In August, we surveyed 50 CFOs from Australian companies with revenue between $100M and $2B. We wanted to understand the financial reality of AI investment – not the vendor hype, not the media narrative, but actual spending patterns and attitudes.

The results were illuminating.

The Headline Numbers

Average AI budget for 2024-25: $1.2M (median: $800K)

Percentage of IT budget allocated to AI: 8.4% (up from 4.1% in 2022-23)

CFOs who believe AI spending will increase next year: 74%

CFOs who are satisfied with AI ROI to date: 31%

That last number deserves attention. Almost 70% of CFOs are not satisfied with the returns on their AI investments so far. Yet 74% expect to spend more. We’ll get into why.

Where the Money Goes

We asked CFOs to break down their AI spending across categories:

Category	% of AI Budget
Productivity tools (Copilot, etc.)	34%
Custom AI/ML development	28%
Data infrastructure	18%
Consulting and implementation	12%
Training and change management	8%

The dominance of productivity tools isn’t surprising. It’s the path of least resistance – buy a license, roll it out, hope for efficiency gains. Whether that strategy delivers is another question.

The underspend on training and change management (8%) is consistent with every AI project failure pattern I’ve seen. Technology without adoption is just expensive shelfware.

The Top Three Concerns

We asked CFOs to rank their concerns about AI investment:

1. Difficulty measuring ROI (67%): The most common concern by far. CFOs are approving budgets without clear metrics for success. One CFO told us, “I can measure the cost precisely. Measuring the benefit? Complete guesswork.”

2. Integration complexity (54%): AI doesn’t exist in isolation. Getting it to work with existing systems is consistently underestimated. Many CFOs reported that integration costs exceeded original estimates by 40-100%.

3. Data quality and availability (49%): The eternal problem. AI is only as good as the data it trains on, and most organisations know their data isn’t good enough.

Interestingly, concerns about AI ethics and job displacement ranked lower (23% and 18% respectively). At the CFO level, it’s execution challenges, not existential questions, that keep people up at night.

The Satisfaction Gap

Why are 69% of CFOs unsatisfied with AI ROI but 74% planning to increase spending?

The interviews gave us context:

Fear of falling behind: “We’re not satisfied with results yet, but our competitors are investing heavily. Not investing isn’t an option.” This sentiment appeared repeatedly. As Harvard Business Review has noted, FOMO is a real driver of enterprise technology investment.

Long-term positioning: Several CFOs framed current spending as capability building rather than immediate returns. “We’re building the foundation. The ROI will come in years 2 and 3.”

Sunk cost effect: “We’ve invested too much to stop now.” Not the most rational approach, but an honest one.

Bright spots: Even among the dissatisfied, most could point to at least one use case that was working. Success in one area justified continued investment in others.

What the Satisfied CFOs Did Differently

The 31% who were satisfied shared common characteristics:

Started with a clear business problem. Not “we need AI” but “we need to reduce invoice processing costs by 40%.” The AI was incidental to the business outcome.

Invested heavily in data preparation. Satisfied companies spent an average of 35% of their AI budget on data work. Unsatisfied companies spent 19%.

Had realistic timelines. Satisfied CFOs expected ROI in 18-24 months. Unsatisfied CFOs originally expected 6-12 months.

Owned the AI strategy at business level. In satisfied companies, business units drove AI priorities. In unsatisfied companies, IT or a central innovation team typically led.

Industry Breakdown

AI spending varied significantly by industry:

Industry	Average AI Budget	Primary Use Case
Financial Services	$2.1M	Risk/fraud detection
Healthcare	$1.4M	Clinical documentation
Manufacturing	$1.1M	Predictive maintenance
Retail	$900K	Demand forecasting
Professional Services	$750K	Productivity tools

Financial services leads in absolute spending, driven by regulatory use cases where AI has clear applications and measurable outcomes. Professional services, despite being knowledge-work heavy, lags – largely because use cases are harder to define and measure.

The Vendor Landscape

We asked which vendors CFOs were working with:

• Microsoft (Azure AI/Copilot): 78%
• Google Cloud AI: 34%
• AWS AI services: 31%
• Salesforce Einstein: 24%
• Local Australian partners: 41%

The Microsoft dominance reflects enterprise reality – most large Australian organisations are already in the Microsoft ecosystem. Google and AWS are more common in companies with existing cloud relationships on those platforms.

The 41% working with local partners is notable. Several CFOs mentioned preferring Australian firms for data sovereignty, time zone alignment, and understanding of local regulations.

What This Means

A few implications from the data:

The shakeout is coming. When 69% of CFOs are unsatisfied with ROI, some of that spending will eventually be cut. Vendors and internal AI teams that can’t demonstrate value are vulnerable.

Data is the bottleneck. Consistently, data preparation emerged as the difference between success and failure. Organisations planning AI initiatives should budget accordingly.

Measurement frameworks matter. CFOs are approving budgets they can’t measure. This isn’t sustainable. Developing clear AI ROI metrics should be a priority.

The winners are focused. Satisfied companies didn’t try to do AI everywhere. They picked specific problems, solved them well, then expanded.

Final Thought

Australian enterprises are spending real money on AI. Whether that investment pays off depends less on the technology chosen and more on the fundamentals: clear problems, good data, realistic expectations, and proper measurement.

The CFOs who get this right will look prescient in two years. The ones who don’t will be explaining budget write-offs to their boards.