AI Business Case Template - How to Justify AI Investment
You know AI could help your business. Your team knows it. But the CFO wants a business case, and "everyone else is doing it" isn't going to cut it.
We've helped dozens of Australian organisations build AI business cases that actually get approved. Not because we inflate the numbers, but because we structure them in a way that addresses what decision-makers actually care about.
Here's the exact template we use, along with guidance on how to fill in each section.
Why Most AI Business Cases Get Rejected
Before we get to the template, let's talk about why AI proposals fail.
Reason 1 - Too vague. "We should implement AI to improve efficiency." That's not a business case. That's a wish. Decision-makers need specific processes, specific costs, and specific returns.
Reason 2 - No baseline. If you can't clearly describe how things work today and what they cost, you can't convincingly argue for a change.
Reason 3 - All upside, no risk. Every executive has seen technology projects go sideways. A business case with no risks section looks naive. Including honest risks and mitigations actually builds trust.
Reason 4 - Wrong time horizon. AI projects have upfront costs and ongoing returns. If you only show year one, the numbers might look marginal. Show three years and the compound returns become clear.
Reason 5 - No comparison. What happens if you don't invest? The status quo has a cost too, growing labour costs, competitive disadvantage, mounting technical debt.
The AI Business Case Template
Section 1 - Executive Summary (1 page)
Keep this to one page. If the executive summary doesn't sell it, the rest won't get read.
Include:
- The business problem in one sentence
- The proposed AI solution in 2-3 sentences
- Total investment required (AUD)
- Expected annual return (AUD)
- Payback period
- Key risks and mitigations (top 3)
Example:
Our customer service team processes 3,200 enquiries per week at a cost of $9.50 per interaction. An AI agent can handle 55-65% of Tier 1 enquiries at $0.80 per interaction, reducing annual customer service costs by $520,000 while improving response times from 4 hours to under 2 minutes. Total investment is $180,000 in year one with $85,000 annual operating costs. Payback period is 4.2 months.
That's the kind of summary that gets attention.
Section 2 - Current State Analysis
Document exactly how things work today. Be specific and use real numbers.
Process description:
- What is the process being targeted?
- Who performs it? How many people?
- What tools do they use?
- How long does it take per unit of work?
- What's the current volume?
Current costs (annual):
| Cost Item | Annual Cost (AUD) |
|---|---|
| Staff (fully loaded) | $X |
| Tools and software | $X |
| Error/rework costs | $X |
| Opportunity costs | $X |
| Total current cost | $X |
Current performance metrics:
| Metric | Current Value |
|---|---|
| Processing time per unit | X minutes |
| Error/rework rate | X% |
| Customer satisfaction | X |
| Throughput per day | X |
| Staff utilisation | X% |
In our experience, the current state analysis is where most teams under-invest. They rush to describe the shiny future and skip documenting the painful present. But this section is what makes the financial case credible.
Section 3 - Proposed Solution
Describe what you're proposing to build, without getting lost in technical details.
Solution overview:
- What will the AI system do?
- What technology approach will be used? (e.g., large language model, machine learning, computer vision)
- How will it integrate with existing systems?
- What will the user experience look like?
Scope definition:
- What's in scope for phase 1?
- What's explicitly out of scope?
- What are the dependencies?
Keep it accessible. The audience for a business case is typically executives, not engineers. "An AI agent powered by a large language model will process customer enquiries through our existing ticketing system" is better than three paragraphs about model architecture.
Section 4 - Financial Model
This is the heart of the business case. We recommend a 3-year model.
Investment costs:
| Cost Item | Year 1 | Year 2 | Year 3 |
|---|---|---|---|
| Discovery and design | $X | - | - |
| Development and testing | $X | - | - |
| Integration | $X | - | - |
| Data preparation | $X | $X | - |
| Training and change management | $X | $X | $X |
| Infrastructure | $X | $X | $X |
| AI model/API costs | $X | $X | $X |
| Monitoring and maintenance | - | $X | $X |
| Total investment | $X | $X | $X |
Returns:
| Return Item | Year 1 | Year 2 | Year 3 |
|---|---|---|---|
| Labour cost reduction | $X | $X | $X |
| Error/rework reduction | $X | $X | $X |
| Throughput improvement value | $X | $X | $X |
| Revenue impact | $X | $X | $X |
| Risk/compliance value | $X | $X | $X |
| Total returns | $X | $X | $X |
Summary financials:
| Metric | Value |
|---|---|
| Total 3-year investment | $X |
| Total 3-year returns | $X |
| Net present value (NPV) | $X |
| Internal rate of return (IRR) | X% |
| Payback period | X months |
| Year 1 ROI | X% |
Pro tip: Use conservative, moderate, and aggressive scenarios. Present the moderate case as your primary recommendation but show the range. This builds credibility and gives decision-makers confidence that even the conservative case is worthwhile.
Section 5 - Risk Assessment
Be honest here. A business case with no risks listed isn't trustworthy.
Risk matrix:
| Risk | Likelihood | Impact | Mitigation |
|---|---|---|---|
| AI accuracy below expectations | Medium | High | Pilot phase with defined accuracy thresholds before full rollout |
| Integration complexity higher than estimated | Medium | Medium | Technical spike in discovery phase; contingency budget of 20% |
| Low user adoption | Medium | High | Change management plan; early stakeholder involvement |
| Data quality insufficient | Low | High | Data audit before development begins |
| Ongoing costs higher than projected | Low | Medium | Usage monitoring; cost caps on API calls |
| Regulatory changes | Low | Medium | Modular architecture allowing quick adjustments |
Section 6 - Implementation Plan
Show you've thought about how this gets done, not just what gets done.
Recommended phases:
Phase 1 - Discovery (4-6 weeks, $20,000-$40,000)
- Detailed process mapping
- Data audit and readiness assessment
- Technical feasibility validation
- Refined business case with real data
Phase 2 - Pilot (6-10 weeks, $60,000-$120,000)
- Build MVP targeting highest-value use case
- Test with subset of users/volume
- Measure against defined success metrics
- Go/no-go decision based on pilot results
Phase 3 - Production (6-12 weeks, $40,000-$100,000)
- Scale to full volume
- Complete integration with existing systems
- Training and change management
- Monitoring and optimisation setup
Phase 4 - Optimise and Expand (Ongoing, $60,000-$100,000/year)
- Performance monitoring and tuning
- Additional use cases based on pilot learnings
- Model updates and improvements
The phased approach is important because it builds in decision gates. Nobody's approving $300,000 on faith. They're approving $30,000 for discovery, with a clear decision point before committing further.
Section 7 - Success Metrics and Governance
Define clear success metrics for each phase:
| Phase | Metric | Target |
|---|---|---|
| Pilot | Accuracy rate | >90% |
| Pilot | Processing time reduction | >60% |
| Pilot | User satisfaction | >4/5 |
| Production | Cost per transaction reduction | >50% |
| Production | Monthly cost savings | >$X |
| Production | Error rate reduction | >70% |
Governance:
- Who owns the AI system after deployment?
- Who monitors performance?
- What triggers a review or rollback?
- How often will ROI be reassessed?
Filling In the Numbers - Where to Get Data
The template is only as good as the numbers you put in it. Here's where to find them:
For current state costs:
- HR for fully loaded salary costs (remember to include super, leave, overhead - typically 1.3-1.5x base salary in Australia)
- Finance for tool and software costs
- Operations for volume and throughput data
- Quality team for error and rework rates
For AI implementation costs:
- Get quotes from 2-3 AI consulting firms (we're happy to provide a scoping estimate)
- Include 15-20% contingency for unknowns
- Budget for ongoing costs at 15-25% of build cost annually
For expected returns:
- Use industry benchmarks as a starting point (we publish these regularly)
- Apply a 30% discount to vendor claims
- Use your pilot results to validate before scaling
Common AI Business Case Scenarios in Australia
To give you a sense of scale, here are business cases we've seen approved recently:
Scenario 1 - Document processing for a financial services firm
- Investment: $150,000 (build) + $60,000/year (run)
- Annual savings: $320,000
- Payback: 5.6 months
- Approved: Yes, funded from operational budget
Scenario 2 - Customer service AI for a mid-market retailer
- Investment: $200,000 (build) + $85,000/year (run)
- Annual savings: $480,000
- Payback: 5 months
- Approved: Yes, phased rollout starting with discovery
Scenario 3 - Internal knowledge assistant for a professional services firm
- Investment: $90,000 (build) + $40,000/year (run)
- Annual value: $180,000 (productivity gains)
- Payback: 6 months
- Approved: Yes, but started with 3-month pilot
The Conversation With Your CFO
Here's what CFOs actually care about, based on every business case conversation we've sat in:
- What's the payback period? Under 12 months is strong. Under 6 months is exceptional.
- What happens if it doesn't work? Having a phased approach with decision gates answers this.
- What are the ongoing costs? Don't hide these. Transparency builds trust.
- Can we start smaller? Always say yes. A $30,000 discovery phase is easier to approve than a $200,000 project.
- What's the cost of doing nothing? This is your secret weapon. Rising labour costs, competitive pressure, and growing process volumes mean the status quo gets more expensive every year.
Get Help Building Your Business Case
If you're putting together an AI business case and want expert input, our AI strategy consulting team can help. We run AI readiness assessments that produce the data you need for a credible business case, including process analysis, cost modelling, and ROI projections.
You can explore our full AI consulting services or our AI automation capabilities to understand what's possible for your organisation.
Ready to start the conversation? Get in touch and we'll help you build a business case that gets approved - because it's built on real numbers, not hype.