AI in Banking: Why ASX Banks Lifted Into Year-End

The ASX 200 finished last Friday (December 19) up 0.4% to 8,621.40, with the strongest push coming from banks and tech—exactly the combination you’d expect to benefit when markets start sniffing out rate cuts. US equities also rebounded, with the S&P 500 up 0.8%, after a softer inflation read lifted sentiment.

That’s the market headline. The more useful story for anyone building, buying, or selling in financial services is underneath it: banks are being re-priced as “operationally smarter” businesses, and a big slice of that belief is tied to AI in banking—not as hype, but as measurable improvements in fraud losses, cost-to-serve, and customer retention.

This post is part of our AI in Finance and FinTech series, and it looks at what the bank-and-tech rally signals, where AI is already showing up on P&Ls, and how to evaluate AI initiatives that actually move the needle.

What the ASX move really says about banks and AI

Answer first: The rally says investors are rewarding businesses that can protect margins and manage risk if rates fall—and AI is one of the clearest tools banks have to do both.

Banks were broad-based higher in the session: Commonwealth Bank +1.8%, Westpac +1.3%, NAB +0.8%, while ANZ ended flat amid news of increased penalties in a misconduct case. When a whole sector moves together like that, it usually isn’t about a single product launch or one earnings beat. It’s about macro expectations plus confidence in execution.

Here’s why AI belongs in that “execution” bucket:

• Net interest margins compress when rates fall. Banks need to offset that with lower operating costs and better risk pricing.
• Credit quality gets harder to read when the economy slows. Better early-warning systems and dynamic affordability models matter.
• Scams and fraud don’t take holidays. December is peak season for social engineering and card-not-present fraud.

AI systems that reduce fraud losses, automate servicing, and improve credit decisions translate into something investors love: more earnings resilience.

A useful way to think about it: when markets price “banks up,” they’re often pricing “risk down.” AI is one of the few levers that can reduce risk while also cutting cost.

Why rate-cut optimism boosts AI-heavy bank strategies

Answer first: If rate cuts arrive, bank growth shifts from “margin-led” to “volume- and efficiency-led,” and that’s where AI-driven operations outperform.

In the source story, US inflation printed at 2.7% last month—still above the Fed’s 2% target, but enough to restart the conversation about cuts. Whether cuts come early or late, the direction of travel changes boardroom priorities.

AI that protects revenue when margins compress

When net interest margin tightens, banks typically fight back with:

1. Better pricing and segmentation (who gets which rate, fee, limit)
2. Lower cost-to-serve (automation and smarter routing)
3. Lower losses (fraud, credit impairment, operational risk)

AI supports all three—but only if it’s connected to decisions. A model sitting in a dashboard is a science project. A model embedded in credit policy, fraud rules, or call-centre workflow is a commercial asset.

Where banks are most likely to deploy AI in 2026 budgets

Over the last year, I’ve noticed a consistent pattern: banks are funding AI where ROI can be audited. That means fewer “innovation labs” and more production deployments in:

• Fraud/scam detection (real-time risk scoring, mule account networks)
• Collections and hardship (next-best action, vulnerability detection)
• Contact centre automation (summaries, suggested replies, compliance prompts)
• AML alert reduction (triage, entity resolution, prioritisation)

The shared theme is simple: reduce manual work per customer, and catch the bad stuff earlier.

Fraud and scams: the AI use case investors actually believe

Answer first: Fraud and scam prevention is the most “bankable” AI use case because it produces direct, countable savings and reduces regulatory pain.

Australia’s scam environment has become more sophisticated—multi-channel, fast-moving, and coordinated. Traditional rules still matter, but they struggle with novel patterns and cross-account networks.

What modern AI fraud stacks look like

The strongest bank fraud programs typically combine:

• Graph analytics to detect mule networks and “hub” accounts
• Behavioural biometrics (typing cadence, device signals, session behaviour)
• Anomaly detection on transfers and payee creation
• Natural language models for call/chat red flags (coaching, coercion, urgency)

The goal isn’t “stop all fraud.” The goal is reduce losses without wrecking customer experience. That last part is why AI matters: it can deliver more accurate risk scores with fewer false positives.

A practical metric: cost per prevented loss

If you’re scoping a fraud AI program, force the discussion into one number:

• Cost per prevented loss = (Program cost) / (Estimated losses avoided)

It’s not perfect, but it makes the project comparable to other investments. It also exposes weak assumptions early.

Personalisation and customer retention: the quieter AI advantage

Answer first: AI-driven personalisation is less visible than fraud prevention, but it’s a major driver of lifetime value—especially as competition heats up from digital banks and fintech apps.

When markets rally into late December, banks benefit from sentiment, but they also benefit from something structural: customers are less “sticky” than they used to be. Switching friction is lower, price comparison is easier, and fintech onboarding is faster.

AI helps banks compete on experience without ballooning headcount.

Examples of personalisation that actually pays off

The profitable forms of personalisation tend to be unglamorous:

• Next best action in-app (e.g., prompting offset account setup when a mortgage settles)
• Proactive cashflow nudges (predicting bill stress before a missed payment)
• Merchant-level insights (categorising spend correctly, surfacing subscriptions)
• Limit and repayment optimisation (right-sizing limits to reduce risk and increase usage)

Done well, this reduces churn and increases product-per-customer. Done badly, it feels creepy or irrelevant. The difference usually comes down to governance and data quality.

AI risk and governance: the ANZ headline is a reminder

Answer first: The same AI that boosts bank performance can create compliance and conduct risk if governance is weak—especially in trading, pricing, and customer communications.

The market update referenced an increase in penalties related to ANZ’s management of a bond trading deal, with strong language from the court about transparency and conduct. That isn’t an “AI story,” but it’s absolutely an automation story.

As banks add AI into:

• pricing decisions,
• trade surveillance,
• marketing copy,
• complaints handling,

…the risk of “we didn’t mean to” outcomes rises unless controls keep pace.

Minimum viable AI governance for banks

If you’re a banking leader trying to move quickly without stepping on rakes, these controls are non-negotiable:

1. Model inventory (what models exist, where used, who owns them)
2. Data lineage (what data is used, where it came from, retention rules)
3. Explainability standards for customer-impact decisions
4. Human override and escalation paths (especially for vulnerable customers)
5. Monitoring for drift, bias, and performance decay

The point isn’t bureaucracy. It’s protecting the licence to operate.

Tech stocks rose too—here’s how that connects to fintech AI

Answer first: Tech outperformance reflects investor appetite for AI-enabling platforms—cloud, data, security, workflow automation—which are the backbone of modern fintech.

The market wrap noted strong moves in local tech names, and also referenced ongoing global capital flows into AI. For fintech founders and product leaders, the takeaway isn’t “buy tech.” It’s that distribution and infrastructure are converging:

• Banks want AI, but they don’t want to build everything.
• Fintechs want bank-grade trust, but they don’t want bank-grade overhead.

The winners are the teams that can combine:

• bank-compliant delivery (audit trails, controls, resilience), with
• fintech speed (rapid iteration, strong UX), and
• measurable outcomes (loss reduction, faster onboarding, lower servicing cost).

A reality check on AI ROI

Even in a buoyant market, investors are asking tougher questions about AI payback. The fastest way to lose internal support is to promise “productivity gains” without proving where they show up.

If you’re pitching or prioritising an AI initiative, tie it to one of these levers:

• Losses (fraud/scams, credit impairment)
• Time (minutes saved per case, per contact, per claim)
• Conversion (approval rates, onboarding completion)
• Compliance (fewer breaches, faster investigation cycles)

If you can’t connect your model to a lever, it’s not ready for production.

Practical next steps: a 30-day AI plan for banking teams

Answer first: The best way to start is to pick one decision flow, instrument it, and ship a controlled pilot with clear ROI metrics.

Here’s a 30-day plan I’ve seen work inside banks and mature fintechs:

1. Pick a workflow with high volume and clear cost
   • Example: scam triage, disputes, small business onboarding, AML alert review
2. Define the metric that matters
   • Example: false-positive rate, average handle time, loss per 1,000 customers
3. Map the decision points
   • Where can an AI score or summary change an action?
4. Run a shadow mode test
   • AI makes recommendations, humans still decide; measure lift safely
5. Deploy with guardrails
   • Monitoring, fallbacks, escalation for edge cases and vulnerable customers

The discipline is the advantage. Most companies get stuck arguing about tooling. The real battle is decision design.

What to watch next as markets head into 2026

The ASX ended the week down 0.7% overall, but the session strength revived talk of a late-December “Santa rally.” Whether that seasonal tailwind appears or not, the more durable trend is clear: finance is being rebuilt around AI-enabled decisioning.

If you’re in a bank, the mandate is straightforward: use AI to reduce losses, shrink cost-to-serve, and improve customer outcomes—while staying audit-ready. If you’re in fintech, the opportunity is just as clear: ship tools that banks can adopt without inheriting new conduct risk.

If you want to pressure-test where AI in banking will create the next wave of winners—fraud, credit, service, or markets—what’s the one decision in your organisation that still relies on “gut feel” more than evidence?