AI GOVERNANCE MASTERCLASS 2026

Masterclass Agenda

8:30AM Registration | 9:00AM Start | 5:00PM Close

9:00 AM

What You Actually Owe — and Why AI Governance Fails 

The obligation map for your sector, and the evidence on why AI initiatives fall over.

• The Commonwealth policy: mandatory for non-corporate Commonwealth entities, encouraged for corporate ones, excluding defence and the national intelligence community
• State and territory frameworks, and how local government and universities pick up the standard
• Private sector: the Privacy Act automated decision-making obligation from 10 December 2026, applying to all organisations above $3 million turnover
• Directors’ duties, the privacy statutory tort, consumer law and discrimination exposure
• Why 70 to 90 per cent of AI implementations still fail, and why 77 per cent of the factors are non-technical
  

You leave with: A clear map of what binds you, what does not, and where the real exposure sits.

9:45 AM 

Scoping Clinic

The question organisations are genuinely stuck on. You will work on your own systems.

• What counts as an in-scope use case — and where every framework stops being clear
• The hard cases: general-purpose assistants, AI embedded in vendor products, legacy systems with machine learning inside, routing and triage tools
• Shadow AI — what staff are already using that nobody approved
• Working session: delegates bring real systems and argue them in or out
  

You leave with: A defensible scoping rationale for your own systems, including the arguable ones.

11:00 AM

Risk Classification: The Contested Cases

• Classification criteria across the major frameworks, and where they are genuinely ambiguous
• Consequence and likelihood when harm lands on a citizen, a customer or an employee
• Contested cases worked in groups, chosen because two answers are defensible
• What happens when you classify low and are wrong — and when you classify high and stall the organisation
  

You leave with: Calibrated judgement rather than a lookup table.

1:00 PM

An Impact Assessment, End to End

The assessment is the artefact. It has to survive being read by someone with hindsight.

• Working one assessment from scoping through to sign-off, using the Commonwealth tool as the worked reference
• Documenting reliance, assumptions and dissent — the parts people leave out
• Proportionality: matching depth of assessment to consequence
• Where assessments fall apart when read later
• Peer critique — swap, mark up, and decide whether you would sign it
  

You leave with: One completed impact assessment and a critique checklist you can reuse.

1:00 PM

Human Oversight That Survives Volume 

Everyone writes meaningful human oversight into the assessment. Almost nobody can describe what it looks like at scale.

• What meaningful human oversight actually requires operationally
• Automation bias: why the human in the loop stops looking after the first hundred decisions
• Designing review points, sampling rates and escalation thresholds that hold at volume
• When oversight becomes theatre — and how that reads afterwards

 You leave with: An oversight design for one real use case. 

3:15 PM 

Procurement and Vendor Assurance

Most AI is bought, not built. The governance problem moves into the contract.

• What to ask a vendor, and what their answers actually mean
• Contracting for AI: warranties, model change, data use, retraining and liability
• Assurance when you cannot see inside the model
• Government procurement guidance, and what regulated buyers now expect of suppliers
  

You leave with: A vendor question set, and the contract clauses worth insisting on. 

4:00 PM

Live Incident Simulation 

The morning a system produces the wrong answer about a real person. 

• An unfolding scenario: detection, triage, remedy, reporting
• Incident management, notification duties, and when regulators, media or parliament get involved
• Lessons from Robodebt on escalation, and what the accountable person answers for

 You leave with: An incident response you have rehearsed rather than written. 

4:50 PM

 Close and action commitments