← Back to Insights

Financial Services

Why Most Banks Are Still Stuck in AI Pilot Purgatory

Financial Services

Why Most Australian Banks Are Still Stuck in AI Pilot Purgatory — And the 12% That Aren't

The gap between AI experimentation and AI value in Australian financial services is not narrowing — it is widening. Here's what separates the leaders from the laggards.

Bosley Insights 12 min read February 2026
B
Bosley | AI Strategy & Implementation
We design and build AI-native operating models for Australian organisations. Tier 1 consulting rigour, hands-on build capability.

Here is an uncomfortable truth for Australian financial services leaders: only 12% of CEOs globally report achieving both cost savings and revenue gains from AI, according to the 2025 PwC CEO Survey of approximately 4,500 chief executives. Meanwhile, 56% report no significant financial benefit yet. The gap between AI experimentation and AI value is not narrowing — it is widening.

In Australian banking, insurance, and wealth management, the challenge is compounded by regulatory complexity that has no parallel in most other markets. APRA CPS 230, CPG 235, ASIC responsible lending obligations, AML/CTF requirements — every AI deployment must satisfy multiple overlapping frameworks. The post-Royal Commission environment means boards demand governance before they'll approve investment.

Yet the institutions pulling ahead are not doing so despite regulation — they are using governance as an accelerator.

The Vanguard: What the 12% Are Doing Differently

The PwC research reveals that the top 12% of CEO organisations are 2.6 times more likely to have embedded AI into core processes, with 44% deploying AI at large scale compared to just 17% of others. They are three times more likely to report meaningful financial returns when proper AI foundations exist.

Foundations matter as much as scale. CEOs whose organisations have established strong AI foundations — responsible AI frameworks and technology environments that enable enterprise-wide integration — are three times more likely to report meaningful financial returns.

In Australian financial services specifically, this means the leaders have done three things their peers have not:

What Separates the Leaders
Governance First
Built AI governance frameworks aligned with APRA/ASIC expectations before scaling — not after
Integrated MRM
Integrated AI model risk management into existing three-lines-of-defence frameworks rather than creating parallel structures
Smart Use Cases
Chose use cases where regulatory alignment was achievable, rather than chasing the most technically impressive opportunities

The Five Barriers Keeping Australian Financial Services in Pilot Mode

1 Regulatory Complexity Without a Governance Blueprint

APRA is increasingly asking institutions about AI governance. CPS 230 now requires operational resilience for AI systems in critical processes. CPG 235 demands model risk management frameworks that accommodate AI's different characteristics — including opacity, drift, and data dependency. Without a coherent framework that satisfies all regulators simultaneously, institutions default to caution.

2 Legacy Systems That Predate the Internet

Core banking systems in Australia's Big Four are often decades old. AI needs real-time data access, integration flexibility, and clean data — precisely what legacy environments struggle to provide. The institutions moving fastest have created API and integration layers that allow AI to work alongside core systems rather than requiring replacement.

3 The Talent War Australia Cannot Win Conventionally

Competition for AI/ML talent is fierce, with financial services competing against technology companies and consultancies. Institutions making progress combine external partnerships for specialised capability with internal upskilling — building sustainable capability rather than trying to out-bid Google for data scientists.

4 Fintech Competition Raising Customer Expectations

Neobanks, payment fintechs, and global technology players are setting customer experience standards that traditional institutions struggle to match. AI-native products from competitors create an expectation gap that grows wider every quarter.

5 Data Fragmentation from Decades of M&A

Multiple acquisitions have created siloed data environments with inconsistent definitions and quality issues. Without a unified data foundation, AI models are trained on incomplete or inconsistent information.

Where the ROI Actually Sits

The AI Daily Brief's benchmarking study of 1,200+ respondents and 5,000+ use cases reveals that 82% of AI deployments deliver positive ROI, with 37% delivering high or transformational returns. But the distribution matters enormously for financial services.

25%
of risk reduction AI use cases deliver transformational ROI — despite comprising only 3.4% of total deployments. For financial services — where compliance, fraud detection, and risk management involve massive volumes — this is precisely where AI excels.

Fraud pattern recognition, AML/CTF detection improvement, false positive reduction, and regulatory reporting automation represent the highest-value, lowest-risk entry points for Australian financial institutions.

The AI Tax: Why Quick Wins Become Expensive

A critical finding: 37% of time saved through AI is offset by rework. For every 10 hours of efficiency gained, nearly 4 hours are lost to fixing AI output. In financial services, where errors carry regulatory and reputational consequences, this "AI tax" is significant. This is not a reason to avoid AI — it is a reason to invest in proper implementation with training, workflow integration, and quality assurance.

A Practical Framework: From Pilot to Production

Four-Phase Approach for Regulated Environments
Phase 1 (4–6 wks)
Governance-First Strategy. Develop AI governance framework aligned with APRA, ASIC, and AML/CTF. Map use cases against regulatory risk and business value simultaneously.
Phase 2 (8–12 wks)
Low-Risk Proof Points. Deploy in back-office, compliance, and risk functions where regulatory alignment is clearest and volume creates immediate ROI.
Phase 3 (12–20 wks)
Customer-Facing Extension. Extend to customer service, lending decisions, and product recommendations with proven governance.
Phase 4 (Ongoing)
Systematic Scaling. Move beyond spot experiments to cross-organisational adoption. The more diverse the AI use cases, the higher the overall ROI.

The Perception Gap That Must Be Addressed

Research across 5,000 white-collar workers found that 81% of C-suite believe their company has a clear AI policy, while only 28% of employees agree — a 53-point gap. Tool access shows a 48-point gap. Training shows a 54-point gap. In financial services, where frontline staff process transactions and serve customers, this gap will not self-correct without intervention.

Frequently Asked Questions

How do we start AI adoption when APRA is watching?
Start with governance. Build an AI governance framework aligned with CPS 230 and CPG 235 before deploying your first model. Institutions that build governance first move faster than those that bolt it on after problems emerge.
Where should Australian banks focus AI investment first?
Risk reduction and compliance automation. These use cases deliver the highest ROI in financial services, carry the lowest regulatory risk, and build organisational confidence. Fraud detection, AML/CTF enhancement, and regulatory reporting are proven starting points.
How do we compete with fintechs on AI without their risk appetite?
Leverage your advantages: data depth, customer trust, and regulatory expertise. AI-enabled products built on deep customer data with proper governance create experiences fintechs cannot replicate because they lack the data foundation.

Want to discuss how this applies to your organisation?

We'd love to have a conversation about your specific challenges and how AI can help.

Start a ConversationMore Insights