The Broker Times · Technology Brief
Willingness to Delegate Falls as the Stakes Rise
Experian and Forrester Consulting asked 481 credit-active Australians which lending tasks they would hand to an AI agent. The answers form a gradient, and it runs downhill toward commitment.
Share comfortable handing each task to an agent
Experian research conducted by Forrester Consulting, July 2026, as reported 18 September 2026.
What they expect an agent to do well
Same survey, as published.
The biggest number in the survey is not about capability
80% named concern about personal data being accessed or misused as their main hesitation. That is the same objection that lands on a broker asking for bank statements, payslips and identity documents — and most brokers have never scripted an answer to it.
What the gradient means on your desk
The comparison step is the exposed one
81% would automate search and compilation, and 80% believe software would look at more options than they could. A pitch built on breadth of search competes directly with that.
Breadth is not the real constraint
Accreditation, panel access and lender credit policy decide a residential recommendation. None of them are visible to a consumer-facing comparison tool.
Stated willingness is not behaviour
This measures what 481 people said in July 2026, and Experian is a credit bureau with an interest in the subject. Attribute the numbers rather than treating them as settled.
The broker takeaway
Consumers in this survey are comfortable automating finding out and reluctant to automate committing. The 22-point spread between those two ends is the honest map of where the channel is exposed.
The near-term change is likely clients arriving with a compiled shortlist and firm expectations, not clients who never arrive. That is a conversation starting at step three, and it rewards brokers who can talk about policy and accreditation rather than lender counts.
Sources: Experian research conducted by Forrester Consulting — 481 credit-active Australian consumers, July 2026, within a 6,247-person study across 13 markets — as reported by Australian Broker and The Adviser, 18 September 2026. Figures measure stated willingness, not observed behaviour.
Technology · Broker Analysis
81% Would Let an AI Agent Compile Their Loan Options. 59% Would Let It Apply. The Gap Is Where Brokers Still Sit
Experian’s numbers are not a single verdict on AI. They are a gradient — and it slopes down exactly as the task moves from gathering information to committing a client to a contract.
Experian, with Forrester Consulting, surveyed 481 credit-active Australian consumers in July 2026 as part of a 6,247-person study across 13 markets. The headline is that 59% would let an AI agent apply for a loan or credit card on their behalf. The more useful finding is what happens to that willingness as the task changes.
In this article
The gradient, not the headline
Most coverage of this research led with the 59%. Read on its own it sounds like a channel-threat story. Read against the other numbers in the same survey it is something more specific and more useful: consumers are sorting lending tasks by consequence, and delegating in inverse proportion to it.
| Task the consumer would hand to an AI agent | Share comfortable |
|---|---|
| Search for and compile offers | 81% |
| Negotiate on their behalf | 77% |
| Handle identity verification | 69% |
| Accept an offer against criteria they set in advance | 68% |
| Apply for a loan or credit card on their behalf | 59% |
Twenty-two percentage points separate the top of that list from the bottom. The tasks at the top are information tasks — gathering, comparing, summarising. The tasks at the bottom commit the consumer to something. The survey is not measuring enthusiasm for AI in general; it is measuring where people are prepared to stop supervising.
That gradient matters commercially because the broker’s value has never been evenly distributed across those five tasks either. The comparison step is the one clients most often assume they are paying for, and it is the one they are most willing to automate.
What consumers think an agent is good at
The survey also asked what consumers expect these tools to deliver, and the expectations are high.
| Consumer expectation of AI agents | Share agreeing |
|---|---|
| Trust a large language model to compare loans across providers | 79% |
| Believe an agent could evaluate more options than comparing manually | 80% |
| Think an agent could help secure a better rate or price | 78% |
| Expect an agent to identify hidden fees or contract terms | 76% |
Take the second one seriously. Eight in ten believe software can evaluate more options than a person working manually — and on raw breadth of search they are right. A broker’s honest answer is not that this is wrong, but that breadth is not the constraint on a home loan recommendation. Accreditation, panel access, lender credit policy and what a particular assessor will accept as income evidence are the constraints, and none of them are visible to a consumer-facing comparison tool.
The fourth expectation is the one worth watching. If 76% expect an agent to surface hidden fees and contract terms, the bar for a broker’s own disclosure conversation rises accordingly. A client who has been told by software that there is an annual package fee will not be impressed to discover you had not mentioned it.
Where the willingness stops
The two lowest numbers in the survey are the two that involve handing over consequence: accepting an offer on preset criteria (68%) and applying (59%). Roughly four in ten credit-active Australians would not let an agent submit an application in their name.
There is a reason those sit lowest that has nothing to do with technology. An application is a set of representations about the applicant. Getting it wrong is the applicant’s problem, not the tool’s. Most people understand that intuitively even if they could not name the mechanism.
Which points at where a broker’s work actually concentrates. Establishing what a client’s income really is, which lender will accept the evidence for it, what the file needs before it is submitted, and what happens when the assessor comes back — none of that is the comparison step that 81% are happy to automate.
The 80% who are worried about their data
The survey’s largest single number is not about capability. Eighty percent named concern about personal data being accessed or misused as their main hesitation about AI agents.
For a broker this is the most immediately usable finding in the study, because it describes an objection that already exists in your client base and has nothing to do with whether they like your service. Every client you ask for bank statements, identity documents and tax returns is a client in that 80%.
The practical read: being able to explain clearly what you collect, why you collect it, where it goes, who at the lender sees it and how long you keep it is now a competitive answer rather than a compliance chore.
It is also the one question a consumer-facing agent tends to answer worst, because the consumer genuinely does not know what the tool does with what it is given.
A note on what this survey is and is not
This is 481 credit-active Australian consumers surveyed in July 2026 by Forrester Consulting for Experian, within a 6,247-person international study. It measures stated willingness, not behaviour. People routinely say they would use a tool they then do not use, and the gap between the two is usually large.
Experian is a credit bureau with a commercial interest in how lending data and decisioning develop. That does not invalidate the research, but attribute it rather than citing the figures as settled market fact.
What an agent cannot carry
There is a structural point underneath the survey that no amount of consumer enthusiasm changes, and it is worth stating carefully.
The best interests duty is an obligation on mortgage brokers providing credit assistance. ASIC’s Regulatory Guide 273 sources it at RG 273.6 to sections 158LA and 158LE of the National Consumer Credit Protection Act 2009, and RG 273.7 defines mortgage brokers by reference to carrying on a business of providing credit assistance in relation to credit contracts secured by mortgages over residential property. The duty attaches to a licensed person and their business.
How that framework applies to consumer-operated AI tools is a question for regulators and legislators, and this article does not attempt to answer it. What can be said plainly is narrower and more useful: when a broker makes a recommendation, there is an identifiable party carrying a statutory duty, a licensee behind them, a file note explaining the reasoning, and an external dispute resolution path if it goes wrong. Those are the things the 59% are implicitly declining to give up.
General information, not compliance advice
This is a plain reading of published ASIC guidance, not advice about your obligations or about the regulatory status of any product. Nothing here asserts how the law applies to any particular tool or service. Check your own process with your licensee or aggregator’s compliance team, and seek independent legal advice where you need a view you can rely on.
What to review this week
- Write your data answer down, in plain words. What you collect, why, who receives it, how long you hold it. 80% named data concern as their main hesitation about agents; the same concern applies to you, and most brokers have never scripted the answer.
- Stop selling the comparison step as the whole service. 81% would delegate search and compilation. If your pitch is that you look at lots of lenders, you are competing on the one task consumers are most willing to automate.
- Get specific about panel and policy in client conversations. Breadth of search is not the constraint; accreditation and credit policy are. That distinction is invisible to a comparison tool and it is the honest version of what you do.
- Raise fees and contract terms before the client’s software does. 76% expect an agent to surface hidden fees. Being second to mention an annual package fee is a bad position to be in.
- Review what your own tools do with client data. If you use AI in your workflow, you should be able to say which tool, what it sees, and whether client data leaves your systems — before a client asks.
- Watch for agent-prepared clients rather than agent-replaced ones. On this data the near-term change is clients arriving with a compiled shortlist and a set of expectations, not clients who never arrive. Prepare for the conversation that starts at step three.
What to watch next
Three things. Whether stated willingness converts into actual use, because survey intent and behaviour diverge sharply in consumer finance. Whether lenders build agent-facing interfaces, since an agent can only compare what it can reach, and today that is mostly public advertised rates rather than the pricing a broker can escalate to. And whether the data-protection concern that 80% named turns into a real constraint on adoption or fades the way privacy concerns often have elsewhere.
The strategic read is more settled than the technology. Consumers in this survey are happy to automate finding out and reluctant to automate committing. Brokers whose value proposition sits in the first half of that sentence have a problem worth addressing now. Brokers whose value sits in the second half have a better story than they are currently telling.
Key takeaways
- Experian, with Forrester Consulting, surveyed 481 credit-active Australian consumers in July 2026 within a 6,247-person study across 13 markets.
- Willingness to delegate falls as consequence rises: 81% would let an agent search and compile offers, 77% negotiate, 69% handle identity verification, 68% accept an offer on preset criteria and 59% apply on their behalf.
- Expectations of capability are high — 80% believe an agent could evaluate more options than manual comparison and 76% expect one to identify hidden fees or contract terms.
- 80% named concern about personal data being accessed or misused as their main hesitation, which is the same objection brokers face when collecting statements and identity documents.
- Breadth of search is not the real constraint on a home loan recommendation; accreditation, panel access and lender credit policy are, and none are visible to a consumer-facing comparison tool.
- The survey measures stated willingness rather than behaviour, and Experian is a credit bureau with a commercial interest in the subject — attribute the figures rather than treating them as settled fact.
Common questions
Does 59% mean AI is about to replace brokers?
It means 59% of a 481-person sample of credit-active Australians said in July 2026 that they would be comfortable with an agent applying on their behalf. It is stated willingness, not behaviour, and the same survey shows willingness dropping steadily as the task carries more consequence. Treat it as a signal about which parts of the service are most exposed, not a forecast.
Which part of my service is most exposed?
On these numbers, the comparison step. 81% are comfortable handing search and compilation to an agent, and 80% believe software could evaluate more options than they could manually. If your client-facing pitch rests mainly on the number of lenders you look at, that is the part most easily replicated.
What is the strongest honest counter-argument?
That breadth of search is not what decides a residential loan recommendation. Accreditation, panel access, lender credit policy and what an assessor will accept as income evidence decide it, and those are not visible to a consumer-facing tool. That is a factual distinction rather than a sales line.
Does the best interests duty apply to an AI agent?
This article does not answer that, and brokers should not assume an answer either way. RG 273.6 sources the duty to sections 158LA and 158LE of the National Consumer Credit Protection Act 2009, and RG 273.7 defines mortgage brokers by reference to carrying on a business of providing credit assistance. How that framework applies to consumer-operated tools is a matter for regulators. Raise it with your licensee rather than relying on a general article.
What should I do first?
Script your data answer. 80% named data access and misuse as their main hesitation, and that concern lands on you every time you ask for bank statements and identity documents. Being able to explain clearly what you collect and where it goes is useful regardless of what happens with agents.
Sources and method: Survey findings, percentages, sample size and methodology as reported by Australian Broker and The Adviser, 18 September 2026, describing research by Experian conducted by Forrester Consulting — 481 credit-active Australian consumers surveyed in July 2026 within a 6,247-person study across 13 markets. Quote from Andrew Black, Chief Executive Officer of Experian Australia and New Zealand, as reported in that coverage. Regulatory references are to ASIC Regulatory Guide 273, Mortgage brokers: Best interests duty (June 2020), paragraphs RG 273.6 and RG 273.7. All figures measure stated willingness at a point in time and are subject to the research sponsor’s own methodology.
Breaking news for modern brokers
Lender policy, regulation and market moves, read for what they change on your desk.
More at The Broker Times →Broker Tool
Five Tasks, Five Different Answers
Each tab takes one lending task from the Experian survey, shows what consumers said about it, and sets out what a broker actually does at that step. Use it to work out which parts of your own pitch are exposed.
Search for and compile offers
The task consumers are most willing to automate, and the one most brokers put at the front of their pitch.
Negotiate on their behalf
High willingness, but a task where what is actually available differs sharply between a consumer and an accredited broker.
Handle identity verification
Willingness drops nine points from negotiation, and this is where the data concern starts to bite.
Accept an offer on preset criteria
The second-lowest score. Consumers will let software look, and hesitate to let it decide.
Apply for the loan or card
The lowest score in the survey, and the step that carries the most consequence for the consumer.
How to use this
This is a framing aid built from one survey, not a forecast or a compliance assessment. The survey measures stated willingness in July 2026, not behaviour. Nothing here states how any law applies to AI tools — raise that with your licensee.
Disclaimer: This article is for general information and professional development purposes only. It does not constitute legal, compliance, or financial advice. Brokers should consult their aggregator’s compliance team and, where required, seek independent legal advice regarding their obligations under the National Consumer Credit Protection Act 2009 and ASIC’s responsible lending guidelines.
