Wellbeing answers in plain language

A SaaS B2B business working with Cloud Combinator on AWS. The client is anonymised at their request.

The proof of concept had three focus areas, each needed to turn a stored catalogue into a reliable conversational service.

There are two flows. On the administration side, the client manages course data through an Amazon API Gateway REST endpoint that invokes an AWS Lambda function to add, update or delete courses. Each change is written to Amazon DynamoDB and mirrored as a JSON file with access metadata in Amazon S3. A S3 event then triggers a second synchronisation step that keeps an Amazon OpenSearch Service vector index and the Amazon Bedrock Knowledge Base up to date, so the catalogue behind the assistant is always current.

Cloud Combinator delivered a scoped, architected proof of concept that meets the client's success criteria: exporting course data into permissioned storage, making it queryable in natural language through OpenSearch and Bedrock, and standing up a Lex pipeline that answers employee questions and logs each conversation.

• Conversational access — Employees can ask for wellbeing support in plain language and receive answers grounded in the client's real catalogue rather than generic guidance.
• Always current — Admin changes flow automatically from DynamoDB and S3 into OpenSearch and the Bedrock Knowledge Base, so answers reflect the live catalogue.
• Permissioned by design — Access-control metadata ensures employees only see the opportunities their employer offers, with encryption at rest and EU data residency.
• Handover — The client received a CloudFormation-deployed environment, a working demonstration and documentation, ready to carry the service into continued development.

With a proven conversational layer over its catalogue, the client is positioned to make wellbeing support easier to discover and act on, and to extend the same natural-language approach across more of its service.