Mortgage Broker Lender Policy Comparisons: Use case and AI Confidentiality

A lender policy comparison sets one client’s numbers against several lenders’ rules: how much of a bonus is counted, how much of the expected rent, the highest loan-to-value ratio, the rate applied to a card limit and what each lender would pay the brokerage. Brokers paste the matrix or upload the file to ChatGPT or Claude to ask which lender fits, to find the policy that blocks a deal and to write up why one was chosen. The sheet carries the client’s employer and pay, a rent estimate, existing debts, credit policy read from lender portals, commission rates and sometimes a lender manager’s unwritten remark. This page marks every cell in a lender policy comparison that the National Consumer Credit Protection Act and the Privacy Act bear on, and shows how the same sheet is worked on Ermos.

The lender policy comparison

Built after the fact find and before a lender is chosen, and kept with the file. No provision requires a policy comparison. ASIC’s RG 273 expects a mortgage broker to keep records that show the options and the recommendation given and the reasons why (RG 273.165(g)), and lists outputs of product comparison tools and calculators among the forms a record can take (RG 273.167(e)); RG 273 is guidance and not law. The best interests duty (section 158LA) and the conflict priority rule (section 158LB) of the National Consumer Credit Protection Act 2009 apply to a licensee that is a mortgage broker, and sections 158LE and 158LF apply to its credit representatives.

fx=SUM(C4:C6)
ABCDEF
1Policy itemLachlan PlunkettSouthgate BankTamarind BankOrrin Building SocietyHartwell Credit Union
2Client and purchase
3Policy used (portal copies, 18 Nov)
4Base salary counted ($ a year)138,000138,000138,000138,000
5Bonus counted ($ a year)9,60011,0006,0008,250
6Rent counted ($ a year)23,29621,84020,38423,296
7Total income counted ($ a year)170,896170,840164,384169,546
8Card commitment counted ($ a month)240304240304
9Repayment on $410,000, 30 years P&I ($ a month)2,5082,4692,5352,522
10Annual package fee ($)39503500
11Upfront commission ($)
12Lender manager
13Short list
14Status
Policy matrixClient scenarioNotes

An invented policy comparison as at Wednesday 18 November 2026. The client, employer, property and lender manager are invented, and so are every lender policy, rate, package fee and commission rate. Hover or tap any highlighted cell for the analysis; the full list is below.

Lender policy comparison builder

An Agent+ that lays one client's position against each lender's policy, working from the fact find and the policy notes and commission tables your brokerage holds. It fills the comparison grid, shows where the policies differ, and flags anything resting on a verbal remark rather than a written policy.

What it does

  1. Reads the fact find and your lender policy notes
  2. Fills a comparison grid, one column per lender
  3. Flags points resting on verbal or unconfirmed advice

What stays with your people. A broker chooses the lender and records why. The agent lays out the comparison; it does not recommend a loan.

An Agent+ is built once and anyone in the practice can run it. It works from your own documents and runs on Ermos Dominion, the AI unit in your office, or on Ermos Edge, your own instance in an Australian data centre.

What leaves the brokerage

  1. 1
    The client and the purchase

    A named client, his age, the address of the unit he is buying, the price and the loan. Personal information, and the deal itself. Whether the Privacy Act reaches a particular brokerage depends on its turnover and activities (section 6D); the client expects confidence either way. Privacy Act guide.

  2. 2
    Employer and salary

    A named employer, a role, a start year and a base salary. Section 117(1)(b) of the National Consumer Credit Protection Act requires reasonable inquiries about the consumer’s financial situation, and this is what those inquiries found. Privacy Act guide.

  3. 3
    Bonus history

    Two years of variable pay. Each lender counts it differently, so the figure that matters is the one the broker has verified. Section 117(1)(c) requires reasonable steps to verify the financial situation, and under section 118(4) only information the licensee had reason to believe was true, or would have had reason to believe was true had it made the inquiries and verification in section 117, is taken into account in deciding whether a contract is unsuitable. Privacy Act guide.

  4. 4
    The rent estimate

    An agent’s appraisal is an estimate and no lease exists yet. A model given the figure can treat an estimate as verified income. Under section 118(4) only information the licensee had reason to believe was true, or would have had reason to believe had it made the inquiries and verification in section 117, counts.

  5. 5
    Existing home loan and card

    The lender, the balance and the repayment on his own home, and a card limit. The commitments a lender counts, and a map of his credit relationships. Privacy Act guide.

  6. 6
    The lenders’ credit policy

    Four lenders’ rules as read from their broker portals: loan-to-value limits, how much of a bonus and of rent is counted, the rate applied to a card limit and the investment rate. Commercial information the brokerage holds through its accreditation; check the lender and aggregator terms before pasting it elsewhere. A model asked from memory can also state a policy that has since changed. ASIC’s RG 273.60 recognises that a lender’s credit policy and risk appetite may keep some consumers from the lowest cost products.

  7. 7
    Commission by lender

    What each lender would pay the brokerage on this loan, from $2,460 to $2,870. A difference in what lenders pay is a possible conflict between the consumer’s interests and the licensee’s, and section 158LB requires a licensee that knows or reasonably ought to know of such a conflict to give priority to the consumer’s interests. Also commercial terms the brokerage may hold under confidence; check yours.

  8. 8
    A lender manager’s unwritten remark

    A named lender employee, a date and a time, and an indication that is not policy. A model asked about the sheet may treat it as policy. A broker who relies on it should record it: RG 273.167(b) lists file notes, including records of conversations, among the forms a record can take.

  9. 9
    The short list and the reasoning

    Which lender comes first and why, the commission conflict and a note that the tax effect belongs to his accountant. RG 273.92 says it would generally be helpful to present the consumer with a shortlist with one option recommended, and RG 273.93 says it is important that the consumer understands why options were chosen and others left out, RG 273.54 expects evidence for recommending a higher cost loan, and RG 273.141 to 273.143 do not expect a broker to give detailed tax advice. Section 158LA is the duty that the reasoning has to meet.

  10. 10
    The status line

    A step in the process: fact find done, credit guide sent, no assessment yet. Section 115(1) says a licensee must not suggest that the consumer apply for a particular credit contract with a particular credit provider, or assist the consumer to apply for one, unless within 90 days (or another period the regulations prescribe) before that day it has made a preliminary assessment and made the inquiries and verification in section 117. No personal detail beyond the process.

What happens in Frontier Models (ChatGPT, Claude)

  • The whole document is transmitted to the provider’s servers. On personal plans the user cannot choose where it is stored or processed, and for the major providers that has generally meant outside Australia, under that country’s legal process. Some business plans offer storage in Australia, but the provider still receives the content.
  • On consumer tiers, prompts can be retained and used to train future models.
  • The provider’s staff and systems can reach the content under the provider’s own policies.
  • Once sent, the brokerage no longer controls it. Retention and deletion happen on the provider’s terms.

The same document on Ermos

Ermos Dominion

The document is drafted on a unit in your office, from your own templates and the client file. Every marked item stays on the unit, which keeps working with the internet switched off. Nothing in it is transmitted to an AI provider, so there is no disclosure to analyse: the confidentiality work is the same as for your file server.

Ermos Edge

Your documents, the index and your access boundaries stay on your own instance in an Australian data centre. When a staff member asks for this draft, only their question and the excerpts they are permitted to see go to a named Australian provider, who is contractually barred from keeping the content or training any model on it. Nothing is processed outside Australia.

Frontier Models (ChatGPT, Claude)Ermos EdgeErmos Dominion
Where the document is processedProvider servers, generally offshore on personal plansYour instance in an Australian data centre; answers generated in AustraliaThe unit in your office
Who can read itProvider staff and systems, under policyYour staff, within the boundaries you set. The named Australian provider processes only permitted excerpts, under contract, and keeps nothingYour staff only
Is it keptOn the provider’s termsKept on your instance; the provider is barred from retaining itStays on the unit
Can it train a modelOn consumer tiers, yesNo, by contractOnly your own unit learns from it, on your hardware

Part of Use cases in mortgage brokerages. See how this works in practice: Lender policy lookups on Ermos.

Common questions

Do ChatGPT and Claude use the information in a lender policy comparison?
Everything pasted or uploaded from the sheet goes to the AI company's servers and is processed there to produce a reply. As at October 2026, whether a conversation on a personal plan (ChatGPT Free, Plus or Pro; Claude Free, Pro or Max) can be used to train the models depends on a setting each user controls. Business plans (ChatGPT Business and Enterprise, Claude Team and Enterprise) and the developer APIs sit under commercial terms, and OpenAI and Anthropic each say they do not train on that content by default (OpenAI, Anthropic). On either product, the thumbs up and thumbs down buttons can send a whole conversation back for training. The sheet sets Lachlan Plunkett's employer and bonus history beside four lenders' credit policies and the upfront commission each would pay the brokerage, from $2,460 to $2,870.
Can we use ChatGPT or Claude for a lender policy comparison if we remove the client's name?
Not safely, because the numbers still describe one buyer. The Office of the Australian Information Commissioner (OAIC) recommends in its guidance on commercially available AI products that organisations do not enter personal information, and particularly sensitive information, into publicly available generative AI tools, because of the significant and complex privacy risks. Whether the Privacy Act reaches a particular brokerage turns on section 6D, but the client expects confidence either way. Without his name the sheet still shows an operations manager at Pennywell Packaging, bonuses of $11,000 and $13,000, a Werrington unit at $585,000 and a $410,000 loan. Removing names rarely makes a document anonymous. Information stays personal information while the person is reasonably identifiable, and the OAIC warns that information can be at risk of re-identification even when it has been de-identified or anonymised.
Can ChatGPT or Claude be trusted to draft a lender policy comparison?
Not without a check. Language models can produce fluent text with wrong figures, wrong totals and formulas that look right but are not, and a policy asked for from memory may have changed. A qualified person should check every item against the file and the lender portals. In this sheet the counted bonus and rent, the repayments and the short list would do most harm if wrong, and the rent figure is only an agent's appraisal with no lease. Under section 118(4) only information the licensee had reason to believe was true, or would have had reason to believe had it made the section 117 inquiries, counts.
How can a mortgage brokerage use AI for a lender policy comparison without sending it to ChatGPT or Claude?
Ermos is built so the document stays under your control. On the Dominion AI unit it is drafted inside your office and none of it is sent to an AI provider. On Ermos Edge your documents and index stay on your own instance in an Australian data centre, and only the question and the excerpts the user is permitted to see go to a named Australian provider that is contractually barred from keeping them or training any model on them. The comparison is produced from the client file and the brokerage's own policy notes and template. The broker checks every counted figure and the reasons for the short list, then signs it off.
Does a mortgage broker have to compare several lenders?
The Act sets no number. Section 158LA requires a licensee that is a mortgage broker to act in the consumer's best interests in relation to the credit assistance. ASIC's RG 273, which is guidance rather than law, says a broker should be accredited with a reasonably representative panel (RG 273.113), that presenting more than one option is good practice unless there is a good reason not to (RG 273.89), and that the consumer should understand why options were selected or left out (RG 273.93).
Can a broker recommend the lender that pays the highest commission?
Section 158LB requires a licensee that is a mortgage broker, if it knows or reasonably ought to know that its interests conflict with the consumer's, to give priority to the consumer's interests. ASIC's RG 273.54 says a recommendation of a higher cost loan needs evidence showing why it is in the consumer's best interests. Part 3-5A also deals with conflicted remuneration in Division 4. The recommendation has to rest on the client's interests, so the file should show why this lender suits this client.
What records should a broker keep of a lender comparison?
ASIC's RG 273 expects a mortgage broker to keep records showing how it complied with the best interests obligations, including the options and recommendation given and the reasons why (RG 273.165(g)). Comparison tool outputs, calculators and file notes of conversations can all be records (RG 273.167). How long to keep them depends on factors such as the loan term (RG 273.168). RG 273 is guidance and not law.

See Dominion answering from your own documents.

A demonstration takes 30 minutes, uses no client data, and comes with no obligation.

Book a demo