Automation that drafts. People who decide.
Australia has no AI Act. That makes how a system is governed a matter of how it was built, not what it was promised — so we build to a published standard and say plainly where the human sits.
The line we do not cross.
Every automated function we deliver drafts, sorts or surfaces. No function sends, approves or commits. Human decision is retained at each point by design, not by configuration that can be switched off later.
Classification is probabilistic. Nothing is deleted or actioned on classification alone.
Built to the six essential practices.
The National AI Centre's Guidance for AI Adoption sets six essential practices. We build to them.
Decide accountability
A named person owns each automated function, on your side and ours. Recorded in the documentation, not assumed.
Understand impacts
What the function touches, who it affects, and what happens when it is wrong — established before it is built.
Measure and manage risks
Failure modes are written down and mitigated in the design, rather than discovered in production.
Share information
Staff are told what is automated and what is not. Nothing operates in a way people have to guess at.
Test and monitor
Functions are demonstrated on real data before go live, and remain inspectable afterward.
Maintain human control
The decision point stays with a person. This is the constraint the rest of the design bends around.
It is becoming part of winning work.
Where a client is heading toward prequalification or certification, the document control, access control and training records the automation produces form part of the evidence base. The governance is not overhead sitting beside the system — it is an output of it.
Principals and insurers are beginning to write AI clauses into procurement. Being able to answer them, in writing, is becoming a condition of being considered.
Ask how we would govern your systems.
The answer is part of the written diagnostic, before anything is built.