24/04/2025
LEKIR TECH X JABATAN DIGITAL NEGARA
Last week, Lekir Tech recently had the honour of engaging with the team at Jabatan Digital Negara (JDN) to present our flagship innovation, Assistnify AI Agent Digital Workers. What was initially planned as a one-hour briefing evolved into a 2.5-hour in-depth discussion focused on a key challenge: Ensuring Reliability in AI agents.
โ
๐ Understanding the Challenge :
โ
Many AI solutions today are powered directly by Large Language Models (LLMs) such as ChatGPT, Claude, or Gemini. While these models excel in understanding and generating natural language, they are fundamentally designed to predict the next likely word โ not to guarantee factual accuracy.
โ
โ
This leads to hallucinations โ where the AI confidently produces information that is false, misleading, or completely made up.
In real-world use cases, hallucinations can lead to:
โ Incorrect customer responses
โ Missed or wrongly executed workflows
โ Fake data or reputational harm
โ
We, Lekir Tech believe in building AI that is reliable, auditable, and safe โ especially in mission-critical workflows. Our Large Action Model (LAM) framework ensures that hallucination is avoided through a structured, layered architecture:
โ
LLMs used only for understanding intent (NLP)
โ
RAG (Retrieval-Augmented Generation) to ensure fact-based answers
โ
Intelligent Decision Support Systems (IDSS) for logical decision-making
โ
Governed, closed-loop knowledge base for traceable outputs
โ
๐ค Moving Forward with Purpose
We deeply appreciate the Jabatan Digital Negara team's thoughtful questions and their shared commitment to ethical and responsible AI adoption. Our conversation covered everything from technical depth to why it matters for Malaysiaโs future and we agreed that:
๐ฐHallucination in AI workflows is unacceptable
๐ฐResponsible AI is the only way forward
๐ฐMalaysia has the talent and drive to lead in AI innovation
At Lekir Tech, weโre proud to contribute to that vision.