Problem overview

GrayHair Software offers a wide range of products, services, and deep postal industry expertise. However, users often face challenges in finding answers to their questions. This can lead to time-consuming efforts, such as asking colleagues or manually searching through manuals or Confluence for information.

These situations highlight the need for faster access to relevant documentation and insights. In a world increasingly powered by AI, GrayHair’s goal is to empower internal users with instant access to knowledge about GrayHair Software’s products and the USPS.

Despite the potential of AI, there are natural concerns about its adoption—stemming from issues like hallucinations, perceived complexity, or fears of job displacement. The project, in partnership with Tech 42, aimed to bridge the gap by introducing a practical, user-friendly AI tool to demonstrate how automation can enhance efficiency in everyday workflows while thoughtfully addressing concerns.

Solution

Tech 42 and GrayHair worked together to design a solution that would work well for GrayHair’s organization. To add context-aware AI responses, Bedrock Knowledge Bases were used. Bedrock Knowledge Bases is a pre-built, ready-to-go Retrieval Augmentation Generation (RAG) service provided by AWS. Knowledge Bases have connectors making it easy for AI to be aware of data in S3 and Confluence.

To interact with AI in a conversational manner, Bedrock Agents were leveraged. Bedrock Agents are a severless framework for agentic workloads in AWS. They have built-in memory, guardrails, and integrations with Bedrock Knowledge Bases. This streamlined development and made it easier to integrate knowledge and security with GrayHair’s AI tools.

The agent and knowledge base provided the ability for users to ask questions about GrayHair practices, procedures, and easy access to the vast knowledge available to GrayHair. The rest of the architecture was built to support integration with Slack in a secure manner. An application load balancer with HTTPS and Route53 for domain name integration added to security and reliability.

AWS architecture overview

The AWS architecture to provide context-aware answers to users in Slack was comprised of the following components:

Stack 1 (Slack Bot)

• ECR
• ECS Fargate Service/Taskrunning Java application to interact with Slack API
• Route53 (DNS entry for Slackbot for the load balancer for a constant URL)
• Load Balancer for HTTPS traffic
• WAF Web ACL
• Cloudwatch Logs
• IAM Role

Stack 2 (Bedrock Agent)

• Lambda (Daily sync of data sources)
• EventBridge (To schedule Lambda)
• Bedrock Agent
• Bedrock Knowledge Base
• Bedrock Datasources (S3, Confluence)
• Bedrock Guardrails
• OpenSearch Serverless (Vector database)
• S3 Bucket (For data source)
• Secrets Manager (For Confluence secret)
• IAM Role

Usage and benefits

The AI solution enables users to instantly access GrayHair’s extensive postal knowledge base through natural language queries. Rather than manually searching through documentation or interrupting colleagues, users can ask questions directly such as:

“Does the USPS allow punctuation in an address block?”

“Where should the indicia be placed on retail mail?”

“What’s the aspect ratio for a machinable letter?”

The system leverages AWS Bedrock Knowledge Bases to retrieve contextually relevant information from GrayHair’s Confluence and S3 repositories, delivering accurate answers without requiring expertise in AI or ML.

Measurable impact

• Handled over 100 inquiries within the first week of deployment
• Improved response consistency across departments
• Reduced workload on subject matter experts and managers
• Facilitated seamless cross-departmental knowledge transfer

The regular synchronization process allows GrayHair’s postal experts to update documentation to reflect current regulations and remove outdated information. These updates are immediately available in the knowledge base, ensuring the AI always provides the most current postal regulations and practices.

Tech 42 is thankful to have partnered with GrayHair on this project. We used AWS’s best tools to build a solution that actually helps their staff get work done faster. The solution makes a real difference in their daily operations, giving employees quick access to the information they need when they need it.