Case Studies

AI implementation work at Mekari Qontak

During my time at Mekari Qontak I've implemented AI chatbot and Agentic AI solutions for over 50 enterprise clients across logistics, property, finance, healthcare, and retail. Client configurations are confidential, but the following case studies outline the types of problems I've worked on and how I approached them.

Named ClientLogistics & Fleet Rental

Transgo

Logistics & Fleet Rental

Context

A fleet rental company handling inbound inquiries, availability checks, and order creation entirely through WhatsApp.

Challenge

The client needed to move beyond a simple FAQ chatbot — customers were trying to actually browse available vehicles and complete rentals through the chat interface. That required the AI to maintain context, access live inventory data, and execute multi-step transactions without a human in the loop.

Approach

Designed and implemented Mekari's first Agentic AI production deployment. Built an autonomous flow covering: inbound inquiry triage, fleet availability lookup via API integration, customer registration, and order creation. Prompt architecture was structured around tool-use capabilities — the agent knows what actions it can take and when to take them. Knowledge base covers vehicle catalogue, pricing, terms, and escalation rules.

Outcome

First implementation in Mekari's product history to achieve fully autonomous transactional AI — no human handoff required for standard deal completion. The implementation served as the reference architecture for subsequent agentic deployments.

Agentic AIAPI IntegrationWhatsAppFleet / Logistics

AnonymisedProperty

Property — Anonymised

Context

A property developer handling inbound leads from multiple marketing channels — WhatsApp, web, and social — with a high volume of repetitive pre-sales inquiries.

Challenge

Sales agents were spending most of their time answering the same questions about unit availability, pricing, and payment schemes. The client wanted to automate the first layer of engagement without losing the warmth of a human conversation.

Approach

Built a knowledge base structured around the full product catalogue: unit types, pricing tiers, available payment schemes, promo periods, and location details. Prompt engineering focused on tone — formal but conversational, in Bahasa Indonesia. Implemented handoff logic to route qualified leads to human agents at the right moment based on intent signals.

Outcome

Significantly reduced first-response time and agent workload on repetitive queries. Lead qualification became more consistent — the AI collected standard qualifying information before human handoff.

ChatbotLead QualificationBahasa IndonesiaProperty

AnonymisedHealthcare

Healthcare — Anonymised

Context

A healthcare provider using WhatsApp as the primary patient communication channel for appointment scheduling, clinic information, and general health queries.

Challenge

Healthcare AI requires extra care around accuracy and scope. The client needed a system that was genuinely helpful for logistics (appointments, locations, operating hours) but clearly bounded — never venturing into medical advice territory.

Approach

Designed the knowledge base and prompt constraints specifically around scope management. The system handles scheduling, FAQs about services, and location/hours reliably. Prompt architecture includes hard constraints on what the AI will and won't address, with clear handoff to clinical staff for anything that falls outside scope. Tested extensively for edge cases where users might push the AI toward medical advice.

Outcome

Clean deployment with high patient satisfaction on the logistics layer. Zero incidents of the AI overstepping into clinical territory in production.

ChatbotHealthcareScope ManagementBahasa Indonesia

Additional case studies and detailed implementation notes are available on request. Client-specific configurations, prompt architectures, and knowledge base structures are shared under NDA.