Trust is earned through transparency.
AI sits at the centre of how we work – which is exactly why we're open about how we use it, where it falls short, and the standards we hold ourselves to. These are our commitments, in plain language.
Six commitments, in plain language.
- Human oversight, always. Every AI-assisted analysis is reviewed, refined and validated by experienced practitioners. AI supports professional judgement; it does not replace it. That order never reverses.
- Transparent and reviewable. No black boxes. Findings can be checked against the responses they came from, and corrected where the context demands it. If we can't show you where a finding came from, it doesn't go in the report.
- Accountable. The reports we deliver carry our name. When something is wrong, we say so and we fix it.
- Your data stays yours. You own your data and the outputs – always. Your data isn't retained on our servers, we never use client data to train AI, and on request we delete your data and the record of analysis entirely.
- Privacy by default. Personal details – names, addresses, contact information – are removed on the local machine before analysis, so only de-identified text is ever sent onward for processing. You can do this before sending us anything, or we'll do it for you.
- No hype. We describe what our technology can and can't do – honestly. If AI isn't the right tool for your problem, we'll tell you.
The AI processing that supports our analysis currently runs on infrastructure located outside Australia. Because personal details are removed first, only de-identified text is ever transmitted. How we handle your information →
Behind every response is a person.
Consultation data isn't ordinary data. Behind every response is a person – often speaking about decisions that affect their home, their community or their livelihood. Used carelessly, AI can misrepresent what a community said: sentiment misread across cultural and linguistic lines, quieter voices collapsed into an ‘other’ category, the loudest themes mistaken for the most important ones. When that analysis informs a real decision, the error isn't statistical – it's ethical. Recognising those limitations and impacts is where our practice starts.
We're clear-eyed about the limits of the technology we use – including our own.
- AI models carry bias from the data they were trained on.
- Sentiment analysis can misread cultural context, community language and nuance.
- Theme detection can over-represent the loudest voices and lose the quietest ones.
- No algorithm understands the history behind a community's concerns, or what's at stake locally.
These limitations are why our model exists: practitioners design the inputs, and practitioners review every output. The technology is powerful. It is never left unsupervised.
- Identifying themes and sentiment in large volumes of feedback
- Categorising responses
- Removing repetitive analysis work
- Helping track issues over time
- Making decisions
- Producing findings without practitioner review
- Replacing the professional judgement our clients rely on
Our Code of Ethics
Our commitment to ethical practice is written down, published and holds us to account. It has guided how we work since Yuzu was founded.
Download our Code of Ethics (PDF)