An internal AI workspace for UniMelb staff

Client

The University of Melbourne

Years

2023 – 2025

My roles

Lead Product Designer · Product Strategy

Overview

Spark AI is the in-house AI workspace for UniMelb staff. It gives academics and professional staff a single place to experiment with large language models, work with university documents, and share useful AI tools with colleagues. Instead of everyone quietly pasting content into whatever AI site they prefer, Spark AI offers a supported, governed option that actually fits how people work at the university.

My Role

I was the sole designer for Spark AI. I worked with product and engineering to define the core pillars of the product and how they fit together as well as the minuet details of the interactions and UI.

The Product

Spark AI starts with Chat. Staff can create spaces for different subjects, teams or ideas, and spin up threads that use different AI models and behaviours.

For work that needs to be grounded in university content, staff can use AI Studio to upload and index (vectorise) documents into datasets, then wire those into projects that use retrieval-augmented generation (RAG) to answer questions from those sources.

Spark Apps sit on top as templates: preconfigured projects, connected to the right datasets, so other staff can reuse a pattern without setting everything up from scratch.

THE PROCESS

A little insight into the process for this project and build.

The business problem

Through workshops and conversations with senior staff, we unpacked why the university needed its own AI tool, where people were already using public models, and what “success” should look like for Spark AI.

Governance and risk

We worked with governance, legal and security teams to understand the boundaries: which data is allowed, what needs to be logged, and what patterns are completely off limits.

Staff interviews

I spoke with academics and professional staff across faculties to see how they were actually using AI in their day-to-day work, what they were worried about, and where a supported tool could genuinely help.

Use case mapping

We turned a long list of ideas into a clear set of patterns – quick chat, working with documents, support-style flows – which later became chat, datasets/projects and Spark Apps.

Landscape analysis

I looked at how other tools and universities were approaching AI workspaces, paying attention to where people got lost or overloaded, and where we could keep the model simpler.

Design and tech workshops

Product, design and engineering met regularly to sketch flows, test assumptions and line up the product vision with the technical realities of the underlying platform.

Interaction design and prototyping

I designed and prototyped the core flows for spaces, threads, datasets, projects and apps, then used clickable prototypes with staff to check that the end-to-end experience actually made sense.

Naming and language

We iterated on labels and microcopy for things like datasets, indexing and retrieval so the product could stay honest about what it does without feeling like a research tool.

UI design

Once the model was stable, I translated it into a consistent interface and component set for Spark AI, designing key screens and states so future features could plug into the same system.

Impact

Spark AI is now the endorsed way for UniMelb staff to use AI at work. Staff use it for everyday drafting and analysis, and also for more advanced projects that involve indexing sensitive data and building Spark Apps for their teams. The structures and patterns from this work now carry through into the other AI tools within the university.

Feedback

— Jane Smith, CEO

“The professionalism and expertise demonstrated by the team were unparalleled, and we couldn’t be more pleased with the results. We’ve been able to achieve new levels of innovation and success in our industry.”