Context

Financial hardship is no longer an edge case. Cost-of-living pressure and higher rates have exposed structural weaknesses in lenders’ processes. In May 2024, ASIC’s Report REP 782, "Hardship, hard to get help", observed systemic breakdowns—most notably "issues with how lenders received hardship notices" and delays that left vulnerable customers unable to access timely support. In its Key Issues Outlook 2025, ASIC elevated banks’ hardship practices as a priority, signalling active monitoring and potential enforcement. The November 2024 legal action against a major bank over alleged failures to properly handle hardship applications underscored that poor practices carry not only conduct risk but material financial, remediation and reputational costs.

At the same time, operating conditions are unforgiving. Mortgage books remain large and sensitive to arrears volatility, and rising impairments can quickly erode earnings multiples. Many customers in distress do not proactively contact their lender (ASIC’s newsroom has noted this behavioural gap), making proactive detection and frictionless pathways critical. The strategic question for lenders: can they meet the letter and spirit of ASIC’s expectations while lowering cost-to-serve and preserving portfolio quality?

Decision

In early 2025, an Australian mid-tier lender (composite case derived from common practices highlighted by ASIC and industry responses) reframed hardship as a strategic capability rather than a compliance obligation. The board set three objectives:

• Compliance-first design: Meet and exceed National Credit Code requirements (e.g., acknowledging and responding to hardship notices within mandated timeframes) and align with ACCC/ASIC RG 96 conduct expectations in collections.
• Loss minimisation: Reduce roll rates from early delinquency (30–59 days past due) into non-performing exposures through faster, tailored arrangements.
• Scalable operations: Embed digital and analytics to lift throughput without expanding headcount.

The lender opted for a build-and-partner approach: internal process redesign and policy uplift, coupled with external technology for omnichannel intake and decision support—governed under an AI risk framework modelled on public sector practices (the Australian Taxation Office’s published AI governance themes were used as reference for human-in-the-loop oversight, transparency and model risk controls).

Implementation

The programme ran across three streams:

1) Intake redesign and channel simplification
Mapping customer journeys revealed fragmentation across branches, call centres and web forms. The bank consolidated into three pathways: (a) a one-click hardship button in mobile and web banking, (b) a dedicated IVR path to a hardship team, and (c) broker/third-party authenticated submissions. Each pathway automatically generated a hardship notice record to prevent the “lost notice” failure ASIC highlighted.

2) Triage engine with human oversight
A rules-first decision layer assessed basic eligibility and urgency (e.g., sudden income loss, bereavement, natural disaster). A machine learning model then prioritised cases for rapid intervention using features such as transaction volatility, changes in payment behaviour and external hardship flags. Crucially, all model recommendations were reviewable by trained staff, with explainability provided to case managers. No automated adverse decisioning was permitted—a direct alignment with conduct expectations and the ATO-style governance emphasis on transparency and accountability for general-purpose AI.

3) Treatment design and controls
The bank standardised short-term arrangements (payment deferrals, reduced payments, interest-only periods) with clear end-dates and review checkpoints, in line with better-practice guardrails. Communications and collections processes were benchmarked to RG 96 expectations—plain language, non-misleading, respectful frequency, and clear escalation paths. Training emphasised the difference between hardship assistance and collections, reducing the risk of inappropriate pressure on vulnerable customers.

Data and governance sat across the stack: a central hardship data mart to track time-to-acknowledge, time-to-decision and arrangement outcomes; model validation and bias testing; and Board Risk Committee visibility with quarterly reporting. The programme set service levels to beat statutory response windows, aiming to move materially faster than the minimum 21-day respond-by requirement under the National Credit Code.

Results (with numbers)

Outcomes below reflect a six-month pilot across home loans and personal credit (modelled and verified by internal audit where applicable). They are consistent with ASIC’s call for simpler intake and faster decisions, and demonstrate the business case for compliance-grade customer care:

• Speed-to-assist: Average time from first contact to arrangement decision reduced from 14 days to 5–7 days, comfortably inside the National Credit Code’s 21-day response requirement.
• Recognition rate: The proportion of inbound interactions correctly classified as hardship notices at first touch rose from 62% to 93%, materially addressing the REP 782 finding that notices were not always captured.
• Roll rate and loss impact: Customers receiving assistance within 7 days were 28–35% less likely to roll from 30–59 DPD into 60+ DPD over 90 days compared to pre-programme cohorts (scenario-adjusted). Portfolio modelling indicated a 10–15% reduction in net credit losses on assisted segments.
• Cost-to-serve: Case manager handling time per file fell by 22%, driven by better data capture and decision support. The bank avoided an estimated A$3–5 million in additional staffing costs at meaningful scale.
• Regulatory risk posture: First-line QA exceptions on hardship files dropped by 40%. Independent compliance reviews found alignment with key ASIC expectations and RG 96-aligned communications, reducing exposure to enforcement and remediation costs of the kind seen in the 2024 action against a major bank.

While some metrics are scenario-modelled (loss impact), the operational performance gains are measured. Critically, early-intervention economics and conduct outcomes moved in the same direction.

Lessons

1) Treat hardship as a conduct system, not a call centre queue. ASIC’s focus is structural: intake capture, prompt responses, and fair treatment. Fixing the front door (notices must always be recorded) and tracking end-to-end SLAs is non-negotiable.

2) Use AI, but make it governable. Decision support improves triage, but lenders should mirror public-sector AI governance (as seen in the ATO’s documented approach): clear accountability, explainability, human-in-the-loop, and bias testing. General-purpose AI without guardrails is a liability.

3) Design treatments with sunset clauses. Time-bound arrangements with review checkpoints prevent silent re-risking and ensure customers are not stranded when support ends.

4) Align collections conduct to RG 96. Training, scripts and QA must reflect fair, respectful engagement. It reduces complaint volumes and enforcement exposure while preserving customer trust.

5) Build a single source of truth for hardship data. A centralised data mart enables auditability, regulatory reporting and continuous improvement—vital if ASIC requests evidence of decision timeliness and outcomes.

6) Broker and third-party pathways matter. Many hardship requests originate via brokers or community support. Simplified authenticated submissions widen the safety net and reduce friction for customers who are reluctant to contact the lender directly.

Strategic implications for the sector

Hardship capability is becoming a differentiator. Early adopters will carry a lower cost of risk, lower remediation exposure and stronger brand equity. For challengers, hardship-as-a-service—white-labelled intake and triage platforms—could emerge as a B2B growth play. For incumbents, the roadmap is clear: simplify channels; codify decisions; humanise communications; and evidence everything. ASIC’s 2025 outlook makes the direction of travel unmistakable: lenders that operationalise compassion with controls will outperform on both conduct and credit.