Redpumpkin.AI at AWS Agentic AI in Production Roadshow 2026

At the AWS Agentic AI Showcase in Jakarta on May 12, 2026, five AWS partners demonstrated production-ready AI solutions to a room of enterprise decision-makers. Hosted at Hotel Indonesia Kempinski's Grand Ballroom, the half-day event wasn't about roadmaps or prototypes, it was about live systems already running in production across ASEAN.

Among the five solutions presented, Redpumpkin.AI and Innovation Cloud Services demonstrated something that resonated with the room: an agentic AI system that doesn't just answer customer questions, but completes entire workflows, from booking appointments, guiding purchases, to resolving inquiries, all without human handoff.

Here's what we showed, and why it matters for any business dealing with growing customer demand and limited team bandwidth.

Why do most AI chatbots still frustrate customers?

The event opened with a challenge every attendee recognized. Customers reach out through WhatsApp, website chat, social media, and they expect instant, complete answers. Not a link to an FAQ. Not "an agent will get back to you." They want their appointment booked, their order tracked, their question resolved.

Most chatbot deployments fail here because they're retrieval-only: they can surface information but can't act on it. They can tell a patient that Dr. Rani is available on Thursday, but they can't book the slot. They can describe a product's features, but they can't guide a customer through checkout. The gap between "answering" and "doing" is where customer experience breaks down and revenue leaks.

What does agentic customer experience actually look like?

What we demonstrated at the showcase was a system built on two core capabilities that bridge that gap.

Healthcare: booking appointments through MCP

In the healthcare scenario, a patient messages a clinic on WhatsApp, the channel most Indonesians already use daily. The AI assistant understands the request, checks real-time appointment availability, and confirms the booking. End to end, no human intervention required.

The key enabler is MCP (Model Context Protocol) which connects the AI directly to the clinic's scheduling system, patient database, and internal tools through a single standardized interface. No custom API integration for each system. No manual lookups by staff. The assistant operates as an agent that can read from and write to the systems it needs to complete the task.

This matters because the bottleneck in most healthcare operations is the AI's ability to act on requests rather than the ability to understand a request. MCP eliminates the integration friction that keeps most chatbots stuck in "information-only" mode.

Retail: turning inquiries into purchases through RAG

In the retail scenario, a customer visits an online store and asks about a product: sizing, availability, alternatives. The AI assistant answers instantly, drawing from the store's actual product catalog, inventory database, and policy documents through RAG (Retrieval-Augmented Generation).

The distinction matters. RAG means the system retrieves relevant information from your data before generating a response. It doesn't guess. It doesn't hallucinate product specs or invent pricing. Every answer is grounded in your actual catalog, and when the customer is ready, the assistant guides them through checkout in the same conversation.

The result is a single conversation that moves from product discovery to purchase, with intelligent cross-sell and upsell recommendations at the right moment. Customer service becomes a revenue channel, not just a cost center.

What's behind the system that makes this work?

Behind both demonstrations is a production-grade architecture built on four capabilities.

RAG ensures every answer is grounded in your documents and data rather than stale training memory. This is critical for businesses where product catalogs, policies, and schedules change frequently.

MCP connects the AI to your internal systems such as scheduling, inventory, CRM, and knowledge bases through a single protocol. No custom integration per system, no rip-and-replace.

Guardrails keep every response safe, on-brand, and within defined boundaries. The AI won't go off-script, make unauthorized commitments, or surface information it shouldn't.

Agentic workflows let the AI execute multi-step tasks autonomously from initial inquiry through resolution, escalating to humans only when confidence drops below a set threshold, with a clear explanation of why.

Everything runs in a secure, cloud-hosted environment: VPC-isolated, load-balanced, and accessible only through controlled endpoints.

What results did the demonstration show?

Based on proof-of-concept deployments and business case analysis, the agentic customer experience system demonstrated measurable impact across both verticals.

In healthcare:
- Up to 85% faster response times compared to staff-dependent processes,
- Up to 90% reduction in booking errors through AI-driven scheduling, and
- Up to 60% recovery of after-hours inquiries that would otherwise be lost, enabled by 24/7 availability.

In retail:
- Up to 35% higher conversion from inquiry to purchase when AI guides the full journey in a single conversation, - Up to 25% increase in average order value through intelligent cross-sell and upsell, and,
- Up to 50% reduction in pre-sale support tickets as RAG handles product questions instantly.

These are projected estimates from previous business case analysis. Actual results vary by implementation, data quality, and operational context.

Why does this matter for enterprises across ASEAN?

The showcase demonstrated a shift that's already underway. The question for most enterprises is no longer "should we use AI for customer experience?" but "how do we move from a chatbot that answers questions to a system that completes workflows?"

The businesses seeing real returns are the ones connecting AI to their operational systems, not just their knowledge bases. MCP and RAG together create a system that knows your data and can act on your behalf, with the guardrails and auditability that enterprise deployments require.

Whether you're in healthcare, retail, financial services, or any sector facing growing customer demand, the architecture demonstrated at the Jakarta showcase is designed to move from proof-of-concept to production in weeks, not months.

See it working with your data.

Every business has different systems, channels, and customer workflows. We're happy to walk you through how the agentic customer experience platform would work with your specific setup and what results you could realistically expect.

Book a Technical Discovery Call

About Redpumpkin.AI

Redpumpkin.AI exists for AI projects where the hard part is to make AI work reliably inside complex enterprise environments. We help organisations choose, build, and operate the right AI architecture across commercial and open-weight models, multiple cloud environments, and demanding business workflows. Our strength lies in structured evaluation, deep engineering, and production deployment, turning AI ambition into systems that are accurate, governed, scalable, and ready for real work.

‍