AI Now Pulls Escalated Pricing From Up to Nine Lenders in Minutes. RG 273 Still Asks You to Show Why You Picked One

Three broker-facing AI launches landed within a week of each other in September. One automates pricing escalation, one audits files at lodgement, and one listens to your client calls. Each speeds up a different stage of the file. None of them writes the part ASIC cares about most: the reasons you recommended one loan over the others you looked at. As the tools widen what you can see, that part gets harder, not easier.

What launched, and where each tool sits in the file

SecondBrain’s Sharp: pricing escalation

On 11 September, Broker Daily reported the launch of SecondBrain, a Sydney-developed AI platform whose lead module, Sharp, automates new-lending pricing escalation. According to Broker Daily, Sharp pulls current rates from lender pricing tools and sets them beside escalated pricing, so the broker sees both in one comparison. Founder James Larkey told the outlet that Sharp “requests pricing from up to nine lenders at once and returns the results in a single comparison, usually inside four minutes.”

Broker Daily’s report included a few details brokers should note before switching it on:

• Sharp currently handles new lending only. It does not reprice existing loans.
• It runs as “attended automation” inside Microsoft Teams. Brokers sign in with their own Microsoft 365 identity, and the platform uses Azure infrastructure in the Australia East region.
• Lender credentials are encrypted in per-firm vaults. Access requires manual multi-factor authentication, and each access logs the broker, lender, outcome and timestamp.
• Borrower data such as income, employment and credit files is not retained. The platform keeps only what Larkey called a “de-identified rate outcome”, which feeds weekly peer-rate reports broken down by LVR bracket, repayment type and borrower category.
• The wider platform also handles document classification, valuations, credit report ordering, bank statement processing and data entry into aggregator platforms such as Connective. Broker Daily reported that SecondBrain handled 31,209 individual jobs in August and is used by more than 40 brokerages.
• The product is sold on subscription, works with any aggregator, and, per Broker Daily, the founder is working with banks towards formal approval.

Salestrekker’s Audit AI: the lodgement check

Also on 11 September, Broker Daily reported that Salestrekker had released Audit AI. The module reviews key parts of a loan file at submission and flags issues that affect quality, accuracy or lender requirements. It also gives aggregators real-time visibility across their brokers’ submissions. It is available in Salestrekker 2.0, is switched on through approved aggregator partners, and joins the platform’s existing Compliance AI, Translate AI and Policy AI modules. Chief executive Dalibor Ivkovic said: “Audit AI closes the loop.”

Yellow Brick Road’s Nectaria: the conversation layer

On 14 September, The Adviser reported that Yellow Brick Road had unveiled Nectaria at its Next Level conference on the Gold Coast. Nectaria is a multi-agent platform that records and transcribes client calls, maps what was said to client profiles, spots missing information and creates follow-up tasks. It also searches lender policy databases and gives brokers compliance feedback and coaching. A business-coaching feature is due in October. YBR chief digital and marketing officer Kevin Mangano said: “We started with common pain points for our brokers and looked to build solutions that didn’t yet exist.”

Why pricing is the pressure point under RG 273

ASIC’s Regulatory Guide 273 is clear about cost. Paragraph RG 273.51 says ASIC generally expects the cost of a credit product, “such as interest rate, fees and charges and the size of repayments”, to be a factor brokers should prioritise. RG 273.54 goes further: “A failure to consider cost and investigate the lowest cost options available to the consumer may suggest non-compliance with the best interests duty.” Where a higher-cost loan is recommended, the same paragraph says that choice “will need to be supported by evidence demonstrating why that recommendation is in the consumer’s best interests.”

Once a tool puts nine escalated prices in front of you in minutes, you have investigated more low-cost options than you would have by hand. That makes it more obvious when you don’t recommend the cheapest one. Before, a file might show two or three lenders priced over two days. Now it can show nine, with timestamps. If the recommended loan isn’t the cheapest of the nine, the file needs to say why.

That isn’t a reason to avoid the tools. RG 273.56 says the lowest rate “is not necessarily the lowest cost option for all consumers”. It points to offset and redraw features that can save some borrowers a considerable amount of interest, and to annual or establishment fees that can make a lower-rate loan more expensive. RG 273.63 adds that quantifiable promotional offers, including cashbacks, waived fees and reduced rates, should be counted as part of the cost. A rate-only comparison is where your analysis starts, not where it finishes.

Timing matters too. For the week reported on 15 September, Australian Broker, citing Canstar data, said five lenders raised 115 fixed rates by an average of 0.27 percentage points, while two lenders made small variable-rate cuts. The RBA board meets on 28–29 September. When fixed pricing moves week to week, a same-day nine-lender snapshot helps clients, as long as the file shows when it was taken.

Five blind spots in automated escalation

1. Who chose the nine lenders?

RG 273.157 says that “product recommendation or aggregator software should not have in-built credit provider or product bias.” A pricing tool that queries a set list of lenders is shaping your shortlist. Find out who set that list, whether it changes by scenario, and whether it matches the lenders on your panel that could actually approve the deal. If a lender that fits the client’s policy needs is never queried, the comparison looks thorough but isn’t.

2. The back book isn’t covered yet

Broker Daily reported that Sharp handles new lending only. Retention repricing, where you ask an existing lender to sharpen a client’s rate, still happens outside the tool. RG 273.117 says that if you suggest a client stay in a loan you previously arranged, you still need to act in their best interests when you make that suggestion. So an existing client can get a less thorough comparison than a new one. Make sure your process covers that gap.

3. Rate is not cost

An escalated rate doesn’t account for package fees, offset value, cashback conditions or a client’s plan to fix part of the loan. Treat the comparison as input to a cost assessment, not the assessment itself.

4. Approval status and lender terms

Automated tools log in to lender portals with your credentials. Broker Daily reported that SecondBrain’s founder is working with banks towards formal approval. Before relying on any tool that works inside a lender’s systems for you, check with your aggregator and the lenders concerned that it is permitted. If you’re not sure, ask your licensee.

5. Peer-rate data can pull you towards the average

Weekly peer-rate reports by LVR and borrower type are useful for checking whether your pricing is competitive. They don’t replace a client-specific decision. Knowing what other brokers got for a similar LVR helps you negotiate. It says nothing about whether that loan suits this client’s objectives.

Lodgement audits and call transcripts: records, not reasons

RG 273.167 lists the forms records can take. They include “outputs of product and feature comparison tools”, “audio recordings”, and “evidence of the compliance systems you used”. So the outputs of all three tools can count as evidence. But RG 273.165(g) also expects records of “the options and ultimate recommendation you gave and the reasons why (including a detailed description of your decision-making process)”. A pricing screenshot shows the options. A lodgement audit shows the file was complete. A transcript shows what the client said. None of them records your reasoning.

Audit AI is designed to catch problems before a file reaches the lender, based on what Salestrekker has described: quality, accuracy and lender requirements. That reduces rework, and aggregator visibility may help licensees supervise. But a file can pass a lender-requirements check and still have a weak best-interests rationale. Those are two separate tests.

Call transcription raises a different issue. RG 273.169 says taking notes throughout the process “may help you to accurately capture your thinking and reasoning at each relevant point in time.” A full transcript is about as contemporaneous as a record gets. That also means it can contradict your fact-find or your file note. If a client says on a recorded call that an offset matters, and you then recommend a loan without one, the transcript is evidence against you unless the file explains the trade-off. If you record calls, read the summaries before you rely on them, and check consent and privacy requirements with your licensee.

A four-step workflow for AI-assisted files

1. Capture. Confirm the client’s objectives in your own words before any pricing runs: rate certainty, offset, speed, cashback, repayment size. If a transcript tool produced the summary, check it against your fact-find and fix any differences.
2. Compare. Run the escalation and save the full output, not just the winner, with its timestamp and the list of lenders queried. Note any suitable lender the tool didn’t query, and why.
3. Choose. Turn rate into cost: fees, offset value, promotional offers and their conditions (RG 273.56, RG 273.63). If you’re not recommending the cheapest option, write down the specific client objective that justifies it (RG 273.54).
4. Confirm. Before submitting, re-run pricing if time has passed or the market has moved, then run the lodgement audit. Treat the audit as a completeness check, not approval of your recommendation.

Scenario: the cheapest of nine isn’t the answer

A refinance client with meaningful savings tells you on a recorded call that cutting interest is the goal and that they keep a large cash balance. The escalation returns nine prices, and the lowest is a no-frills product with no offset. RG 273’s Example 1 describes a similar case, where a broker recommended only the lowest-rate, no-frills loan to a client with $150,000 in cash. ASIC’s commentary says it would expect an option with an offset account to have been presented. In your file, the transcript records the savings balance, the comparison records the nine prices, and your file note should explain why the second- or third-cheapest loan with an offset costs less overall. Remember RG 273.58 too: if refinancing costs outweigh the saving, staying put may be the better recommendation.

What to review this week

• Ask any pricing or automation vendor how the lender list is set, whether it has lender or aggregator approval, and what is logged.
• Decide where comparison outputs are stored and for how long. RG 273.168 says the retention period may depend on the loan term, any interest-only period and whether the client refinances.
• Update your file-note template with a line that starts: “Lowest-cost option was ___; recommended ___ because ___.”
• Write down how retention repricing is handled while automated tools cover new lending only.
• If you’re adopting call recording, confirm consent, storage and access arrangements with your licensee before the first call.

Conclusion

This month’s launches take real admin work off brokers’ desks, and escalation that used to take days now takes minutes. But under RG 273, faster access to more prices raises what your file needs to show. The brokers who get the most from these tools will save every comparison, explain every recommendation that isn’t the cheapest, and treat automated checks as evidence that supports their judgement rather than a replacement for it.

Frequently asked questions

Sources: Broker Daily, “Automated pricing tool for brokers launches”, 11 Sep 2026; Broker Daily, “Salestrekker launches AI-powered audit tool”, 11 Sep 2026; The Adviser, “YBR debuts new AI platform across broker network”, 14 Sep 2026; Australian Broker, citing Canstar, Sep 2026; ASIC RG 273 Mortgage brokers: Best interests duty (June 2020).