Westpac Accelerates AI Adoption To Transform Business Banking Operations
Last updated: October 21, 2025, 4:02 pm
Westpac is expanding AI across its business lending operation-embedding it from credit assessment through customer engagement. The bank's investment follows similar moves by Commonwealth Bank, signaling a broader shift to analytics and automation across business banking.
What's new
Westpac highlighted its partnership with RDC.ai (formerly Rich Data Co), formed in 2021 and built with Amazon Web Services. The work has evolved from traditional credit models to include generative and agentic AI-expanding use cases beyond decisioning.
AI now supports loan evaluation and helps surface new value in Westpac's data assets. The bank is also piloting an AI call assistant with its scam and fraud team to analyze live calls, flag potential scams, and guide operators to consistent responses.
Why operations teams should care
This is a play for speed, risk control, and cost discipline. AI can shrink cycle times, create consistent decisions at scale, and improve first-time-right outcomes across underwriting and servicing.
Where AI fits in the lending flow
- Intake: Pre-qualify and triage applications; collect and validate documents.
- Assessment: Score credit, segment risk, and propose structures based on policy and appetite.
- Decisioning: Recommend approvals with rationale; surface exceptions and required escalations.
- Fulfillment: Automate covenants, documentation checks, and onboarding tasks.
- Servicing and engagement: Trigger proactive outreach on early warning signs, renewals, and upsell.
Fraud operations: live-call assistant
The pilot assistant monitors calls in real time, spots scam indicators, and prompts next-best actions. Expect faster handling, tighter consistency, and better capture of incident data for continuous model improvement.
Data, governance, and model risk
Scaling AI across credit and fraud raises the bar on governance. You'll need clear model ownership, challenge processes, drift monitoring, and auditable explanations-especially where decisions affect access to credit.
- Data: Document lineage, quality checks, and retention rules; separate training, validation, and production feeds.
- Policy: Encode credit policy and fraud playbooks so models align with risk appetite and compliance.
- Controls: Human-in-the-loop for exceptions; bias testing; periodic revalidation and stress tests.
Implementation playbook
- Start with high-friction steps: document extraction, income verification, and risk scoring.
- Instrument the journey: track handle time, queue depth, SLA compliance, and error types.
- Stand up MLOps: version models, automate deployment, and enable rapid rollback.
- Train frontline teams: prompt patterns, escalation paths, and how to interpret AI recommendations.
- Close the loop: feed outcomes back into models to improve accuracy and reduce false positives.
KPIs to watch
- Underwriting: time-to-decision, approval rate by segment, variance to policy, exception rate.
- Quality: first-time-right percentage, rework rate, documentation error rate.
- Fraud: detection rate, false positive rate, average handle time, dollar loss prevented.
- Customer: NPS/CSAT for lending and fraud-touch journeys, complaint volumes.
- Ops cost: cost per application, cost per contact, backlog days.
Agentic AI in practice
Beyond scoring, agentic AI can coordinate tasks across tools: fetching documents, updating core systems, and kicking off workflows with guardrails. Keep actions constrained to policy, log every step, and require human approval for sensitive moves.
What this signals for the industry
Major banks are moving from pilots to embedded AI across core operations. The competitive edge will come from how reliably you integrate AI into processes, not just the model you choose.
Next steps for operations leaders
- Map your lending and fraud workflows and mark steps fit for AI assistance within 90 days.
- Set a lightweight model governance framework before scaling; involve risk and legal early.
- Pilot with clear exit criteria: target SLA gains, loss reduction, and customer outcomes.
- Build capability: upskill teams on prompt patterns, oversight, and exception handling.
If you're building team capability, see curated options by role at Complete AI Training.
Your membership also unlocks:
