For managers reviewing and sponsoring an AI project proposal

Does the proposal solve a real, well-defined service or staff problem Is the problem appropriate for AI, or are there simpler alternatives?

Has the project been classified as low, medium or high risk?
Are the governance requirements proportionate (light-touch for low-risk, robust for high-stakes)?
Have I assessed risk against the Australia's AI Ethics Principles and the national AI assurance framework? For example:
- Fairness – could outputs disadvantage certain groups?
- Privacy and security – is data protected and appropriately de-identified?
- Transparency – can the system’s role be explained clearly to users and the public?
- Accountability – is there a clear chain of responsibility?

Have I allocated enough time, budget, and skills for the team to succeed?
Am I providing psychological safety, making it clear staff won’t be penalised for safe experimentation?

Will there be human-in-the-loop oversight where needed?
Are privacy, ethics and risk reviews embedded from the start (not bolted on later)?
Is there clear accountability (an AI steward or project owner)?
Does the proposal include a transparency statement, in line with DTA requirements (to be published externally if the pilot scales)?

Am I modelling curiosity and willing to learn alongside my team?
Am I acting as a sponsor, not a gatekeeper, giving permission and legitimacy to proceed?
Am I communicating this as a safe, strategic experiment rather than a risky gamble?

AI in government: Practical checklists to set projects up for success

Related content