Capability needs accountability.
LostMind AI builds systems that make context, sources, uncertainty and human control visible. The practice follows Australia's AI Ethics Principles at the level of outcomes.
Context with boundaries
Shared and personal information should stay in the correct room. Products are built around separated spaces, and private context is not pulled into a shared room without permission.
Grounded, checkable output
Where an answer draws on sources, those sources should be available to inspect. Extraction and conversion should flag low-confidence figures instead of presenting a guess as fact.
Review stays visible
High-impact numbers, documents and actions need an appropriate check. DoneSmith work is reviewed by agents with automated checks, including LostMind's automated quality layer. Automation should remove repetitive work without hiding uncertainty.
People stay in control
The products are there to support a person's judgment, including in property and finance work. We do not treat an automated draft as the decision.
Data minimisation
We aim to collect only what a workflow needs, we avoid advertising-driven profiling, and there is a clear contact path for privacy requests.
Continuous improvement
Responsible AI is an engineering and operational practice, not a badge. We review failure modes and what people tell us as the products develop.