On this page
In a broking or advice practice, most of the paperwork sits between the conversation with the client and the lender or product provider. A broker’s support team collects payslips, bank statements and ID, checks them against the lender’s list and keys the application into the aggregator’s platform. An adviser’s client services team turns fact-finds and dictated notes into CRM records, builds the pack for each review meeting and follows up application forms. When the predictable steps run as fixed, tested software, the tracking, checks and record-keeping happen the same way on every file. AI is used only where something has to be read, such as a payslip or a voice note, and its output is checked before it reaches a lender, a provider or the client file. The advice and the credit assessment stay with the licensed person.
These are patterns we see in financial advice and mortgage broking businesses, not client case studies. Your process gets its own map in the assessment.
Loan documents checked before they reach the lender
A home loan application starts with a pile of documents. The client emails payslips, bank statements, ID and rates notices in whatever form they have, and broker support checks each one against the lender’s list, chases what is missing and keys the details into the application. AI reads each document as it arrives and pulls out the income, employer, dates and account balances. Fixed rules then build the checklist for that lender and product, compare what has arrived with what is needed, and chase the gaps on a schedule. Anything that does not match, such as an income that differs between a payslip and a bank statement, goes to the broker. The broker reviews the complete file before anything is lodged, and the credit assessment stays theirs.
Fact-finds and file notes into the CRM
After a meeting the adviser dictates a file note, and the fact-find comes back as a form, a PDF or a scan. Someone in client services retypes both into the CRM, field by field, and the gaps only show up when the file is reviewed. AI reads the voice note and the fact-find and drafts the CRM fields and the file note, but nothing is saved until the adviser has checked it. Once it is approved, the record is written, the follow-up tasks the note mentions are created, and any fact-find answers still missing are listed against the file. The note stays in the adviser’s words, and the record shows who approved it.
Loan applications through to settlement
Each loan moves through valuation, conditional approval, outstanding conditions, formal approval and settlement, and someone tracks every step across lender emails, the aggregator’s platform and a spreadsheet. Where your aggregator’s platform already tracks status, we build around it rather than duplicate it. The gap is usually the email. AI reads lender and valuer emails for status changes and new conditions, and fixed rules update the file, set due dates and remind the right person before a condition holds up settlement. Broker support sees one list of what is waiting on whom. Anything unusual, such as a valuation below the purchase price, goes straight to the broker.
In the weeks before settlement, broker support swaps emails with the lender, the conveyancer, the real estate agent and the client to confirm dates, documents and funds. The standard confirmations go out on a timetable keyed to the settlement date, each reply is recorded against the file, and anything still outstanding close to the date is flagged. AI reads the replies to pick out dates and requests. The broker or their support decides what to do about a delay or a changed date, and the client hears about it from a person.
Once a loan settles, the signed documents, approvals and correspondence have to be filed against the client and kept as your licence requires. It is easy to leave until later, and later becomes the week before an audit. Filing runs on rules with no AI involved. Each document is named and saved to the right client and loan folder, a completeness check runs against your file checklist, and anything missing goes back to broker support with the list attached.
Advice paperwork
Before each review meeting, someone pulls the client’s holdings, super and insurance details, last year’s notes and any changes since, and assembles them into the pack the adviser reads beforehand. The data pull and the layout follow the same template every time, so they run as fixed software from your planning platform and CRM. Where your planning software already produces review reports, we use them as the source rather than rebuild them. AI drafts a short summary of what has changed since the last review, the adviser edits it, and the pack is ready ahead of the meeting.
Insurance and super applications ask for the same client details the CRM already holds, on each provider’s own form. Client services retypes them, then chases signatures and the provider’s outstanding requirements. Each form is filled from the client record, sent for signature and tracked until the provider accepts it, with AI reading provider emails for new requirements. The adviser signs off every application before it goes, and any medical or underwriting question goes to the adviser.
Records and statements of advice are the adviser’s work, and the advice in them stays that way. The repetitive part is the assembly: client details, current holdings, fees and the standard sections your licensee requires, pulled into the template before the adviser writes the recommendations. That assembly runs as fixed software from the CRM and planning platform, and AI is not used to write advice. The adviser writes and signs the advice, and the document goes through your usual compliance check before the client sees it.
Commissions and the back office
Each month the aggregator’s commission statement lists upfront and trail payments, and someone matches it against the loans that settled, the loans discharged early and any clawbacks. Mismatches are easy to miss and slow to chase. We build this without AI. Each statement line is matched to the loan record on rules you set, upfront commissions are checked against settled loans, and clawbacks are matched to the discharged loan. Anything unmatched is listed with the loan and the amount for the office manager to query with the aggregator.
Client documents, lender updates, provider requirements and fee enquiries all arrive in one shared inbox. Someone reads each message, saves the attachments to the right client and forwards it on, and a message that waits becomes a delay on a loan or an application. AI classifies each email and matches it to the client and the loan or policy. Fixed rules then file the attachments, create the task and route it to broker support, client services or the adviser. Anything the classifier is unsure of goes to a review queue, so nothing is lost for being ambiguous.
The principal wants to see what settled, what is in the pipeline, which commissions have arrived and which fees are outstanding, and today someone builds that by hand from the aggregator, the CRM and the accounting system. The extraction, calculations and layout run as fixed software on the same day each month. AI drafts a few lines on what changed for the principal to edit. The figures come from your systems, never from AI.
Retyping between the CRM, the aggregator and lender portals
Client details are entered in the CRM, again in the aggregator’s platform and again on lender or provider forms, and a change made in one place rarely reaches the others. Each hop looks small, so nobody measures it as one process. Almost all of it is fixed software, with field mappings between systems, a scheduled sync, reconciliation checks and an exception queue for records that do not line up. Where a lender portal offers no way in except its screens, that step stays with your team rather than becoming something that breaks. We measure the whole chain as one process first, so the cost is visible before anything is built.
If one of these costs you real money every month, describe it in four answers.
Thinking about what a process actually costs you?
The assessment is small, fixed-price, and tells you the real number, whether or not you ever build.
Worth a 30-minute conversation.