Wathaga: building AI around data sovereignty

Most AI platforms start from the same assumption: pour in all the data, and let the model sort it out. For the communities behind Wathaga, that assumption was a non-starter. Their data included stories, artwork and video, much of it sensitive and some of it sacred, and the one thing they could not do was hand it to a system that might store it, share it, or train on it.

Wathaga is the platform we built with Kowa Collaboration to answer exactly that: a way for First Nations communities to measure and evaluate their work, using AI, without giving up ownership or control of any of it.

The problem with off-the-shelf measurement

Monitoring and evaluation, the work of showing whether a program actually helped, usually runs on tools built for spreadsheets and survey forms. That works until your most important evidence is a recorded conversation in language, a piece of artwork, or a story told on country. Conventional platforms can't hold that kind of data, let alone make sense of it. They also tend to centralise everything in one place on someone else's terms, which is the opposite of what Indigenous Data Sovereignty requires.

The communities needed something that could take their data in its real forms, make quality evaluation affordable, pull scattered records into one place they controlled, and sit inside a governance framework built around community ownership. Nothing on the market did all four.

What we built

Wathaga is a full software platform, built on a Node.js and React stack on Australian-hosted infrastructure. It lets communities collect data in almost any format, organise it through a Theory of Change approach they call the Wisdom Tapestry, and turn it into reports for their own use and for funders.

AI does real work inside it. It transcribes audio and video across Aboriginal English, language and dialects, summarises large bodies of material, and pulls out the snippets that matter for a report. In one evaluation, its thematic analysis produced results closely matching an independent expert assessment, which is the bar that actually counts.

The decision that defines it

Here is the design choice at the heart of Wathaga: the AI is wrapped at the initiative level, and no data is ever shared with or used to train large language models. Everything stays in Australia, and the communities retain ownership of all of it. Sensitive and sacred material is protected by design, not by a line in a terms-of-service document nobody reads.

This is the difference between bolting AI onto a problem and building responsibly around one. The easy version would have piped everything through a public model and hoped for the best. The right version took more thought, and for this work it was the only version worth building.

Why it's on our site

We include Wathaga here for a simple reason: it is the clearest example of how we think custom AI should be built. Production-grade engineering, AI used where it genuinely helps, and governance and data protection treated as the foundation rather than an afterthought. It is also quiet proof that doing AI properly and doing it responsibly are the same thing, not a trade-off.

Wathaga is live, growing a subscriber base, and being used for everything from junior ranger programs to centralising decades of historical records. The communities own it, and their data, completely.

If you want AI built around your business and your data with the same care, the free AI Maturity Assessment is a good place to start, or talk to us about what you have in mind.