Powered by AWS Kendra and Bedrock 

About Our Client

Our client is a social enterprise focused on supporting nonprofits in securing grant funding from local governments and third parties. They recognised that crafting successful bids to do this requires specialised writing and business expertise, along with the ability to reference a broad range of documents. This process is often time-consuming and repetitive, diverting attention away from the core mission of these establishments. To help their clients overcome this challenge, they identified the need to streamline and simplify the bid creation process. 

The Challenge

Many bids require extensive research into past materials, including reference documents and previous bids. While our client excelled at managing this historical information, they realised that much of the process-driven research could be automated using AI. They had previously experimented with publicly available LLMs but were dissatisfied with the results, particularly regarding concerns over data privacy and the quality of output. This is where we came in… 

The Solution

To address these concerns, they commissioned us to develop a Proof of Concept (PoC) using AI to automate parts of the bid management process. After exploring if AWS Bedrock would be a better fit for this use case, we realised that a more comprehensive solution was needed. This resulted in a hybrid system that combined best-in-class tools to deliver both automation and security. 

The tools we implemented included: 

  • Python Streamlit for a user-friendly interface to manage data ingestion and querying. 
  • AWS Bedrock as a private language model for generating structured outputs. 
  • AWS Comprehend to format and process various data sets, including documents and PDFs. 
  • AWS Kendra to organise backend data and generate meaningful search results. 
  • AWS S3 for secure file storage. 
  • PostgreSQL for storing configuration and metadata.

This solution allowed our client to privately manage their data and use an intelligent search engine to quickly generate content for bids. 

In Phase 2 of the PoC, we enhanced the user interface and transitioned to the more robust Python Django framework to meet the growing needs of the project. 

The Outcome

By implementing this AI-based bid management tool, our client achieved numerous key benefits including: 

  • Increased efficiency: Automation reduced the time spent on repetitive research tasks. 
  • Enhanced privacy: Secure handling of sensitive bid data eliminated concerns about public exposure. 
  • Improved accuracy: AI-generated insights from AWS Kendra and Bedrock helped produce more relevant and structured bid content. 
  • Scalable solution: The system is now flexible enough to adapt to future growth and additional features. 
  • Simplified user interface: The intuitive design allowed non-technical users to easily navigate and query the system. 

Our Learnings

This project showed that AI alone is not always the answer. Many business applications of AI require a balance of automation, security and customisation. Off-the-shelf tools like public LLMs may lack the necessary privacy and control for sensitive data. Our experience with AWS frameworks, combined with our expertise in analysis and development, enabled us to deliver a solution that met our client’s needs.  

If you would like to explore this further, feel free to contact us with your questions.