Email Prioritizer Agent

• 

  Architecture Diagram

Inspiration

One day, while having coffee with a friend, we talked about how our lives are often consumed by social media, yet emails remain one of the most important tools in our daily routine. With busy work and personal lives, we often overlook critical emails. Even with filters and rules, some of the most important messages still slip through. This frustration sparked the idea for the Email Prioritizer Agent, a smart way to help users focus on what truly matters.

What it does

The Email Prioritizer Agent helps users automatically sort and prioritize emails based on importance. It currently classifies emails into three categories as High, Medium, and Low priority. With some prompt adjustments, it can easily be customized to suit different users' needs or business contexts.

How we built it

We built the agent using Amazon Q as the base. We began by outlining a simple concept, then created a detailed specification document to define edge and boundary cases. Next, we designed the prompt using few-shot prompting techniques to guide the model’s responses. Once finalized, we integrated it using AWS Bedrock with Claude Haiku and supporting AWS services such as Lambda, IAM, S3, Secrets Manager, EventBridge, and CloudWatch Logs. It took several iterations to fine-tune and perfect the agent’s performance.

Challenges we ran into

We encountered several challenges along the way: 1)Clearly defining the prompt’s purpose and expected outcomes. 2)Narrowing a broad idea into a well-defined, achievable goal. 3)Connecting securely with external email systems like Gmail. 4)Turning a simple idea into a structured specification and architecture was more complex than expected.

Accomplishments that we're proud of

1)Successfully building a fully functional, end-to-end working agent. 2)Achieving the original vision of an intelligent system that prioritizes emails effectively.

What we learned

1)Coming up with ideas is easy — executing them is the real challenge. 2)Big visions often meet harsh realities, but iteration leads to success. 3)Breaking down complex problems into smaller, manageable tasks is a key skill in project development.

What's next for Email Prioritizer Agent

We plan to integrate this agent into a broader agentic workflow, where a human-in-the-loop can review recommendations and take automated actions based on prioritized emails. This will make email management even smarter and more interactive.

Built With

• amazonbedrock
• amazoncloudwatch
• amazoneventbridge
• amazonq
• amazons3
• aws-iam
• aws-secretsmanager
• awslambda
• bash
• css
• flask
• github
• gmailapi
• googleoauth
• gunicorn
• html5
• javascript
• json
• python
• render.com
• vanillajs