AI Policy Implementation

A practical walkthrough of the Commonwealth mandate and what it actually requires of your agency.

• The obligations under the Commonwealth Policy for the Responsible Use of AI, and the June and December 2026 timeframes
• The concrete steps to compliance, and the key points every internal AI policy needs
• How AI obligations fit within your existing legislative, policy and operational landscape (end to end, not bolted on)
• Why letting a tool like Copilot generate a tick-the-box policy fails the intent
• One clear principle: the agency owns the risk

Move beyond principles to a policy that operational teams can pick up and run with.

• Writing for the operational reader: plain language, do/don't, decision trees, embedded examples
• Positioning AI policy as an enabler of responsible uptake and productivity, not a governance-only instrument
• Elevating ownership to executive/CEO level rather than parking it in Data and Information
• One-page summaries, quick-reference cards, and role-based views staff can actually act on
• Knowing which central teams to call on for help

The heart of Day 1. The practical mechanisms that turn a published policy into daily practice.

• Executive sponsorship and visible modelling from the top
• Embedding policy into existing gates: procurement, project intake, privacy and security review, PIA/AIA, records
• Making the compliant path the easy path: approved-tool catalogues, sensible defaults, guardrails inside the tools themselves
• Champions networks and communities of practice
• Comms cadence: launch, reinforce, refresh (not one and done)
• Training and AI literacy tied to roles, not blanket modules
• Attestations, onboarding hooks, and recognition tied to responsible use
• Feedback loops so the policy improves from real use

Participants apply the morning sessions to:

• Pick three levers most relevant to their agency
• Map owners, dependencies and sequencing
• Identify the first artefact they would ship in the next 30 days
• Pressure-test the map with peers

Build an uptake measurement approach your executive and ARC will actually read.

• Coverage metrics: percentage of AI use logged in the register, percentage of teams with a nominated AI lead
• Capability metrics: literacy completion by role, assessor certification, sponsor readiness
• Behaviour metrics: attestation rates, approved-tool usage versus shadow AI signals (personal accounts, Copilot Chat, unapproved SaaS)
• Flow metrics: time to triage, time to approval, rework rate
• Assurance metrics: audit findings closed, incidents and near-misses reported
• Sentiment metrics: pulse checks on staff confidence, clarity and friction
• Designing an uptake dashboard for the exec and ARC

Prioritise AI opportunities using a practical and defensible triage approach.

• Clarifying business purpose, relevance and expected benefits, and who receives the AI-generated outputs
• Testing whether the real issue is people, process, technology or data
• Screening for data quality, bias and false positives
• A simple risk-tiering framework with low, medium and high examples
• Approval pathways and clear escalation points
• How triage decisions feed back into the register and Day 1 uptake metrics

Manage the risks and dependencies that commonly derail AI implementation.

• Privacy, cyber, data governance, records, human oversight, explainability and contestability in practice
• Assessing AI tools and embedded systems: guardrails for systems and guidelines for staff
• Managing shadow AI (personal accounts, Copilot Chat, unapproved SaaS)
• Supply chain, sovereignty (not a binary), T&Cs and vendor/MSP/JV arrangements
• Model dependency and vendor concentration: continuity if a vendor changes terms, pricing or access
• Substitution and exit planning: open-weight, closed and locally hosted options
• Testing vendor claims, and stopping pilots from becoming uncontrolled shadow systems

Build a compelling business case for AI investment and public value.

• Measuring ROI where cost to serve is unclear
• Building the financial value proposition and the evidence of value that nobody can show
• Identifying high-value areas: agents, automation of repetitive tasks, accuracy and error-checking, faster query resolution
• Linking efficiency gains to better community outcomes
• Linking value evidence to Day 1 uptake metrics: uptake without value is activity theatre, value without uptake is a one-off win
• Making the case work for resource- and budget-constrained teams doing AI in house

Develop a practical roadmap for safe, staged AI implementation.

• Sequencing the levers, not just the tech: what to embed in the first 90 days versus the next 6 months
• Starting small with pilots, proof points and iterative delivery to build confidence
• Change management and culture
• Strategic workforce planning, transition of capabilities and human in the loop
• Seeing what good looks like and where to source trusted guidance
• Learning from what other agencies are doing through a cross-cohort community of practice
• Building AI capability through literacy, training and responsible adoption across the workforce

Interactive workshop: Your agency AI implementation roadmap

Participants apply the day's learning to:

• Prioritise AI initiatives
• Develop a one-page implementation plan covering three levers, three uptake metrics and three use cases
• Prepare an executive briefing
• Discuss implementation considerations and next steps