Do lenders use AI to assess business loans? What the software reads, and what changes on 10 December 2026

Yes. Most Australian lenders now run at least part of a business loan application through software before a person sees it, and some small unsecured decisions are made almost entirely by automated rules. Larger, property-secured or unusual applications usually still reach a human assessor. From 10 December 2026, privacy policies must also say when computer programs make significant decisions using personal information.

That last change is new, and it’s prompting a reasonable question from owners: who, or what, is actually reading my application? This guide answers that one step at a time.

Do Australian lenders actually use AI to assess business loans?

Many do, though “AI” covers a wide range. Some systems are simple rules (“decline if the business has traded under six months”). Others are statistical scoring models. A growing number use machine learning to sort transactions or flag risk.

The regulator has noticed. In its 2024 report Beware the gap, ASIC reviewed how 23 financial services and credit licensees were using AI, and warned that adoption was moving faster than the governance around it. It singled out a credit scoring model whose workings the licensee couldn’t fully explain. ASIC also made the point that existing obligations are technology neutral: a lender’s duties don’t change because a computer made the call. You can read the findings in REP 798 on asic.gov.au.

Where automation typically shows up in a business loan:

Stage	What software commonly does	Where a person usually steps in
Identity	Checks ID documents and matches names, dates of birth and addresses	Mismatches, name changes, overseas documents
Business checks	Confirms ABN status, GST registration, company details and trading age	Restructures, new entities, trusts
Bank statements	Categorises income, spending, repayments and dishonours	Seasonal patterns, one-off events, mixed accounts
Credit file	Pulls and scores personal and commercial credit history	Defaults with an explanation, disputed listings
Policy rules	Applies cut-offs for industry, loan size, trading time and debt levels	Exceptions, which only a person can approve
Property	Automated valuation estimates	Formal valuations, unusual or commercial property

The pattern is consistent. Software handles the standard parts well. People handle the exceptions.

Which applications are most likely to be decided by software alone?

The more standard and smaller the request, the more likely it is to be automated end to end. Broadly:

1. Small unsecured online applications. Often sized directly from bank statement data against fixed rules.
2. Short-term cash-flow products. Built for volume, so they lean heavily on automation.
3. Larger unsecured loans. Usually a mix: software prepares the file, a credit assessor signs off.
4. Property-secured loans. A person almost always reviews the security, title, valuation and exit plan.
5. Anything with ATO debt, defaults or a recent restructure. These are the cases rules handle worst, so they tend to need human judgement anyway.

If your situation sits in the last group, a fully automated route is often the wrong place to start. Our answer on what lenders look at explains what a person weighs that a rule can’t.

What changes on 10 December 2026?

From 10 December 2026, new Australian Privacy Principles 1.7 to 1.9 apply. They were added by the Privacy and Other Legislation Amendment Act 2024. Organisations covered by the APPs must explain in their privacy policy when they use a computer program to make, or do something “substantially and directly related” to making, a decision that could reasonably be expected to significantly affect an individual’s rights or interests, using personal information.

The privacy policy must describe:

• the kinds of personal information used by those programs;
• the kinds of decisions made solely by a computer program; and
• the kinds of decisions where a program does something substantially and directly related to the decision.

The OAIC consulted on guidance for the new obligation earlier this year and said it intended to publish that guidance before the start date. Details are on the OAIC consultation page, and the existing rules for privacy policies are in Chapter 1 of the APP guidelines.

For a business borrower, the practical effect is simple. After 10 December, you should be able to read a covered lender’s privacy policy and see, in general terms, whether credit decisions are automated and what personal information feeds them.

Does the new rule give me a right to a human review?

No. The obligation is about transparency. It requires a description in the privacy policy, not a personal explanation of your result, a notice that your application was automated, or a right to have a person look again. That puts Australia’s approach behind some overseas regimes, at least for now.

What you can still do:

• Ask the lender directly. “Was this decided by an automated system? Can a credit assessor review it?” Many lenders will answer plainly.
• Ask which rule or limit applied. Trading time, loan size against turnover, a credit file item, dishonours. Knowing the reason tells you what to fix.
• Check your own credit file. You can get a free copy of your credit report every three months from each credit reporting body. The OAIC’s credit reporting pages explain how, and how to correct errors.
• Choose a route with a person in it. A file a rule declined can look very different to someone who reads the explanation.

Does it apply when the borrower is a company?

The rule protects individuals. A company isn’t one. But very little business lending happens without personal information. Sole traders apply in their own names. Directors provide ID and consent to personal credit checks. Guarantors put their own finances on the table. That personal information is what the new transparency rule covers.

So even if the loan is in your company’s name, the parts of the decision that use your details are within scope where the lender is covered by the APPs.

Why can software say no when a person would say yes?

Because rules are blunt. They’re built for the typical file and can’t read the story behind an unusual one. Common automated trip-wires:

• A seasonal dip read as a decline in the business.
• One or two dishonours from a single bad month, counted without context.
• A recent ABN or restructure treated as a brand-new business, even when the trading history continues.
• ATO debt on a payment plan treated the same as unmanaged tax debt.
• An old default that’s since been paid and explained.
• Industry codes that sweep a low-risk business into a high-risk category.
• A loan amount outside the product’s band, rather than a sign the business can’t afford it.

None of these is necessarily a deal-breaker for a person who can ask questions. See what to do if the bank said no for the steps after an automated decline.

If you’d rather a person read your situation from the start, send a short enquiry and a specialist will call you.

How do I make my application easy to read for software and for people?

Both read better when the file is tidy and consistent. Before you apply anywhere:

1. Use one business account for business money. Mixed personal spending clouds the transaction categories.
2. Keep BAS lodgements up to date. Missing lodgements are a common automatic stop.
3. Match your details everywhere. Legal name, trading name, ABN and address should be the same on every document.
4. Pull your own credit report first and fix errors before a lender finds them.
5. Write a one-paragraph explanation for anything unusual: a quiet quarter, a default, an ATO plan.
6. State the amount and purpose precisely. A clear purpose sized to the need reads better than a round number.
7. Have the documents ready. Our list of what documents you need covers the usual set.

Our loan navigator is a useful pre-check. It’s rules-based and asks five questions, but it collects no personal details and makes no credit decision.

Should I worry about giving a lender access to my bank data?

Not if you know what you’re agreeing to. Most lenders need several months of statements, and many collect them electronically with your consent. Before you click “connect”, check three things:

• Who receives the data. One lender, or a platform that shares it further?
• How long access lasts. One-off collection or ongoing?
• What the privacy policy says. From 10 December, that includes whether the data feeds automated decisions.

If the answers are vague, ask. A lender who can’t explain how it uses your data isn’t one to rush into.

What does this look like in practice? (Illustrative example)

Illustrative only. The business and figures are invented.

A Brisbane joinery company applies online for an unsecured loan to buy timber for a large fit-out contract. Within minutes the lender’s system declines. The director asks why. The lender explains that its rules flagged two dishonoured payments in March and a 40% drop in deposits over winter.

The director knows both have a simple story. A large customer paid late in March, and winter is always quiet in their trade. They pull their credit report (it’s clean), write a short explanation of the seasonal pattern, attach the signed contract and send an enquiry to a specialist instead. A credit assessor reads the full year of statements, sees that the March gap was cleared within a fortnight and that the quiet months repeat every year, and finds a lender comfortable with the pattern.

Same business, same numbers. A different reader.

When you want a person, not a rule, to read your file

Automated systems are good at standard files and fast at saying no to everything else. If your business has a seasonal pattern, a past credit issue, ATO debt or simply a story that needs telling, you’re better off starting with someone who’ll read it properly.

Enquiring with us takes about 60 seconds and there’s no credit check when you first enquire. We don’t send your details to a pile of lenders, so your phone won’t light up with strangers. A real person looks at your individual situation and calls you to talk it through. Please fill in the form accurately, including anything unusual, so we can match you to the right option the first time.

See if you qualify →