---
title: "Odie Open Data Bot Case Study | Data-Driven AI"
canonical: "https://data-driven.com/case-study/odie-bot-the-open-data-hub-virtual-helpdesk-agent/"
description: "See how TfNSW used the Odie helpdesk bot to guide Open Data Hub users, answer common questions, and learn from search patterns."
---

TfNSW · Open Data

# Odie helps people _find transport data_

The Open Data Hub bot answered common questions, guided people to datasets, and gave the support team a view of common searches.

![Illustration of Transport for NSW Open Data and the Odie helpdesk bot](/_astro/About-us-image-12.BxJBM-Qn_Z13KoIk.webp)

The service problem

## Public data was available, but it was not always easy to find

TfNSW publishes transport datasets through its Open Data Hub. Users still had trouble with dataset names, search, and API registration. Basic questions then reached the support team.

[Read the TfNSW platform FAQ](https://opendata.transport.nsw.gov.au/sites/default/files/2023-08/Open%20Data%20Platform%20Replacement%20FAQs.pdf)

> “The Open Data Hub needed a solution to help data users navigate the Open Data Hub website and easily find, view and use transport public data. This solution just gives us that, so that we can understand how our users are searching for data and improve our services.”

Yvonne Lee

Program Manager, Open Data Hub, TfNSW

Where users got stuck

## The helpdesk load began with findability

The project focused on common public-data questions, not transport operations or travel advice.

1.  Datasets used different names and could be hard to locate.
2.  Users needed help to understand which dataset matched their task.
3.  API registration questions added manual work for support staff.

How Odie worked

## A managed answer layer sat inside the Open Data Hub

Odie became the first line for common questions. Natural-language matching handled misspelled words and phrases. The support team could add and update the question-and-answer content.

1.  A visitor asked a question in the embedded bot.
2.  Odie matched the request to managed help content and relevant data.
3.  The Open Data team reviewed search reports and updated the answers.
4.  Questions outside that content still needed a person and an escalation path.

Recorded result

## Faster access stayed a qualitative result

The published project record says Odie surfaced less visible datasets, answered common questions at any time, and showed the team how people searched. It does not publish query counts, response-time measures, or a cost baseline.

### Scope the next bot on its own evidence

Set the approved answer set, owner, fallback route, quality measure, and support baseline before comparing a new service with Odie.

## Discuss a public helpdesk workflow

Bring the common questions, content owners, search logs, and escalation path. We will use them to define the first useful scope.

[Contact Data-Driven](/contact-us/)

[Back to home](/)

## Want to learn more?

Explore the rest of the site or get in touch with the team.

[Back to home](/)