How to set up AI that passes an aggregator or BID compliance review

BLUF: how to make AI pass an aggregator or BID review

You pass an aggregator or Best Interests Duty (BID) review when your file shows contemporaneous records, clear reasoning, and a human sign-off that maps to your aggregator’s checklist. AI helps you capture and structure that evidence. It does not remove your duty, you remain responsible. Build your AI around evidence-first capture, instruction logs, decision logic, and an audit pack your aggregator can read. That is the standard. That is how we configure @Briicky to work for brokers.

Two live facts set the bar in Australia. ASIC is actively reviewing brokers’ BID compliance and sampling files, including data requested from aggregators. The Adviser reported ASIC is “undertaking an information-gathering exercise on broker files” and working with some aggregators on how brokers meet BID. MPA Magazine likewise reported an “ongoing review” requesting extensive data from national aggregators and analysing broker file samples. In parallel, the Australian Energy Regulator’s August 2026 Compliance Bulletin on automated bidding states software does not shift responsibility away from market participants, and that contemporaneous records should include the instructions the program relied on, relevant triggers and thresholds, and standing instructions for any third-party provider. The Bulletin also indicates detailed system logs can satisfy record-keeping if they capture inputs, decision logic, and outputs for each decision. Translate that to broking and you have your AI compliance blueprint.

Sources, so you can show your compliance manager: see The Adviser’s coverage of ASIC’s information-gathering on broker files, MPA Magazine’s report on ASIC’s ongoing BID review, the AER August 2026 Compliance Bulletin on automated bidding, and the National Law Review’s summary of the AER position.

What “AI that passes review” actually means

Vendors love a slogan. Strip the marketing. There is no stamp that says an AI “passes” BID. There is the file you produce and the review your aggregator or ASIC conducts. Your file passes when your evidence meets the standard. AI can help you meet it every time by making the record complete, contemporaneous, and readable.

Across sectors, the pattern is the same. Regulators accept that software can draft, suggest, and act. They expect the licensed party to own the outcome, keep detailed logs, and show why a recommendation was in the client’s best interests at that time. That is exactly what the AER spells out for automated bidding: responsibility remains with the participant, and contemporaneous records need the instructions, triggers, thresholds, and outputs. Do that for BID and your AI will hold up under review.

Translate the regulator’s position to mortgage broking

Here is the translation for an Australian broker file, whether you sit with AFG, Connective, or another aggregator, and whether your daily system is Salestrekker, Mercury Nexus, or BrokerEngine:

The seven-part BID evidence framework for AI agents

If you want your AI to survive an aggregator or ASIC review, build this into your workflow. We implement these seven parts for every finance client.

Where AI helps, and where you stay on the hook

AI can do real work in a broker file. It can ask missing-Info questions by SMS, summarise a discovery call, assemble a comparison table from policy notes you have configured, pre-draft a BID statement, and prepare a credit proposal shell ready for you to enrich. It can keep you on top of follow-ups and move a client from intent to application without you re-keying data.

Here is the line. The AI agent does not provide credit assistance. It does not choose a product or make a recommendation without your review. It drafts. You review and decide. Your signature closes the loop. That is how we keep files safe under scrutiny while still giving you the time saving.

The toolchain reality: Salestrekker, Mercury Nexus, AFG, Connective, BrokerEngine, Xplan, Midwinter

Your systems are the backbone. The AI must work with how your team files. You do not need to rip anything out to get the compliance benefit.

Important: you choose how the AI interacts with your stack. Some teams prefer export-ready artefacts they paste into Salestrekker or Mercury Nexus. Others enable secure connections so notes flow in automatically. Either way, the evidence framework above is the same. Your compliance manager gets the same audit pack.

How we set this up at Briick

You do not need another dashboard. You need an operator that does the work, shows its thinking, and keeps you safe. That is @Briicky, the AI Chief of Staff in Briick. You talk or type to one thing. It orchestrates the other AI agents that sit on your front desk, SMS, email, and WhatsApp, and it builds the evidence pack as it goes.

This is done-for-you. Our team builds the Briick Workflow, configures your prompts and checklists, and aligns outputs to your aggregator. You never learn new software. You keep working in the tools you already use.

Example: a refinance file that stands up in review

Walk a refinance for Sarah and Tom, PAYG with a small side business. You start with a voice discovery. The AI voice agent records and tags their objectives: consolidate debt, reduce repayments, keep offset, retain redraw. It captures constraints: stable PAYG income, fluctuating side income, preference for major banks, aversion to break fees.

That afternoon, @Briicky prepares a comparison of three products you short-listed, using lender policy excerpts you have already approved as reference content. It does not choose. It lays out interest rates, fees, offset features, cashback time limits, and serviceability notes. Each line item links to the source document or your policy library and is timestamped.

Over two days, the SMS agent requests updated payslips, bank statements, and a credit consent. Each response is logged with time, content hash, and client identity. Missing living expense categories are flagged and collected with a guided form link. When the bank statement shows an afterpay-style commitment, the AI flags a potential conflict with the client’s stated expense estimate and prompts you to confirm the treatment. You decide and annotate. That decision rationale is versioned.

When you are ready, @Briicky drafts the BID note. It references the time-coded parts of the discovery call where the clients asked for an offset, the email where they confirmed their preference for a major bank, and the policy extract noting the lender’s treatment of side income. It sets out two alternatives you considered and why they were unsuitable at that time. You add a paragraph about potential refinance costs and break fees. You sign. The log shows your sign-off handover time.

For the aggregator, you export the AFG-aligned pack. The reviewer sees headings that match their checklist. Under each, the AI has placed the evidence, with links and timestamps. The pack includes the instruction log of the prompts and rules the AI used, so a reviewer can see there was no hidden instruction that could bias the result. The file passes. Not because AI got lucky, but because the evidence is complete and contemporaneous.

Risk controls and red flags to remove early

Here are the traps that trigger queries in review. Remove them in your workflow.

‍

Set-up plan: two weeks to an audit-ready AI workflow

This is the build we run for a finance team. Keep it simple, get to live, then optimise.

What your reviewer should see in your export pack

Make the reviewer’s job easy. Your pack should include:

Security and data sovereignty matter

AI that helps with compliance must keep client data safe and in your control. With Briick, your vectors live in your own secure data store so you keep your IP and client records. You decide which channels are in-bounds. You decide what gets retained, for how long, and who can export it. That is how you use AI without handing your book to a black box.

The honest bit on “pass rates”

You will see claims about pass rates. Ignore them. There is no authoritative statistic on the percentage of AI-assisted broker files that pass BID review. What we do have are the regulator positions and what your aggregator asks for. Build your workflow to that, and you will give a reviewer exactly what they need. That is the only number that matters.

Ready to run this with an operator, not another tool

If you want the AI Chief of Staff that runs this for your team, sets up the audit logs, and delivers aggregator-ready packs, that is @Briicky in Briick. Done-for-you. Voice-first discovery, SMS and email follow-ups, audit-ready BID notes, and your data in your own AI database. See the finance guide and talk to us.

FAQ

Does AI actually pass an aggregator or BID compliance review?

Files pass, not AI. You pass when your file shows contemporaneous records, clear decision logic, and your sign-off aligned to your aggregator’s checklist. AI helps you capture and structure that evidence. You remain responsible for compliance.

What records do I need if AI helps write my BID notes?

Keep the instruction log, inputs, decision logic, and outputs. That means prompts and standing instructions, time-stamped client facts, versioned reasoning, and the final signed note. The AER’s guidance for automated systems highlights this model: instructions, triggers, thresholds, and outputs recorded contemporaneously. It is a useful benchmark for broker AI record-keeping.

Who is responsible if AI drafts my BID note?

You are. The regulator position in other sectors is clear, and it applies in spirit here. Responsibility remains with the licensed participant even when software or a third party is involved. Use human approval gates and a clear sign-off to make that responsibility visible in your file.

How do I set this up safely if I use Salestrekker or Mercury Nexus?

Keep your current system. Configure your AI agents to produce export-ready BID notes and evidence summaries that mirror your fields and aggregator headings. Store the audit log alongside the deal. If you later enable a connection, keep the same evidence framework and permissions model.

Can @Briicky join client calls and prepare BID notes?

Yes. A voice AI agent can record and tag discovery calls, then @Briicky assembles a first-draft BID note with sources linked. You review and sign. The pack exports in your aggregator’s format.

Will ASIC accept AI-generated logs?

ASIC has not issued a template for AI logs. Their current activity focuses on broker BID compliance via file sampling and aggregator data. The AER’s model in energy shows that detailed, contemporaneous system logs that capture inputs, instructions, logic, and outputs can satisfy record-keeping. Build to that standard and you will be ready for scrutiny.

Does this replace my compliance team?

No. It gives them better files. Your compliance manager gets a pack that mirrors aggregator headings, with sources and timestamps. Reviews are faster because the evidence is already organised.

What about privacy and client consent?

Use approved consent wording for recordings and data processing, sent via your AI agents with timestamps. Keep your data in your own secure AI database. Limit channels to what your policy allows.