Skip to content

Deploy GenAI Tools for Your Team

You want to enable your team to use GenAI (like ChatGPT, Copilot, or Claude) safely and effectively. This journey helps you roll out GenAI tools responsibly.

Step 1
Assess
Step 2
Select
Step 3
Safeguard
Step 4
Enable
Step 5
Govern

Step 1: Assess the Opportunity

Understand why GenAI and what use cases make sense.

Use Case Identification

Good first use cases for GenAI:

Category Examples Risk Level
Drafting Emails, reports, briefs Low-Medium
Research Summarising documents, finding information Low-Medium
Coding Code generation, debugging, documentation Medium
Analysis Data exploration, pattern identification Medium
Ideation Brainstorming, option generation Low

Proceed with caution:

Category Concerns
Decision-making Accountability, bias, explainability
Citizen-facing Quality control, accessibility, trust
Legal/compliance Accuracy requirements, liability
Sensitive data Privacy, security, data handling

Readiness Check

  • Leadership support exists
  • Budget identified
  • IT/security engaged
  • Policy position understood (agency and whole-of-government)
  • User demand validated

Reference

Check what's already approved:

  • GovSafeAI Platform - Whole-of-APS option
  • Agency-specific approved tools
  • Existing procurement arrangements

Step 2: Select the Right Tool

Choose based on your needs, constraints, and risk appetite.

Decision Factors

Factor Questions
Use case fit Does it do what your team needs?
Security Where does data go? Who can access?
Privacy Is data used for training? Retention?
Cost Per user? Per token? Predictable?
Support Vendor support? Community?
Integration Works with existing tools?
Compliance Meets government requirements?

Tool Options

Option Pros Cons
GovSafeAI (Finance) Approved, secure, supported May have limitations
Enterprise tools (Copilot, etc.) Integration, features Cost, vendor lock-in
API integration Control, customisation Development needed
Consumer tools Free/cheap, easy Security, compliance concerns

Vendor Evaluation

Tool: Model Evaluation Calculator

Read: Questions Vendors Won't Answer

Key questions for GenAI vendors:

  • Where is data processed and stored?
  • Is data used to train models?
  • What content filtering exists?
  • What's the SLA and support model?
  • How are model updates handled?
  • What happens if the service is discontinued?

Step 3: Put Safeguards in Place

Manage the risks before enabling access.

Acceptable Use Policy

Define clearly:

Area Policy Decision
What's allowed Approved use cases, tool list
What's forbidden Sensitive data, certain decisions
Human oversight When human review is required
Quality checks How to verify outputs
Incident reporting How to report problems

Data Handling Rules

Critical decisions

  • What data can be input to GenAI?
  • What classification levels are permitted?
  • Can PII be used? Under what conditions?
  • What about intellectual property?

Use PII Masking Tool to identify sensitive data.

Technical Controls

Control Purpose
Access management Control who can use tools
Logging Track usage for audit/review
Content filtering Block inappropriate outputs
Data loss prevention Prevent sensitive data input
Cost controls Prevent budget blowouts

Step 4: Enable Your Team

Roll out with training and support.

Training Program

Essential topics:

  1. What GenAI is (and isn't)
  2. Capabilities and limitations
  3. How it works at high level

  4. Why it makes mistakes

  5. How to use it effectively

  6. Prompting techniques
  7. Iterative refinement

  8. When to use vs not use

  9. How to use it safely

  10. Data handling requirements
  11. Quality verification

  12. Policy compliance

  13. What to do when things go wrong

  14. Error recognition
  15. Incident reporting
  16. Escalation paths

Rollout Strategy

Approach When to Use
Pilot group High uncertainty, need to learn first
Phased by team Different needs, want controlled growth
Use-case gated Start with low-risk, expand over time
Full rollout High confidence, good controls

Support Model

  • Designated GenAI champions
  • Clear help channels
  • FAQ documentation
  • Regular tips and updates
  • Feedback mechanisms

Step 5: Govern and Improve

Keep it working well over time.

Ongoing Governance

Activity Frequency Purpose
Usage review Monthly Understand adoption, patterns
Policy review Quarterly Update based on experience
Cost review Monthly Budget management
Incident review As needed Learn from problems
User feedback Ongoing Continuous improvement

Metrics to Track

Metric What It Tells You
Adoption rate Is the team using it?
Use case distribution What's it being used for?
User satisfaction Is it helping?
Quality issues Are there problems?
Cost per user Is it efficient?
Incident rate Are safeguards working?

Scaling Considerations

As usage grows:

  • Expand use cases carefully
  • Review and update policies
  • Invest in advanced training
  • Consider customisation/fine-tuning
  • Build internal expertise
  • Share learnings across organisation

Common Pitfalls

Watch out for

  • Rushing without safeguards - Pressure to keep up, skip controls
  • Over-restricting - So locked down it's useless
  • No training - Expecting people to figure it out
  • Ignoring shadow IT - People already using unapproved tools
  • Set and forget - Not monitoring, reviewing, improving
  • Underestimating costs - Token costs can surprise you