Case Study · 2024

AI Knowledge Assistant

Trilingual knowledge platform for Sri Lanka's Ministry of Labour

B2GGen AIKMSMultilingual

RoleHead of Product & Operations

Users10,000+ officers

LanguagesSinhala · Tamil · English

ClientMinistry of Labour, Sri Lanka

Overview

Instant Trilingual Knowledge at the Field Level

The AI Knowledge Assistant is a lightweight, mobile-first platform designed to provide development officers with quick, reliable access to essential protocols and knowledge. Tailored for the unique challenges faced in the field, it serves as a centralised trilingual knowledge repository, ensuring consistency and accuracy in decision-making at every level of the service.

The Problem

A Service Delivery Gap Built on Assumptions

Support GapOfficers relied on assumptions or sporadic office communication, leading to inconsistent guidance and delays in public service delivery.

AvailabilityNo 24/7 access to accurate information meant critical decisions were deferred or made with incomplete context.

ScaleA solution needed to work seamlessly across mobile devices in three languages and two character sets (Sinhala, Tamil, Latin).

Research & Discovery

Listening Before Building

MethodologySeveral rounds of focus group discussions with Ministry of Labour stakeholders and frontline development officers across regions.

Device InsightMost officers owned smartphones but not PCs — mobile-first was non-negotiable, not a preference.

Language InsightProficiency highest in Sinhala and Tamil; officers typed in both native character sets and Latinised forms requiring flexible NLP.

Trust RequirementPublic service protocols demand high accuracy — officers needed to trust every answer before acting on it.

Strategy

Prioritising for the Field

Mobile-FirstOptimised for smartphones as the sole access device — no desktop dependency in the design or architecture.

Trilingual AIEnabled seamless interaction and knowledge retrieval in Sinhala, Tamil, and English with a single query interface.

Scalable BackendDesigned to manage a vast multilingual knowledge repository that grows as Ministry policies evolve.

Simple Auth & UXAccessibility for users with minimal technical expertise — friction in sign-in or navigation would kill adoption.

Design & Build

From Knowledge Base to Conversational Interface

FrontendMinimalistic chat-based interface with secure authentication, prioritising ease of navigation for non-technical users.

BackendIndexed an extensive multilingual knowledge base in PDF and markdown format with AI-powered query routing.

AI ChatbotNatural language processing across three languages with robust accuracy and reliability checks before surfacing responses.

UI Showcase

Metrics

What We Track

EngagementDAU/MAU (stickiness), average session length, and time between sessions as proxy for knowledge dependency.

AccuracyRegular assessments by Ministry subject-matter experts to maintain a high standard of knowledge reliability.

ImpactHours saved per officer and measurable improvements in public service delivery timelines.

Next Steps

Areas for Improvement

Predictive AI

Expand capabilities to support proactive suggestions and push notifications before officers need to ask.

Offline Access

Integrate offline mode for officers in remote areas with limited or no mobile data connectivity.

Ministry Analytics

Add admin dashboards providing real-time insights into usage patterns and knowledge gaps across regions.