AI Use Case Identification¶

Template Download Template

Purpose: Systematically identify and evaluate potential AI use cases. This template helps you assess whether AI is appropriate for your problem and understand the requirements for successful implementation.

At a Glance

Time to complete: 1-2 hours
Who should complete: Business owner with technical advisor
Key output: Go/no-go decision on pursuing AI approach
Next step: Business Case Template if proceeding

Section 1: Problem Definition¶

1.1 What Problem Are You Trying to Solve?¶

Brief Description (2-3 sentences):

Describe the current problem, challenge, or opportunity

Current State: - How is this problem currently addressed? - What are the pain points or inefficiencies? - What is the impact of not solving this problem?

Desired Future State: - What would success look like? - What specific outcomes are you seeking? - How will you measure improvement?

1.2 Who Is Affected?¶

Primary Users/Beneficiaries: - [ ] APS Staff - [ ] Citizens/Public - [ ] Businesses - [ ] Other government agencies - [ ] Other: _______________

Stakeholder Map: | Stakeholder Group | Role/Interest | Impact Level (High/Med/Low) | |-------------------|---------------|------------------------------| | | | | | | | |

Section 2: AI Suitability Assessment¶

2.1 Is AI Appropriate for This Use Case?¶

Answer these questions (Yes/No/Unsure):

Question	Answer	Notes
Is the problem clearly defined and measurable?
Is there relevant data available (or can it be collected)?
Is the problem pattern-based rather than rules-based?
Would a human be able to perform this task given the same data?
Is the cost/effort of AI justified by the potential benefit?
Are there ethical and responsible ways to use AI for this?
Is there support from leadership and stakeholders?

Interpretation: - Mostly Yes: Good candidate for AI - Mix of Yes/No: May be suitable but requires careful scoping - Mostly No: AI may not be the right solution - consider alternatives

2.2 AI Use Case Type¶

Select the category that best fits (check one):

Classification/Categorization - Assigning labels or categories (e.g., document classification, risk scoring)
Prediction/Forecasting - Predicting future outcomes (e.g., demand forecasting, risk prediction)
Natural Language Processing - Understanding or generating text (e.g., chatbots, summarization, sentiment analysis)
Computer Vision - Analyzing images or video (e.g., object detection, facial recognition)
Recommendation - Suggesting options or next best actions (e.g., content recommendations, case routing)
Anomaly Detection - Identifying unusual patterns (e.g., fraud detection, system monitoring)
Optimization - Finding optimal solutions (e.g., resource allocation, scheduling)
Generation - Creating new content (e.g., text generation, synthetic data, image generation)
Other: _______________

Section 3: Data Assessment¶

3.1 Data Availability¶

What data exists or can be collected?

Data Source	Description	Volume	Quality (High/Med/Low)	Accessibility

Data Characteristics: - Is the data labeled (for supervised learning)? - [ ] Yes, fully labeled - [ ] Partially labeled - [ ] No labels (unsupervised)

Is historical data representative of future scenarios?
Yes
Mostly
No - significant changes expected
Data format:
Structured (databases, spreadsheets)
Unstructured (text, images, audio)
Semi-structured (JSON, XML)
Mixed

3.2 Data Sensitivity¶

What type of data is involved?

Personal information (PII)
Sensitive personal information (health, financial, etc.)
Classified or protected information
Commercially sensitive
Public/non-sensitive

Data Classification (PSPF): - [ ] OFFICIAL - [ ] OFFICIAL: Sensitive - [ ] SECRET - [ ] TOP SECRET

Privacy & Security Requirements: - Privacy Impact Assessment required? [ ] Yes [ ] No [ ] Unsure - Security assessment required? [ ] Yes [ ] No [ ] Unsure - Data sovereignty requirements? [ ] Yes [ ] No [ ] Unsure

Section 4: Technical Feasibility¶

4.1 Solution Approach¶

Preferred approach (select one or more):

Build custom model - Develop and train your own AI model
Use pre-trained model - Leverage existing models and fine-tune
Commercial AI service - Procure cloud AI services (e.g., Azure AI, AWS AI)
Open source tools - Use open source frameworks and models
Hybrid - Combination of above

Rationale:

Why this approach?

4.2 Integration Requirements¶

Where will the AI be deployed?

Integration Points: | System/Platform | Integration Type | Complexity (High/Med/Low) | |-----------------|------------------|----------------------------| | | | | | | | |

4.3 Performance Requirements¶

Response Time: - [ ] Real-time (< 1 second) - [ ] Near real-time (1-10 seconds) - [ ] Batch processing (minutes to hours) - [ ] Offline (days)

Accuracy/Quality Requirements: - Minimum acceptable accuracy: ______% - Is explainability required? [ ] Yes [ ] No - Are errors tolerable? [ ] Yes [ ] No [ ] Depends (explain): __________

Section 5: Risk & Ethics Assessment¶

5.1 Ethical Considerations¶

Assess against Australian Government AI Ethics Principles:

Principle	Assessment	Mitigation Actions
1. Human, societal and environmental wellbeing - AI should benefit individuals, society and the environment	[ ] Low Risk [ ] Med Risk [ ] High Risk
2. Human-centered values - Respect human rights, diversity and autonomy	[ ] Low Risk [ ] Med Risk [ ] High Risk
3. Fairness - Inclusive and accessible; avoiding unfair bias	[ ] Low Risk [ ] Med Risk [ ] High Risk
4. Privacy protection and security - Protect privacy and security	[ ] Low Risk [ ] Med Risk [ ] High Risk
5. Reliability and safety - Operate reliably and safely	[ ] Low Risk [ ] Med Risk [ ] High Risk
6. Transparency and explainability - Clear and responsible disclosure	[ ] Low Risk [ ] Med Risk [ ] High Risk
7. Contestability - Provide mechanisms for challenge and redress	[ ] Low Risk [ ] Med Risk [ ] High Risk
8. Accountability - Clear responsibility and governance	[ ] Low Risk [ ] Med Risk [ ] High Risk

5.2 Key Risks¶

Identify top risks:

Risk	Impact (H/M/L)	Likelihood (H/M/L)	Mitigation Strategy

Common risks to consider: - Bias and discrimination - Privacy breaches - Security vulnerabilities - Model drift or degradation - Over-reliance on automation - Lack of transparency - Regulatory non-compliance

Section 6: Business Case¶

6.1 Benefits¶

Quantifiable Benefits: - Cost savings: $________ per year - Time savings: ________ hours/FTE per year - Productivity improvement: ________% - Error reduction: ________% - Other: _______________

Qualitative Benefits: - [ ] Improved citizen experience - [ ] Better decision-making - [ ] Enhanced staff capability - [ ] Increased service quality - [ ] Regulatory compliance - [ ] Other: _______________

6.2 Costs¶

Initial Costs (One-time): | Item | Estimated Cost | |------|----------------| | Technology/software licenses | $ | | Infrastructure | $ | | Data preparation | $ | | Development/implementation | $ | | Training and change management | $ | | Total Initial Cost | $ |

Ongoing Costs (Annual): | Item | Estimated Cost | |------|----------------| | Software licenses/subscriptions | $ | | Infrastructure/hosting | $ | | Maintenance and support | $ | | Model monitoring and retraining | $ | | Staff time | $ | | Total Annual Cost | $ |

ROI Estimate: - Payback period: ________ months/years - Net benefit (3 years): $________

6.3 Alternatives Considered¶

Have you considered non-AI solutions?

Alternative Approach	Pros	Cons	Why Not Selected?
Business process improvement
Rules-based automation
Human-in-the-loop only
Other: _____________

Section 7: Implementation Readiness¶

7.1 Capability Assessment¶

Current capability (rate 1-5):

Capability Area	Rating	Notes
Data science/AI expertise	/5
Technical infrastructure	/5
Data management	/5
Change management	/5
Governance and oversight	/5

Gaps and Actions:

What capabilities need to be built or acquired?

7.2 Dependencies & Prerequisites¶

What needs to be in place first?

7.3 Timeline Estimate¶

Estimated Phases:

Phase	Duration	Key Activities
Planning & Design
Data Preparation
Model Development
Testing & Validation
Deployment
Total

Section 8: Recommendation & Next Steps¶

8.1 Go/No-Go Recommendation¶

Based on this assessment:

Proceed - Strong candidate for AI; move to detailed planning
Proceed with Caution - Viable but significant risks/gaps to address
Further Investigation - Need more information or proof of concept
Do Not Proceed - AI not appropriate; pursue alternatives

Rationale:

Explain your recommendation

8.2 Immediate Next Steps¶

Priority actions (next 30 days):

8.3 Approvals Required¶

Sign-offs needed:

Role	Name	Approval Date
Business Owner
IT/Technical Lead
Privacy Officer
Security Officer
Executive Sponsor

Appendix: Additional Resources¶

Related Templates: - Privacy Impact Assessment Guide - Security Assessment Checklist - Business Case Template - Risk Register Template

References: - Australian Government AI Ethics Framework - APS Digital Service Standard - Protective Security Policy Framework - Privacy Act 1988

Support: For assistance with this template, contact: [GovSafeAI Team]