NT Data Insight Assistant

Team Name

ZeroHallucinations (0H)

Team Members

Project Description

Our project is the development of an HR Leave Data Assistant prototype — a conversational AI system that enables government agencies to interrogate HR datasets with high accuracy, transparency, and accountability. Unlike general-purpose chatbots, this assistant is designed to meet government-grade standards, where accuracy must approach 100% and every response requires a clear audit trail.
The system is built on a modular coding approach, separating dataset management, query analysis, and response generation to ensure traceability and reliability. It integrates trust scoring, statistical validation, audit logging, and privacy safeguards such as role-based access and anonymisation.
From a Human-Centred AI (HCAI) perspective, we have prioritised explainability, ethical safeguards, and user control, aligning our work with the Northern Territory Government’s Artificial Intelligence Assurance Framework. This ensures the assistant operates with security, fairness, and transparency at its core.
As this is a prototype, we are actively exploring ways to improve accuracy as more data becomes available and the system learns from user interactions. To support continuous improvement, we will also publish user surveys, allowing feedback to directly shape future iterations.
Ultimately, this project demonstrates how AI can transform HR data analysis in government settings — making information not only easier to access, but also trustworthy, auditable, and aligned with ethical and compliance standards.

Data Story

Government agencies manage thousands of datasets but often struggle to extract insights with the accuracy and accountability required for decision-making.
Our HR Leave Data Assistant prototype addresses this challenge by combining:
Government-grade accuracy (near 100 percent correctness required, not 90)
Explainability and transparency (every answer has an audit trail and trust score)
Privacy safeguards (role-based access, anonymisation, and suppression of small groups)
Human-Centred AI principles — keeping humans in control, ensuring fairness, and preventing bias
We have aligned with the NT Government’s Artificial Intelligence Assurance Framework, ensuring that our prototype respects security, transparency, and fairness.
As this is a prototype, accuracy will improve as more data becomes available and as the system learns from user feedback surveys.
Our vision is to make government HR data accessible, trustworthy, and auditable — helping decision-makers ask questions like:
What is the average sick leave in Finance this year?
Are there red flags in leave-taking patterns?
Which departments are at risk of exceeding leave quotas?
This project is a step toward responsible, human-centred AI that makes data interrogation safe, transparent, and reliable for government use.