We’ve spent the past six weeks surveying technology leaders across 140 Australian enterprises to understand where AI adoption actually stands. Not where vendor marketing says it stands. Not where LinkedIn posts suggest. Where the real money is being spent, what’s actually in production, and what’s stalled.

The picture is more mixed than either the optimists or sceptics would have you believe.

The Headlines

78% of surveyed enterprises have at least one AI initiative in progress. That’s up from 62% in our September 2025 survey. However, “in progress” covers everything from a funded proof of concept to a fully deployed production system. The devil is in the detail.

Only 23% have AI systems running in production that directly affect business operations. The rest are in various stages of exploration, piloting, or waiting for business case approval to proceed beyond proof of concept.

Average AI spending per enterprise is $1.2 million annually for organisations with revenue between $100 million and $1 billion. This includes vendor licenses, consulting, internal headcount, and infrastructure. It does not include general cloud computing costs that may partially support AI workloads.

The most common use cases in production are customer service automation (chatbots and email triage), document processing and data extraction, and demand forecasting. These aren’t flashy, but they’re generating measurable returns.

The Adoption Curve

Australian enterprises broadly fall into four categories.

Leaders (roughly 12% of respondents): These organisations have multiple AI systems in production, dedicated AI teams, established governance frameworks, and are actively scaling. They’re past the experimentation phase and treating AI as core operational infrastructure. Most are in financial services, mining, or telecommunications — industries with data density and clear ROI pathways.

Fast followers (roughly 25%): They’ve run successful pilots and are working to move one or two use cases into production. The primary blockers are data quality, integration with legacy systems, and organisational readiness rather than technical capability. They have budget and executive support but are navigating implementation complexity.

Experimenters (roughly 40%): They have AI on the strategic agenda and have funded some exploration, but haven’t committed to production deployment. Many have run proof of concepts that produced promising results but haven’t secured the organisational commitment (budget, process changes, talent) required for production.

Observers (roughly 23%): They’re watching the market but haven’t made meaningful AI investments. Some are constrained by budget, some by talent, some by uncertainty about where AI applies to their business. A few are sceptical of the business case and waiting for clearer evidence before committing.

Where Money Is Being Spent

The spending breakdown across our survey tells an interesting story:

• Vendor software licenses (35%): Microsoft Copilot, Salesforce Einstein, and industry-specific AI tools dominate. Platform-level AI features embedded in existing enterprise software account for most AI spending, even though many organisations don’t classify this as “AI investment.”
• Consulting and implementation (28%): External expertise for strategy development, use case identification, implementation, and integration. A Sydney-based firm specialising in practical AI consulting noted that demand for implementation services has grown three times faster than demand for strategy work — organisations know what they want to do, they need help doing it.
• Internal headcount (22%): Data engineers, ML engineers, and AI product managers. Hiring remains difficult, with median time-to-fill for senior AI roles at 4.5 months in the Australian market.
• Infrastructure (15%): GPU compute, data platforms, and AI-specific infrastructure. Organisations running models locally rather than using API-based services have higher infrastructure costs but lower per-inference costs at scale.

The Pilot-to-Production Gap

The most significant finding is the persistent gap between pilot success and production deployment. Of organisations that completed AI pilots in 2025, only 34% progressed to production deployment. The remaining 66% either abandoned the initiative, are still evaluating, or are stuck in implementation.

The primary barriers to progression aren’t technical. They’re organisational:

Data readiness (cited by 67% of stalled projects): The pilot worked on clean sample data. Production requires integration with messy, distributed, inconsistent enterprise data. The data preparation work required to bridge this gap exceeds the original project budget.

Change management (cited by 52%): AI systems change how people work. Process changes, role modifications, and trust-building with end users take longer than anyone budgets for. Technically successful AI systems get abandoned because users don’t adopt them.

Unclear ROI measurement (cited by 44%): The pilot demonstrated capability, but translating that to measurable business value is harder than expected. When the CFO asks “what’s the return on this investment?” and the answer is vague, funding decisions get deferred.

Governance and risk concerns (cited by 38%): Particularly for AI systems that make or influence decisions affecting customers, legal and compliance teams want governance frameworks that many organisations haven’t built yet.

Industry Variations

Financial services leads adoption by a significant margin. APRA’s operational resilience standards have paradoxically accelerated AI adoption in banking by creating clear governance requirements that give risk-averse organisations a framework to work within.

Mining and resources is the surprise performer. Predictive maintenance, autonomous operations, and geological analysis use cases are well suited to AI, and the sector’s comfort with capital-intensive technology investments translates well to AI programs.

Healthcare adoption is slower than expected despite the obvious applications. Privacy requirements, clinical validation needs, and the sector’s inherently conservative approach to technology change are creating longer adoption timelines.

Retail is a mixed picture. Large retailers are investing heavily in demand forecasting and personalisation. Mid-market retailers are struggling to justify AI investment against tight margins and competing priorities.

What to Watch

Three trends will shape the next 12 months of Australian enterprise AI adoption.

First, the embedded AI wave. As Microsoft, Google, and Salesforce embed AI capabilities into their core platforms, adoption will accelerate without organisations making explicit “AI investment” decisions. This will blur the line between AI adoption and software upgrades.

Second, regulation. The Australian government’s AI governance framework, expected to move from voluntary to mandatory for certain applications in late 2026, will force organisations to either formalise their AI practices or pause deployments until they can comply.

Third, talent. The Australian AI skills shortage isn’t resolving. Organisations that can’t hire are turning to external partners, but the consulting market is capacity-constrained too. Talent availability, not technology capability or budget, may be the binding constraint on adoption growth.

The overall picture is one of steady but uneven progress. AI is moving from strategic aspiration to operational reality for Australian enterprises, but the pace varies enormously by industry, company size, and organisational readiness. The gap between leaders and laggards is widening, and closing it is becoming harder as leaders accumulate data assets, institutional knowledge, and competitive advantages that compounding makes increasingly difficult to replicate.