--------------------------------------------------------------------------------------------

Case Study: GenAI Knowledge Base for RAS's Audit Staff

About the Client

RAS (Retail Asset Solutions) are a leading provider of inventory management and stocktaking services in the UK and Europe. Their audit teams work across diverse retail environments, supporting clients with compliance-critical stock checks, loss prevention, and operational insights. With a workforce that often operates out of hours and across multiple locations, fast and accurate access to training and customer documentation is essential to maintain consistency and efficiency at scale.

Challenge

RAS's audit staff frequently needed to reference customer and company training documents during audits, but accessing the right information quickly was difficult. Documentation was often unstructured, inconsistently formatted, and time-consuming to search. This slowed audit efficiency and sometimes forced staff to rely on guesswork or leave items unrecorded. RAS wanted a solution that would provide staff with fast, accurate, and compliant access to knowledge in real time.

Solution

Cloud Combinator partnered with RAS to design and deploy a production-level Generative AI workflow on AWS. The solution was built around Amazon Bedrock for retrieval-augmented generation, AWS Lambda for orchestration, and Amazon S3 for authoritative document storage. Guardrails within Bedrock were configured to ensure responses remained safe and brand-compliant.

Audit staff now interact with the system through an API-driven chatbot built on AWS Lambda Function URLs. When a query is submitted, the workflow retrieves the most relevant information from Bedrock knowledge bases, applies prompt-engineering logic to refine accuracy and tone, and returns the output in real time. A DynamoDB layer tracks knowledge base metadata and sessions, enabling scalability and flexibility.

Automation was central to the design. Whenever training documents are uploaded, updated, or deleted in Amazon S3, the knowledge base syncs automatically, eliminating manual upkeep. Infrastructure deployments were delivered via AWS CloudFormation for repeatability, while CloudWatch monitoring provides visibility into workload health.

Security was addressed at every layer, including encryption at rest and in transit, IAM policies with least privilege, VPC private-by-default design, and AWS account governance guardrails. Finally, RAS’s technical team received detailed runbooks and a knowledge transfer session, giving them ownership of operations and future improvements. Cloud Combinator also coordinated with AWS to secure funding credits, offsetting initial adoption costs.

Outcomes

• Faster access to information:

  Audit staff can now get accurate, context-aware answers in real time.

• Consistent and compliant outputs:

  Guardrails ensure responses align with brand and safety standards.

• Reduced manual work:

  Automatic syncing between Amazon S3 and Bedrock keeps knowledge bases always up to date.

• Scalable architecture:

  Built on AWS-native services with DynamoDB session tracking, ensuring the solution can grow with RAS’s needs.

• Empowered teams:

  Technical staff were onboarded with documentation and training, giving them ownership of the system.

  
  

• Cost efficiency:

  Leveraged AWS funding and credits to lower overall deployment costs.

Results Summary

The deployment of the GenAI knowledge base has transformed how RAS audit staff access critical documentation. Previously, auditors spent variable amounts of time searching through files, often resorting to guesswork or omission. Now, staff receive consistent and accurate answers in seconds, improving audit speed and removing uncertainty in stocktaking tasks.

The system was architected for resilience. Automated deployments, S3 versioning, DynamoDB point-in-time recovery, and multi-AZ redundancy ensure high availability and disaster recovery readiness. Security baselines—covering encryption, IAM, and VPC design—further aligned the solution with AWS Well-Architected standards.

From a financial perspective, the architecture is cost-efficient and aligned with AWS funding support. The main driver of consumption is Amazon Bedrock model usage, with additional overhead from Lambda, DynamoDB, and S3 services. Cost guardrails and monitoring are in place, allowing RAS to scale confidently while keeping spend predictable and sustainable.

RAS’s technical team now manages the platform confidently using established KPIs and runbooks, supported by monitoring dashboards. The engagement has been validated as a strong fit for AWS’s Generative AI Consulting Services specialization, demonstrating secure, scalable, and outcome-driven delivery.