Team Name

DLPT

Team Members

Project Description

GovAsk Project Description

1. Problem Statement

The Australian Government holds credible datasets across all areas of operation. However, it has not yet identified a suitable AI chatbot to form part of its long-term digital strategy. Existing options such as ChatGPT or Gemini fall short of government requirements, particularly around reliability and accuracy, where standards must exceed 90%. The core challenge is not the lack of data but a trustworthy AI system that can access, interrogate, and present this data with the level of trust, accuracy, and accountability that government decision-making demands.

2. Proposed Solution

In order to address this challenge, we proudly present an AI chatbot explicitly designed for government use. Delivers accurate, auditable, and transparent answers by working directly with datasets provided by government agencies. Rather than generating speculative responses, it grounds every answer in verified data sources. Its ability to learn from and operate on inserted datasets enables agencies to interrogate their information conversationally and trust that the insights they receive are reliable and accountable. Using the RAG (Retrieve Augmented Generation) model for fast dataset query and the ability for the application to query a large database with high efficiency. Moreover, with the help of a local LLM model, data privacy is ensured as no data is uploaded onto the Internet.

3. Technology Stack

• Large Language Model: Llama 3 8B Instruct
  
• Front-End Development: Next.js, TailwindCSS, Streamlit UI
  
• Back-End Development: Python
  
• Data Management: Vector database for indexing and storing embeddings, enabling fast and accurate retrieval of relevant data
  

4. Key Features

Key features of GovAsk:
- Conversational Data Interrogation: Allows staff to ask natural-language questions across multiple government datasets without needing advanced technical skills

- Grounded Responses Only: GovAsk will give answers based on what it has learnt from the datasets inserted into itself

- Auditability and Transparency: Each interaction produces an audit trail, showing the data used and the reasoning steps taken, ensuring accountability

- Ethical and Secure by Design: Built with strict attention to privacy, fairness, and transparency in algorithmic decision-making

- Guided Question Scaffolding: Offers prompts and suggestions to help users refine their queries and reach meaningful insights faster

5. Example

Users can ask GovAsk questions relating to the datasets they inserted into GovAsk:

Meanwhile, GovAsk will refuse to answer questions that do not relate to the datasets:

6. Impact Statement

By enabling government staff to interrogate datasets reliably and transparently, GovAsk supports evidence-based, accountable decision-making and helps agencies unlock the full value of their data assets while maintaining the highest standards of trust and accuracy.

Try It Out

You can head to our website GovAsk to install and test the application!

Data Story

The chart shows the rapid rise of the global AI market from 2018 to 2025. Growth was slow and steady in the early years, staying under $100 billion until 2021. From 2022 onward, the market began climbing sharply, reaching around $250 billion in 2023 and $400 billion in 2024. The most dramatic change comes in 2025, where the market value skyrockets to nearly $1.85 trillion, highlighting how quickly AI is expanding and how its economic impact is accelerating worldwide.

With the ambition of creating a chatbot to help the government, we came up with GovAsk. GovAsk obtains datasets from the Northern Territory Government’s Open Data Portal to provide staff with accurate, grounded insights. By enabling conversational interrogation of these datasets, it reduces manual workload, highlights issues such as coverage gaps more efficiently, and marks the next chapter of working with technology - helping the government gradually adapt to a modern world where AI is a key driver of success. GovAsk is also programmed to refuse questions outside its datasets to ensure accuracy. When a user submits an unrelated query, the system alerts them that the information is unavailable rather than generating a speculative response. This safeguard prevents misinformation and strengthens user trust in the chatbot’s reliability.