images, and technical datasheets to answer customer and distributor inquiries.
Location
United States
Industry
Industrial Manufacturing
CORE AWS SOLUTION
GenAI Product Knowledge Assistant
Executive Summary
RSNA Cloud Connect delivered a production-ready product knowledge assistant within 30 days. The solution reduced manual lookup time by approximately 60 percent, achieved more than 95 percent accuracy, and established a reusable platform for future distributor and ERP integrations.
Challenges
Product information was spread across separate documents, email threads, and static knowledge sources. Employees spent too much time searching for accurate answers, which slowed responses and made consistent service harder to maintain.
The company wanted an AI assistant that could centralize product knowledge and answer technical questions in real time. The first release also needed to support later expansion. Planned phases included distributor access and ERP integration, so the initial platform required stable APIs, repeatable deployments, version control, rollback support, and operational monitoring from launch.
Solution
RSNA Cloud Connect designed and deployed a multi-phase GenAI product knowledge assistant on AWS. The architecture combined retrieval-augmented generation with managed AWS services to improve answer accuracy while keeping the application modular and measurable.
Centralized product knowledge
- Amazon S3 stores product datasheets, images, and structured metadata in Markdown and JSON formats.
- Amazon OpenSearch Service indexes product catalogs, FAQs, and technical documents for hybrid semantic search.
- Amazon Comprehend extracts key phrases that enrich the searchable product content.
Context aware answers
- Amazon Bedrock with Claude 3.5 Sonnet provides the generative model for product and technical responses.
- Retrieval-augmented generation supplies relevant company content before the model creates an answer.
- Prompt engineering and cache optimization improve response time and reduce repeated model calls.
Repeatable DevOps delivery
- AWS Lambda runs the ingestion, transformation, and search functions as Docker container images.
- Amazon Elastic Container Registry stores tagged image versions, making releases reproducible and supporting rollback to a known version.
- Amazon API Gateway publishes stable API contracts for current internal applications and planned distributor-facing clients.
Monitoring and analytics
- Amazon CloudWatch captures performance, latency, and feedback logs.
- Amazon QuickSight dashboards track usage, system health, engagement, and product search activity.
- Structured logging and feedback data support KPI-based tuning after launch.
Implementation Approach
- Centralize product documents and metadata in Amazon S3.
- Create a retrieval layer with Amazon OpenSearch Service and semantic enrichment from Amazon Comprehend.
- Connect retrieved context to the generative model in Amazon Bedrock.
- Package application functions as versioned container images and deploy them through AWS Lambda and Amazon ECR.
- Expose stable application interfaces through Amazon API Gateway.
- Measure latency, usage, system health, and feedback through Amazon CloudWatch and Amazon QuickSight.
Success Metrics
Phase 1 delivered a fully functional assistant for the Inside Sales team within 30 days. The reported results compare the new assistant with the previous manual and self-hosted baseline.
Delivery
Within 30 days
Lookup time
About 60 percent lower
Answer accuracy
More than 95 percent
Query throughput
More than 70 percent higher
Infrastructure cost
About 45 percent lower
Response latency
Under 15 seconds on average
The architecture also gave the company versioned releases, stable API contracts, live operational dashboards, and a tested foundation for planned distributor and ERP phases.
Customer Voice
Inside Sales and Distribution teams needed a faster way to find accurate product and technical information without searching through separate emails, FAQs, and datasheets. The new assistant brought those sources into one searchable experience and gave employees real-time answers for common product questions. The platform also addressed the customer’s longer-term requirement for a solution that could extend to distributors and ERP workflows without replacing the first release.
Partner Impact
RSNA Cloud Connect turned the customer’s product knowledge requirements into a production-ready AWS platform within 30 days. The team designed the retrieval workflow, implemented the GenAI application, defined stable API contracts, and created a versioned release process for the Lambda functions. RSNA Cloud Connect also built monitoring and analytics into the first phase. This gave the customer measurable data on accuracy, latency, usage, and feedback from launch. The resulting platform supports current Inside Sales users and provides a controlled path for distributor access and ERP integration in later phases.
AWS Enablement
AWS services provided the compute, storage, search, AI, API, release, and monitoring capabilities required for the solution. Amazon Bedrock supplied the generative model, while Amazon OpenSearch Service retrieved relevant product content. Amazon S3 centralized source documents and structured metadata, and Amazon Comprehend enriched the content with extracted key phrases. AWS Lambda and Amazon ECR supported containerized, versioned application releases. Amazon API Gateway created stable interfaces for current and future applications. Amazon CloudWatch and Amazon QuickSight gave the customer operational logs and dashboards for performance, usage, and product search activity. This managed architecture allowed the platform to scale with demand while keeping releases repeatable and costs tied more closely to actual usage.