Responsible lending file tests, hardship notice tracking, credit reporting corrections and collections conduct reviews are the strongest uses of AI agents in Australian consumer lending compliance. Compliance officers keep every finding, breach report and remediation decision. ASIC supervises credit conduct, APRA's CPS 230 applies to banks, and the Privacy Act's credit reporting rules apply.
What does a consumer lending compliance team in Australia spend its time on?
Responsible lending is the core. Before lending, a licensee must make reasonable inquiries about the consumer's requirements, objectives and financial situation, take reasonable steps to verify that situation, and assess whether the loan is unsuitable. Compliance tests files to see whether inquiries were real, verification happened, and the assessment holds up against the bank statements on file.
Hardship is the second load. Borrowers can give a hardship notice, and lenders must respond within set timeframes and consider variations. Missed or poorly handled notices are a common source of complaints and breach reports.
Credit reporting adds its own work: correct listings, timely corrections and handling disputes about repayment history. Collections conduct, product design and distribution obligations, complaints that escalate to the external dispute resolution scheme, and breach reporting to ASIC complete the picture.
Where can AI agents help?
| Use | What the agent does | Depth |
|---|---|---|
| Responsible lending file testing | Reads application, bank statements and the lender's assessment, and checks that declared expenses, verified income and the assessment agree | 1 |
| Hardship notice tracking | Identifies hardship notices in calls, emails and chat, logs them and tracks response deadlines | 1 |
| Hardship response drafting | Drafts the acknowledgement and information request for a hardship officer to send | 2 |
| Credit reporting checks | Compares repayment history listings with the account record and flags listings made during a hardship arrangement | 1 |
| Collections conduct review | Reviews call transcripts and messages for contact frequency, tone and pressure, and flags conduct concerns | 1 |
Hardship detection is often the first win. Customers rarely use the words "hardship notice"; they say they lost shifts or cannot make the next payment. An agent that reads every contact and flags these catches notices that agents on the phone missed. This connects to the wider question of whether an AI agent can monitor customer outcomes, which applies to Australian lenders even though the UK framing is different.
What stays with a person, and why?
Decisions to grant or refuse hardship variations stay with hardship officers. They involve judgement about a person's circumstances and have real consequences for the borrower. The agent drafts and tracks; a person decides.
Breach reporting stays with compliance. Australian credit licensees must report certain breaches and suspected breaches to ASIC within a fixed period, and whether an incident is reportable is a judgement the licensee owns. The agent can draft the assessment and calculate the deadline.
Remediation stays human. When file testing finds a pattern of unsuitable loans, deciding which customers are affected, how much to refund and how to contact them requires care. This is why we describe AI compliance agents as second line, never the last word.
Which rules apply in Australia?
ASIC regulates consumer credit, including responsible lending obligations, hardship, credit licensing, breach reporting, design and distribution obligations, and guidance on debt collection. Licensees must belong to the external dispute resolution scheme. APRA supervises banks and other authorised deposit-taking institutions, where CPS 230 on operational risk management covers the agent and its supplier.
The Privacy Act 1988, overseen by the OAIC, contains specific credit reporting provisions governing what credit providers may disclose to and use from credit reporting bodies, correction obligations, and complaints. An agent that reads or corrects credit information must stay within those rules.
There is no AI-specific law; the government has published a voluntary AI safety standard. For comparison with other markets, the US approach is set out in our piece on the CFPB and AI agents in consumer finance.
This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.
How should you start in the first 90 days?
First month: run hardship detection at depth 1 over a past month of contact. Compare the notices it finds with those the business logged. The gap is your baseline and usually your business case.
Second month: add responsible lending file testing at depth 1 on a sample of recent approvals. Compliance reviews every exception the agent raises and records whether it agrees.
Third month: if detection is reliable, trial hardship acknowledgement drafting at depth 2, each draft held for a hardship officer. Complete the CPS 230 documentation if you are APRA-regulated.
Where Pryme fits
Pryme certifies hardship detection separately from hardship response drafting, each on the lender's own past contacts, through Attest's seven gates. The escalation gate tests whether the agent routes a hardship signal to a person rather than handling it itself. Certificates lapse after 90 days, and every verdict is filed.
Questions people ask
Can an AI agent assess whether a loan is unsuitable?
It can test the assessment and flag inconsistencies, but the lender's credit decision and its responsible lending judgement stay with people under the lender's credit policy.
Can an agent stop collections when it detects hardship?
It can flag the account and draft a hold request, but the decision to pause collections should be made by a person within a clear policy, ideally quickly.
How does Canada's approach differ?
Canada splits consumer lending oversight between FCAC for banks and the provinces for other lenders. See our brief on compliance agents in Canadian consumer lending.