OmniChannel Agentic Core (OAC)

Inspiration

Businesses and non-technical end-users often struggle to navigate complex enterprise dashboards just to perform routine tasks like checking status, updating records, or querying internal tools. We wanted to eliminate this friction by bringing agentic AI directly to the tools people already use every single day: WhatsApp and Email. By turning simple chat threads and inbox messages into active control interfaces, we realized we could make cloud automation completely invisible and effortless.

What it does

How we built it

We built OmniChannel Agentic Core (OAC) using a modular AWS-native architecture: ​Ingestion & Gateway: Configured AWS End User Messaging Social for WhatsApp webhooks and Amazon Simple Email Service (SES) via @aws-sdk/client-ses for email parsing. ​Orchestration Layer: Used Amazon Bedrock alongside AWS Lambda to run an autonomous agentic control loop. The agent evaluates intent, manages dynamic context, and invokes tools. ​Storage & Security: Managed session state in Amazon DynamoDB and logged artifacts to Amazon S3, securing all public entry points with Amazon CloudFront and AWS WAF.

Challenges we ran into

​State Normalization Across Async Channels: Synchronous API requests and multi-step email threads carry completely different timing profiles. We had to design an event-driven session manager in DynamoDB to maintain user context cleanly across asynchronous channels. ​Payload Parsing & Formatting Constraints: Stripping unstructured HTML signatures from inbound SES emails while enforcing WhatsApp's strict template formatting rules required building dedicated string-sanitization adapters. ​Execution Latency Optimization: Agentic tool-calling loops can introduce delay. We optimized prompt sizing and tool schema schemas to keep total response latency \Delta t \le 2.5\text{s} for WhatsApp transactions.

Accomplishments that we're proud of

Production-Ready SES Implementation: We didn’t just integrate email; we fully implemented the modern @aws-sdk/client-ses (v2) client and adhered to high-volume SES SMTP standards, making the agent instantly viable for high-traffic environments. ​Meta WhatsApp Prize Qualification: We successfully deployed a non-trivial integration of AWS End User Messaging Social to manage real-time, interactive conversations with agentic tool execution. ​ACE Opportunity Generation: Securing the required opportunity ID (O20261028XXXXX) with the exact marketing campaign code within Partner Central confirmed our ability to align development with key AWS partner ecosystem milestones.

What we learned

Serverless Agent Execution: Mastering the orchestration of multi-step, asynchronous agentic workflows purely within transient AWS Lambda and Fargate environments was crucial for maintaining a responsive system on messaging platforms. ​The Power of Minimalist UX: We confirmed that a complex backend system does not require a complex frontend; if an agent is reliable, a simple WhatsApp or email interface is often preferred by users for efficiency. ​Zero-Trust Integration: We gained deep expertise in securing third-party SaaS hooks (e.g., WhatsApp webhooks) using AWS WAF and CloudFront, ensuring no backend calls were triggered without robust validation.

What's next for OmniChannel Agentic Core (OAC)Channel Expansion: Our immediate roadmap includes expanding beyond WhatsApp and SES to integrate SMS (via Amazon SNS/EUM) and corporate chat platforms like Slack or Microsoft Teams, making the core truly omni-channel.

​Sophisticated Memory Management: We plan to transition from simple session state storage in DynamoDB to integrating vector databases for long-term user memory and context-aware responses, improving personalization. ​Proactive Notification Agent: The system will evolve from being entirely reactive (triggered by a message) to being proactive—using backend system alerts (e.g., inventory drops) to initiate conversations with users over WhatsApp or email

Built With

• amazon
• amazon-web-services
• amazonbedrock