AMAZThai-AI ID

Development Internet of Things Machine LearningManagement & Governance Media Services Migration & Transfer Networking & Delivery Quantum Technologies Satellite Security Identity & Compliance Storage

Comment

AMAZThai-AI ID

The Autonomous, Continuous Omnichannel Customer Concierge Powered by AWS CDS & Amazon Bedrock AgentCore Human-First, Problem-First, Privacy-First Architecture.

Inspiration

"Human-First, Problem-First, Privacy-First | AI Workflow & Logic Architect | Building responsible, real-world AI for modern organizations."

Every system, software, and workflow we build stems from a single fundamental reality: unresolved friction among human stakeholders. Whether in large industrial supply chains, traditional procurement, or the everyday disconnect between customers and corporate call centers, broken communication wastes immense value.

Coming from a non-traditional tech background and navigating massive industry shifts, I experienced these pain points firsthand. When left as a solo operator, I refused to halt. Over the past year, I proved that deep domain logic coupled with AI could replace the workload of an entire conventional enterprise team—orchestrating end-to-end blueprint estimation, quantity surveying, automated Value Engineering (VE), vendor compliance, and 3-tier risk-margin pricing models within a single month.

That journey forged a core conviction: AI succeeds only when human operational empathy and technical orchestration understand each other.

For decades, customer engagement has been trapped in a broken dynamic: customers trapped in rigid, frustrating IVR trees, and call centers drowning in repetitive operational strain. AMAZThai-AI ID was born to bridge this exact gap. It takes the battle-tested, resilient operational logic developed through relentless bootstrapping and maps it directly onto modern agentic communication infrastructure—delivering real value, dignity, and seamless continuity to end users.

What it does

AMAZThai-AI ID transforms fragmented messaging channels into a unified, stateful, and autonomous conversational ecosystem built on AWS Communication Developer Services (CDS) and Amazon Bedrock AgentCore.

Instead of static chatbots or one-way notification blasts, AMAZThai-AI ID delivers continuous, contextual customer engagement across the entire lifecycle:

  • True Cross-Channel Continuity: A conversation can initiate via WhatsApp for an initial customer query, transition seamlessly into an interactive RCS (Rich Communication Services) action card for authenticated confirmation or scheduling with one tap, fall back to high-deliverability SMS when offline, and finalize formal documentation and audit trails via Amazon SES (Email)—never losing conversational context.
  • Autonomous Intent & Action Orchestration: Powered by Amazon Bedrock AgentCore, the agent dynamically resolves multi-step customer inquiries, triggers backend workflows (e.g., ticket creation, service dispatch, price tier calculations), and detects when to gracefully escalate to human operators with full context summaries.
  • Continuous Behavioral Feedback Loop: Gathers first-party conversational signals to continuously optimize product offerings, detect service bottlenecks, and refine response parameters without rigid hard-coded rules.

How we built it

AMAZThai-AI ID is engineered with a modular, serverless, and event-driven architecture designed for enterprise-grade resilience and zero downtime:

Architectural Pillars:

  1. Communications Backbone (AWS CDS / EUM):
    • AWS End User Messaging (EUM): Coordinates two-way messaging across WhatsApp Business API, RCS Business Messaging, and SMS.
    • Amazon Simple Email Service (SES): Automatically generates and delivers official receipts, audit summaries, and technical specifications directly to customer inboxes.
  2. Cognitive Agent Engine:
    • Amazon Bedrock AgentCore: Acts as the brain, maintaining dynamic session memory, task decomposition, and prompt guardrails.
    • Custom Action Groups (AWS Lambda): Translates customer intents into real-time operational commands (API calls, data retrieval, scheduling).
  3. State Management & Governance:
    • Amazon DynamoDB: Stores customer session states, channel preferences, and interaction histories to guarantee context persistence across channel hops.
    • Privacy & Security First: Adheres to strict international data protection standards (GDPR, PDPA, CCPA alignment) with end-to-end encryption in transit and at rest.

Challenges we ran into

  • Cross-Channel Session Reconciliation: Ensuring that an interaction starting on WhatsApp immediately links to the correct customer record when an SMS or RCS trigger fires without duplicate identity collisions or dropped state.
  • Context Continuity vs. Latency: Fine-tuning Bedrock AgentCore prompts and tools to deliver sub-second responses while preserving deep multi-turn context and handling nuanced, mixed-language (Thai-English code-switching) phrasing.
  • Extreme Resource Constraints: Building and iterating without large corporate funding or paid enterprise tiers required ruthless architectural discipline—optimizing serverless invocations, minimizing token overhead, and engineering every component to run lean, stable, and cost-efficient.

Accomplishments that we're proud of

  • Zero-Drop Context Switching: Successfully routed live customer inquiries from WhatsApp to RCS and Amazon SES while maintaining 100% conversation state continuity.
  • Production-Grade Logic Built Solo: Demonstrated that deep field domain experience combined with AWS cloud-native primitives can outpace legacy, bloated enterprise call center software.
  • Empowering the Underserved: Designed the interaction flows specifically with accessibility in mind—ensuring intuitive, low-friction access for non-technical users, older demographics, and field operators.

What we learned

  • Speed Is Irrelevant Without Purpose: Fast outputs mean nothing if they don't solve real, measurable user pain. Empathy and operational understanding must always precede AI generation.
  • True Omnichannel Means Relationship Continuity: It is not about blasting the same message across four platforms; it is about meeting the customer where they are, remembering where they left off, and completing their job-to-be-done with minimal friction.
  • AWS CDS + Bedrock AgentCore Is a Game Changer: Decoupling communication channels from rigid business logic and connecting them directly to agentic workflows transforms messaging from a cost center into a strategic asset.

What's next for AMAZThai-AI ID

  • AWS Marketplace Packaged Solution: Package the core architecture into a repeatable, multi-tenant solution template ready for AWS Partners and enterprise clients across Southeast Asia.
  • Physical-to-Digital Field Mobility Integration: Connect the conversational layer directly to real-time telemetry, localized logistics dispatch, and community solar infrastructure monitoring.
  • Advanced Multi-Modal Agents: Expand Bedrock AgentCore integrations to ingest and audit field images, PDF blueprints, and voice memos submitted via WhatsApp/RCS with immediate structured feedback.

Inspiration

Customer engagement is stuck between rigid scripted bots and one-way notification blasts. We wanted a conversation that reaches the customer in the right channel, remembers context, and ends with an action.

What it does

AMAZ-OTS Concierge uses AWS CDS as the communications backbone for an airport-handover concierge.

  1. A customer starts on [FILL: WhatsApp via AWS End User Messaging Social / SMS via AWS End User Messaging].
  2. An Amazon Bedrock-powered adviser [FILL: Bedrock AgentCore if used, otherwise "Bedrock Converse"] reads the shared case brief and drafts a reply.
  3. A human reviews and approves the reply.
  4. The approved summary is delivered via [FILL: SES / SMS / WhatsApp]. [FILL: describe ONLY the channel hand-offs you actually demonstrated in the video.]

How we built it

  • CDS used at runtime: [FILL: AWS End User Messaging SMS v2 SendTextMessage / Social Messaging SendWhatsAppMessage / SES v2 SendEmail; list exactly what you ran]
  • AI: Amazon Bedrock Converse via an authenticated Lambda/API Gateway handler in ap-southeast-1.
  • Safety controls: server-owned recipient allowlist, explicit approval flag, delivery disabled by default, no credentials in the repo, fail-closed AI route.
  • Evidence: [FILL: provider message IDs, CloudWatch log screenshots, timestamps]. We report provider acceptance separately from delivery.

Challenges

Cross-channel identity linking, consent, idempotent sending and delivery callbacks. [FILL: which of these are implemented vs. planned.]

Accomplishments

[FILL: only verified results, e.g. "End-to-end WhatsApp → Bedrock draft → human approval → SES summary in X seconds."]

What we learned

[FILL]

What's next

Durable conversation store, delivery callbacks, multi-tenant packaging for AWS Marketplace.

Meta WhatsApp prize (only if applicable)

[FILL: how AWS End User Messaging Social was used, template/session message details]

ACE opportunity ID

[FILL] with campaign code: AWS CDS Agentic AI Hackathon -Sept. 2026.

Built With

Share this project:

Updates