Artificial intelligence (AI) for mortgage brokers: capture the fact-find, screen lender criteria and package the case faster
See how AI can be applied to the real processes of a UK mortgage brokerage: the enquiry and its qualification, the fact-find and document capture, lender-criteria research and case packaging, borrower communication and protection prompts, retention at rate expiry and the compliance file, and the practice's own fees. Intermediaries now arrange close to 89% of all UK mortgage lending, so the channel has won and the constraint is internal: how fast an enquiry becomes a packaged, submitted case. Each process comes with examples and an honest view of the technology. The golden rule runs through all of it: a UK mortgage sale is an advised sale, so you own the recommendation and the suitability, and AI captures, screens, packages and chases without ever advising.
Lead intake and qualification: every enquiry triaged and ready before you pick it up
Enquiries do not arrive on your schedule. They land from portals, referrals, social and your own website at all hours, and the broker who qualifies and responds first usually wins the case. With intermediaries now arranging close to 89% of UK mortgage lending, the channel has been won; the race is on speed of response, not on whether a borrower reaches a broker at all.
The problem is that a first-time buyer messaging at nine in the evening, or a remortgage enquiry over the weekend, sits unanswered until someone is free. By then the same person has often filled in two more forms on two more websites. A slow first reply is a lost lead, and it is lost quietly.
An assistant built for your firm can take that first contact the moment it comes in. It reads the enquiry, asks the two or three qualifying questions that matter, gathers the basics into a structured brief and hands the adviser a lead that is already triaged and ready to progress, rather than a cold form to start from scratch.
The line is drawn from the first message. The assistant captures facts and sets expectations; it never tells a borrower what they can borrow, never recommends a product and never implies an agreement in principle, because that is regulated advice and a lending decision, and both sit with a person and a lender, not with software.
Qualification gathers information and manages expectations calmly. It must never steer someone toward a fast outcome to hit a target, which is exactly the high-pressure selling the FCA has told mortgage intermediaries it is watching.
A structured first response on every channel, day or night
The assistant sits on the channels enquiries actually arrive on, the website form, email, WhatsApp and social, and replies in plain English within seconds at any hour. It works out whether this is a purchase, a remortgage, a product transfer or a buy-to-let enquiry, asks the handful of qualifying questions that decide next steps, and confirms what happens next. It books nothing binding and quotes no figure; it captures and organises.
At 9:40 on a Sunday evening a first-time buyer messages that they have found a flat and want to know what to do next. The assistant welcomes them, confirms it is the firm's AI assistant, asks about deposit, employment and timescale, and tells them an adviser will pick this up first thing. On Monday the adviser opens a complete brief instead of a one-line form.
The out-of-hours enquiry that used to leak to the next website in the search results now lands in your pipeline, triaged and ready. Your first response is instant and consistent, and the adviser starts warm.
Qualifying questions that build the brief, not a sales script
The assistant asks only what the adviser needs to progress the case: deposit and its source, rough income and employment type, whether it is a purchase or a refinance, the target property and the timescale. It records the answers into a structured lead brief, notes anything unusual for the adviser to probe, and stops short of any assessment. There is no pressure and no push toward a quick decision.
A self-employed enquirer mentions a recent contract change and a small deposit. The assistant captures both, tags the case as one that needs a careful income conversation, and routes it to the adviser with those flags visible, rather than trying to reassure or qualify the borrower in or out itself.
The adviser sits down to an already-triaged, ready-to-progress lead with the awkward points already surfaced. The qualification informs the advice; it never pretends to be it.
Expectations set without a figure or a promise
Borrowers almost always open with how much can we borrow and how fast can we get an agreement in principle. The assistant explains the process in general terms, that an AIP is an early indication from a lender and not an offer or a guarantee, and that the borrowing figure and the recommendation come from the adviser and the lender. It gathers the basics that make that conversation quick, without pre-empting its outcome.
An enquirer asks for a borrowing figure up front. The assistant explains how affordability and the AIP work, notes that only a lender issues an AIP and only after its own checks, and books the adviser call, so the borrower feels answered and correctly informed rather than fobbed off or misled.
The borrower gets a clear, honest first answer that manages expectations, and the adviser inherits a lead that already understands the process. Nothing regulated has been said in the firm's name by a machine.
Clean handover with a human check on every qualified lead
Each qualified enquiry passes to the adviser as a tidy brief: who they are, what they want, the qualifying answers, the flags and the full conversation thread. The assistant triages and prioritises, but a person reviews every lead before it enters the advice process. Automation sorts the inbox; the adviser decides what happens to each case.
Monday morning, the adviser sees eight overnight enquiries already sorted into ready-to-book, needs-more-information and not-yet-in-scope, each with its thread attached. They spend the morning advising the ready cases rather than reading raw forms and working out who is who.
The triage load lifts off the adviser without automation ever making the judgement call. Every lead reaches the advice process through a human, which is exactly where the responsibility sits.
Triaging and qualifying an enquiry is a realistic build today, and the commercial case for speed is well documented.
- With intermediaries arranging close to 89% of UK mortgage lending, and the share still climbing toward 91%, almost every borrower now reaches a lender through a broker. The channel has won, so the constraint is your own throughput: how fast an enquiry becomes a qualified, ready-to-progress lead. An assistant that reads an enquiry, asks the qualifying questions and drafts a structured brief can be connected to your channels now.Mortgage Solutions / IMLA
- Qualification has to stay on the right side of a hard regulatory line. The FCA's 30 January 2025 portfolio letter makes the Consumer Duty, the quality of advice, conflicts of interest and high-pressure selling its supervisory priorities for mortgage intermediaries. That is precisely why a good qualifier gathers facts and sets expectations, and never nudges a borrower toward a fast sale to hit a target.FCA
- Read any headline adoption figure as direction, not a result you can bank. Surveys report that a large majority of brokers already use general AI tools for admin, yet only a small minority let AI anywhere near the advice itself, which is the correct instinct and the model this build follows. The number that matters is your own: how much faster you respond first, and how many more qualified leads reach an adviser.Mortgage Solutions
Qualification touches regulated territory from the first message, so these limits are designed in, not bolted on.
- The assistant can triage and qualify, but the moment it recommends a product or a specific borrowing amount it is giving regulated mortgage advice. Under FCA MCOB 4.7A that is an advised sale only a qualified adviser can make, and the adviser owns the recommendation and the suitability that goes with it.FCA Handbook
- An automated qualifier must never imply an agreement in principle or a lending decision. An AIP is only an early indication of what a lender might advance, not an offer, and only the lender issues it after its own checks. Stating or hinting at a figure would mislead the borrower and cross into the lender's territory.MoneyHelper
- Enquiry data, income, deposit and contact details, is personal data under the UK GDPR. You need a lawful basis and clear privacy information before that data goes into any AI tool, and the ICO can fine up to 17.5m pounds or 4% of worldwide turnover.ICO
- Keep a human check on every qualified lead before it enters the advice process. Automation triages and prioritises; the adviser decides. A qualifier that quietly pushed borrowers toward a quick outcome would create the high-pressure-selling and conflict-of-interest risk the FCA has singled out.FCA
Fact-find capture and document collection: the evidence base assembled, without the week of email tennis
The fact-find is where an advised case is won or lost. Under MCOB 4.7A the suitability of your recommendation is judged on the facts the customer discloses and the other relevant facts you are or should be aware of, so a thin or inconsistent fact-find is where cases go wrong later. It is the evidence base the whole recommendation stands on.
Gathering that evidence is slow and repetitive. Payslips, bank statements, SA302s, ID and proof of address get chased over email for a week, one missing item at a time, while the borrower sends a blurry passport photo and half of a self-employed income picture. With house-purchase lending up 22% in 2025, each adviser is collecting and checking more of this than ever.
An assistant built for your firm can structure the intake and carry the chase. It requests the right documents for each borrower's circumstances, reads income and outgoings from what they upload, spots gaps and inconsistencies before the case moves on, and keeps a clean record of what has and has not arrived.
What it extracts, the adviser verifies. AI can pull the figure off a payslip, but suitability rests on facts the firm is or should be aware of, so a person confirms the numbers before they carry any weight. Automated reading of scans and non-standard or foreign documents can misfire, which is another reason identity and income are checked by a human before the case is submitted.
And it never gives a verdict. Affordability is ultimately the lender's assessment under MCOB 11, so a document tool that told a borrower they can afford it would be presenting a lending decision it has no authority to make.
The right document list for each borrower's circumstances
From the fact-find, the assistant works out which evidence this particular case needs, employed payslips and P60, or self-employed SA302s and tax year overviews, plus ID, proof of address and proof of deposit, and requests them through a clear, secure checklist. It tailors the list to the borrower rather than sending a generic form, and marks each item as outstanding or received as things arrive.
For an employed applicant with a self-employed partner, the assistant asks the employed side for three payslips and the self-employed side for two years of SA302s and tax year overviews, and explains why each is needed. The borrower stops guessing what to send, and the adviser is not chasing the missing half a week later.
The document request is right first time and specific to the case, so fewer rounds of back-and-forth. The evidence base for the recommendation comes together in days rather than a fortnight of email tennis.
Income and outgoings read from the documents, then verified by you
As payslips, bank statements and SA302s come in, the assistant reads the figures, income, regular commitments and any unusual outgoings, and lays them out against the fact-find so the picture is consistent and complete. Every extracted figure is presented for the adviser to confirm, never treated as fact, because suitability rests on the facts the firm is or should be aware of.
The assistant extracts net pay from three payslips and flags that one month includes a large bonus that does not repeat, so the adviser checks how the lender will treat it rather than discovering the anomaly at submission. The adviser confirms the figures; the assistant simply saved the retyping and surfaced the point.
The tedious transcription is done in the background while the judgement, what the figures mean and how a lender will read them, stays with the adviser. The fact-find arrives already cross-checked against the evidence.
Gaps and inconsistencies caught before the case moves on
The assistant compares what the borrower said in the fact-find against what the documents show, and flags the mismatches: a stated income that the payslips do not support, an address that does not match, a bank statement missing a month, an undeclared commitment showing on the statements. Each flag comes with the reason it was raised, for the adviser to resolve.
A bank statement shows a regular loan repayment the borrower did not mention. The assistant flags it against the fact-find, so the adviser addresses the commitment before choosing a lender, rather than having it surface in the lender's own checks and stall the case.
The thin or inconsistent fact-find that causes trouble later is caught early, while it is still a two-minute conversation. The adviser builds the recommendation on evidence that has already been reconciled.
Identity and income checked by a person, never presented as a verdict
The assistant reads and organises identity and income evidence, but a human verifies it before the case is submitted, because automated reading of scans, foreign or non-standard documents can misread them. Crucially, the tool assembles the affordability evidence and never presents a result, since affordability is the lender's assessment under MCOB 11, not the broker's and certainly not the software's.
A borrower asks whether their income is enough to get the mortgage. The assistant confirms it has all the income evidence in order for the adviser, but explains it cannot judge affordability, because that decision belongs to the lender. The adviser then has a complete, verified pack to work from.
You get a complete, ordered evidence base without automation ever overstepping into a lending decision. Identity and income are confirmed by a person, so the case is submitted on facts the firm can stand behind.
Capturing the fact-find and collecting documents is a strong fit for automation now, because so much of it is structured, repeatable work.
- Under MCOB 4.7A, suitability is judged on the facts the customer discloses and the other relevant facts the firm is or should be aware of, so the fact-find is the evidence base for the whole recommendation. An assistant can structure the intake, chase missing answers and read income and outgoings from uploaded documents, provided the adviser confirms the facts before they carry any weight. That build is realistic today.FCA Handbook
- The volume behind this is rising. UK Finance describes document handling as the mortgage bottleneck AI is starting to unblock, from reading statements to spotting gaps before submission, and house-purchase lending rose 22% in 2025. More cases per adviser means more ID, payslips and statements to collect and check, which is exactly the repetitive load an assistant can carry so advisers stay on the parts that need judgement.UK Finance
- Keep one expectation in check. Affordability is ultimately the lender's assessment under MCOB 11, even though the broker gathers most of the evidence the lender relies on. The tool assembles payslips, bank statements and SA302s so the case is complete; it must not present an affordability verdict to the borrower as if it were a lending decision.FCA Handbook
The fact-find holds a borrower's most sensitive data and underpins the whole recommendation, so these limits come first.
- Fact-find data is sensitive financial personal data. Minimise what you feed into AI tools, secure it, and run a data protection impact assessment for higher-risk processing under the UK GDPR. The ICO can fine up to 17.5m pounds or 4% of worldwide turnover, so this is a governance decision, not an afterthought.ICO
- AI can extract figures from payslips and statements, but the adviser must verify them, because suitability under MCOB 4.7A rests on the facts the firm is or should be aware of. A wrong extraction is no excuse for an unsuitable recommendation, so a person confirms the numbers before they carry any weight.FCA Handbook
- Affordability is assessed by the lender under MCOB 11. Do not let a document tool present an affordability result to the borrower as a lending decision; it assembles the evidence, and the verdict belongs to the lender after its own checks.FCA Handbook
- Automated document checks can misread scans, or foreign and non-standard documents, so keep human verification of identity and income before the case is submitted. Retain the fact-find and its evidence as part of the suitability record for the file, because that record is what the advised sale stands on later.
Lender criteria research: a fast shortlist for the adviser, who still owns the recommendation
Matching a case to lenders that will actually accept it is core research work, and it is fiddly. Criteria shift constantly by loan-to-value, employment type, adverse credit and income mix, so a scenario like self-employed with one missed payment at 85% loan-to-value has a shortlist that changes month to month. Getting it wrong wastes an application and the borrower's time.
This is back-office research support for the adviser, not client-facing advice. An assistant can screen lenders whose published criteria fit a scenario far faster than manual lookup, and lay the options out clearly, leaving the adviser to apply judgement and make the recommendation.
It matters most at renewal, where the money is. Internal product transfers reached 256bn pounds in 2025 against 71bn pounds of external remortgaging, so criteria research often decides whether switching lender genuinely beats the existing lender's retention deal. A good recommendation depends on comparing the new lender's criteria and pricing against staying put, and the assistant can gather and lay out that comparison.
The line is firm. Criteria screening narrows the field; presenting a single best product to the borrower is a personal recommendation and a regulated advised sale under MCOB 4.7A. The adviser weighs the shortlist and recommends; the assistant never does.
And it is only as current as its data. An assistant working from stale criteria can surface a withdrawn or mispriced product, so the shortlist is verified against the lender's live criteria before anyone relies on it, and it must never drift into an affordability or lending decision, which belongs to the lender under MCOB 11.
A criteria shortlist built from the case in minutes
The adviser describes the scenario, loan-to-value, employment type, any adverse credit, the income mix, and the assistant screens lenders whose published criteria fit and lists them with the relevant criteria points against each. It works as research support behind the adviser, not a client-facing tool, and surfaces the matches far faster than manual lookup so the adviser starts from a shortlist rather than a blank page.
For a self-employed applicant with one missed credit card payment fourteen months ago at 85% loan-to-value, the assistant returns the lenders whose criteria plausibly accept that profile, with the adverse-credit and self-employed rules noted for each. The adviser reviews a focused list instead of trawling lender by lender.
Hours of manual criteria lookup collapse into minutes of review. The adviser spends their time on the judgement and the recommendation, not on assembling the longlist by hand.
The stay-or-switch comparison laid out at renewal
For a remortgage or product transfer, the assistant gathers the existing lender's retention options alongside the criteria and pricing of realistic alternative lenders, and lays the comparison out for the adviser. It presents the trade-offs, criteria fit, rate, fees and any criteria hurdles to switching, without drawing a conclusion, because whether switching beats staying is the adviser's call.
A client's two-year fix is ending and their current lender has offered a product transfer. The assistant lays out that retention deal next to two external options the client's profile fits, with the pros and cons of each, so the adviser can weigh a genuine comparison rather than reconstruct it from scratch.
With product transfers dwarfing external remortgages, the renewal comparison is where advice adds the most value, and it arrives ready to weigh. The adviser recommends from a clear picture; the assistant never picks the winner.
Live criteria verification before anything is relied on
Because lender criteria change constantly, the assistant treats its shortlist as provisional and flags each match to be confirmed against the lender's current criteria before the adviser relies on it. It surfaces when a rule looks borderline or a product may have moved, so a withdrawn or mispriced option is caught in research rather than in a declined application.
The assistant shortlists a lender but flags that its self-employed policy was updated recently, prompting the adviser to check the live criteria before proceeding. What could have been a wasted application is caught while it is still a research note.
The risk of building a case on stale criteria is contained. The adviser proceeds on verified, current rules, and the borrower is not put through an application that was never going to fit.
Research that stays research, never an affordability or lending call
The assistant screens published criteria and stops there. It does not calculate a lending decision or an affordability verdict, which belong to the lender under MCOB 11, and it does not present a single best product to the borrower, which would be a regulated recommendation. Every shortlist is framed as options for the adviser to assess, matched to the customer's needs rather than to the procuration fee.
The adviser asks for a recommendation outright. The assistant declines to pick, restates the shortlist and the criteria trade-offs, and leaves the choice, and the suitability assessment, with the adviser. The conflict-of-interest risk of optimising for fee over fit never arises.
You get the speed of automated criteria research without any drift into regulated advice or a lending decision. The recommendation and its suitability stay firmly with the adviser, where the FCA expects them.
Criteria research is a natural fit for automation now, precisely because it is structured lookup work that sits behind the adviser.
- Lender criteria vary constantly by loan-to-value, employment type, adverse credit and income mix, so matching a case to lenders that will actually accept it is core research work. An assistant can shortlist lenders whose published criteria fit a scenario, for example self-employed with one missed payment at 85% loan-to-value, far faster than manual lookup, leaving the adviser to apply judgement and make the recommendation. That is a buildable system today.FCA Handbook
- The renewal book is where this pays off. Product transfers reached 256bn pounds in 2025 against 71bn pounds of external remortgaging, so criteria research often decides whether switching lender genuinely beats staying put. A good remortgage recommendation depends on comparing a new lender's criteria and pricing against the existing lender's retention deal, and AI can gather and lay out that comparison for the adviser to weigh.UK Finance via Mortgage Solutions
- The regulatory direction reinforces the model rather than replacing it. The FCA Mortgage Rule Review is simplifying parts of the rulebook, but the Association of Mortgage Intermediaries warns proposals must not sideline professional advice, and the advised model remains central to good outcomes under the Consumer Duty. Criteria research supports the adviser; it does not replace the recommendation. Any adoption percentage you read is direction, so prove it on your own research time and application success rate.FCA
Criteria research sits close to the advice line, so where it stops matters as much as what it does.
- Criteria screening is research support for the adviser. Presenting a single best product to the borrower is a personal recommendation and a regulated advised sale under MCOB 4.7A, so the assistant shortlists and the adviser recommends and owns the suitability.FCA Handbook
- Lender criteria change constantly, so an assistant working from stale data can surface withdrawn or mispriced products. Verify every shortlisted match against the lender's live criteria before relying on it, and treat the AI list as a starting point, not a conclusion.
- Do not let criteria automation drift into an affordability or lending decision, which belongs to the lender under MCOB 11. Screening published rules is not the same as deciding what a borrower can afford, and the tool must not blur the two.FCA Handbook
- The Consumer Duty requires products matched to the customer's needs, so a shortlist that optimises for procuration fee over suitability is a conflict of interest the FCA has said it is watching. The research must serve the borrower's circumstances, not the firm's income.FCA
Case packaging and submission prep: a clean, complete pack, with your sign-off on every case
Clean, complete packaging is what turns a decision in principle into a mortgage offer quickly. Missing or inconsistent documents are the main cause of delay and rework, and every gap the assembler misses is a query, a re-request and a slower case. UK Finance describes packaging and document handling as the bottleneck AI is starting to clear.
An assistant built for your firm can assemble the pack, ID, proof of deposit, income evidence, the fact-find and the suitability note, cross-check it for completeness and consistency, and flag exactly what is still missing before you submit. It does the assembly; you do the sign-off.
The volume makes it worth systematising. With 290.8bn pounds of gross lending in 2025 flowing largely through small intermediary teams, submission prep is high-frequency, repetitive work, and standardising the pack so nothing is missed frees the adviser's time for advice rather than assembling PDFs.
Every advised case must also carry a record that demonstrates why the recommendation is suitable. MCOB 4.7A requires the firm to base and be able to show suitability on the customer's facts and circumstances, so the suitability note travels with the case. The assistant can draft that note from the fact-find, but it must reflect the adviser's actual reasoning, and the adviser reads and owns it before it goes.
The hard rule sits at the end: automation assembles and quality-checks, but it never submits a case or accepts an offer without the adviser's sign-off. The pack is the assistant's; the recommendation and the decision to submit are the adviser's.
The submission pack assembled to a standard shape
The assistant gathers every element the lender needs into one ordered pack, ID, proof of deposit, income evidence, the fact-find and the suitability note, in the same structure every time. Instead of the adviser hunting through email threads and folders, the pack is built to a standard template, so what should be present is obvious and what is absent stands out.
For a case ready to go to a lender, the assistant lays out the ID and proof of address for both applicants, the SA302s, the proof of deposit, the fact-find and the draft suitability note in the firm's standard order. The adviser opens a complete, consistent pack rather than assembling it under time pressure.
Packaging that used to eat an afternoon per case is done in the background, in a consistent shape every time. The adviser reviews a standard pack instead of reconstructing one from scattered files.
A completeness and consistency check before submission
Before the case goes, the assistant checks the pack against itself: are the names, addresses and figures consistent across the documents, is any required item missing, does the income evidence match the fact-find, is anything internally contradictory. Each flag comes with the reason it was raised, so the adviser resolves a short list of concrete points rather than re-reading the whole file blind.
The assistant flags that the deposit shown on the fact-find does not match the amount evidenced in the bank statement, and that one applicant's proof of address is out of date. Both are fixed before submission, so what would have been a lender query is dealt with while it is still a two-minute correction.
The missing or inconsistent documents that cause most delay and rework are caught before the case is submitted, not after. Cases convert from decision in principle to offer faster because the pack goes in clean.
A suitability note drafted from the fact-find, owned by the adviser
The assistant drafts the suitability note from the fact-find and the case, setting out the customer's circumstances and needs in a clear structure. It is a first draft to save typing, never a finished record: the adviser edits it so it reflects their actual reasoning, and reads and owns it before it joins the pack, because MCOB 4.7A requires the firm to be able to show suitability on the customer's facts.
For a remortgage case the assistant drafts a note summarising the client's circumstances, needs and the reasoning captured in the fact-find. The adviser reworks it to match the advice they actually gave, then signs it off, rather than writing the whole note from a blank page late in the day.
The suitability note starts from a structured draft instead of a blank page, saving the adviser time on the write-up. The reasoning and the responsibility stay the adviser's, exactly as the FCA expects.
A chase-list of what is still outstanding, with sign-off held for you
The assistant produces a clear list of anything still needed before submission, a missing document, an inconsistency to resolve, a signature outstanding, and tracks it as items come back. It never submits the case or accepts an offer itself; it prepares everything and holds at the sign-off, so the decision to submit is always a deliberate human act.
The assistant reports that the pack is complete bar one outstanding proof of deposit and a suitability note awaiting the adviser's final read. Once both are in and the adviser has signed off, the adviser submits. The assistant assembled and checked; it did not press send.
Nothing is submitted on autopilot, and nothing goes in half-complete. The adviser reaches submission with a finished, checked pack and makes the final call, keeping the recommendation and the decision where responsibility sits.
Submission prep is a strong candidate for automation now, because it is high-volume, repetitive assembly work with a clear human sign-off at the end.
- Clean, complete case packaging is what turns a decision in principle into a mortgage offer quickly, and missing or inconsistent documents are the main cause of delay and rework. UK Finance describes packaging and document handling as the bottleneck AI is starting to clear, so an assistant that assembles the pack and flags what is missing before the adviser submits is a realistic build today.UK Finance
- The suitability note is part of the pack, and it has a hard requirement attached. MCOB 4.7A requires the firm to base and be able to show suitability on the customer's facts and circumstances, so the note travels with the case. AI can draft it from the fact-find, but it must reflect the adviser's actual reasoning, and the adviser reads and owns it before submission.FCA Handbook
- The volume is what makes standardising it worthwhile. With 290.8bn pounds of gross lending in 2025 flowing largely through small intermediary teams, submission prep is high-frequency, repetitive work ripe for automation with human sign-off. Any productivity figure a tool quotes comes from another firm, so treat it as direction and measure your own packaging time and clean-submission rate before and after.UK Finance via Mortgage Solutions
Packaging ends at submission, which is exactly where the human sign-off must be non-negotiable.
- AI can assemble and quality-check a submission pack, but the adviser owns the recommendation and the suitability evidence that goes with it under MCOB 4.7A. The assistant prepares; the adviser is accountable for what is submitted.FCA Handbook
- Borrower documents in a case pack are personal data. Control which people and systems can access them, and keep them within your lawful basis and retention policy under the UK GDPR, rather than leaving a full financial pack sitting in an inbox indefinitely.ICO
- A generated suitability note must reflect the adviser's real reasoning, not an unread template. The FCA expects a genuine assessment of the customer's circumstances, so a note that only looks complete will not satisfy a Consumer Duty review; the adviser edits and owns it.FCA
- Never let automation submit a case or accept an offer without adviser sign-off. Assembly and checking can be automated; the decision to submit is a deliberate human act, and it stays that way.
Borrower communication and chasing: consistent, clear updates that keep cases moving
A case rarely stalls because of the advice. It stalls waiting on documents, valuations and solicitors, and chasing all of it is one of the biggest time sinks in a brokerage. UK Finance frames much of this progress-chasing as low-value work AI can absorb, freeing the adviser for the parts that need judgement.
An assistant built for your firm can carry the routine touchpoints. It sends timely, plain-English status updates and reminders to the borrower, the solicitor and the estate agent, keeps everyone informed of where the case actually is, and flags a genuinely stuck case to the adviser, without the adviser writing every message by hand.
The volume makes this valuable. With around 1.8 million fixed rates expiring in 2026 on top of a busy purchase market, message volume per adviser is high, and borrowers who are kept informed complain less and drop out less. Consistent, prompt updates protect both the client experience and the adviser's focus.
Every message has a bar to clear. Under the Consumer Duty communications must be clear, fair and not misleading, and must support customers to understand and act. Automated messaging helps you communicate consistently; the content must genuinely help, and anything that strays into advice, a lending decision or a firm timescale belongs with the adviser.
And it has to be right about the case. Automated chasing can misfire, wrong client, outdated status, so the assistant checks the live case state before a message goes out, and it flags any sign of vulnerability to a human rather than automating around it.
Plain-English status updates that keep the borrower informed
Drawing on the live case state, the assistant sends the borrower clear updates on where things stand, valuation booked, documents received, waiting on the solicitor, in plain English and at the right moments. It tells them what is happening and what, if anything, is needed from them, without the adviser drafting each message, and without straying into advice or a promised date.
When the valuation is instructed, the assistant messages the borrower to say so, explains the next step is the lender's assessment, and confirms nothing is needed from them right now. The anxious what-is-happening question is answered before it is even asked, and the adviser has not had to write it.
Borrowers stay informed and reassured through the quiet stretches of a case, so they worry less and chase less. The routine updates lift off the adviser while the client experience improves.
Reminders and chasing across borrower, solicitor and agent
The assistant tracks what each party owes at each stage and sends timely reminders, to the borrower for an outstanding document, to the solicitor or estate agent for a status, so nothing sits waiting because no one chased it. It keeps a record of what was sent and when, and escalates to the adviser when a party goes quiet or a case is genuinely stuck.
A case has been waiting three days on a solicitor's update. The assistant sends a polite chase, logs it, and when there is still no reply flags the case to the adviser with the full history, so the adviser steps in on a real blockage rather than discovering the delay a week later.
The progress-chasing that eats hours of adviser time is carried in the background, and stuck cases surface early. The adviser's attention goes to the blockages that actually need a person.
Communications that clear the Consumer Duty bar
Every message is written to be clear, fair and not misleading and to help the customer understand and act, in line with the Consumer Duty. The assistant sticks to factual status and next steps, avoids jargon, and is built so that anything regulated, advice, a lending decision, a guaranteed timescale, is routed to the adviser rather than said in an automated message.
A borrower asks in reply whether their mortgage is now definitely approved. The assistant explains where the case is factually, does not claim a decision the lender has not made, and passes the question to the adviser, so the borrower gets an honest answer and nothing misleading goes out under the firm's name.
You communicate consistently and promptly without raising Consumer Duty risk. The routine messaging is automated; the regulated content and the reassurance that needs judgement stay with the adviser.
Live-state checks and vulnerability flags before anything sends
Before a message goes out, the assistant checks the current case state so it never sends an outdated or wrong-client update, and it watches for signs a borrower may be vulnerable, distress, confusion, difficulty coping, and flags those to a person rather than handling them automatically. The Consumer Duty requires extra care for vulnerable customers, and that care is a human's.
A borrower replies to a chase sounding distressed about the risk of losing the property. The assistant does not send the next scripted reminder; it flags the message to the adviser to handle personally, with the thread attached, so a person responds with the care the situation needs.
Automated chasing does not misfire on stale data or steamroll a worried client. The routine runs on verified case state, and the moments that need human sensitivity reach a human quickly.
Progress-chasing and status updates are a natural fit for automation now, because they are high-volume, routine touchpoints rather than judgement calls.
- Cases stall waiting on documents, valuations and solicitors, and chasing all of it is one of the biggest time sinks in a brokerage. UK Finance frames much of this progress-chasing as low-value work AI can absorb, so an assistant that sends timely status updates and reminders to the borrower, solicitor and estate agent, and flags a stuck case to the adviser, is a realistic build today.UK Finance
- The volume that makes it worthwhile is measurable. With around 1.8 million fixed rates expiring in 2026 on top of a busy purchase market, message volume per adviser is high, and borrowers who are kept informed complain less and drop out less. Automating the routine touchpoints protects both the client experience and the adviser's focus, and any adoption figure you read is direction, so measure your own case cycle time and drop-out.UK Finance
- The bar on the content is set by regulation. Under the Consumer Duty, communications must be clear, fair and not misleading and must support customers to understand and act, which raises the standard on every status update and reminder. Automated messaging helps you communicate consistently and promptly, but the content must genuinely help the customer, and anything that strays into advice or a firm timescale belongs with the adviser.FCA
Automated messaging speaks to the borrower in the firm's name, so these limits are built in from the start.
- AI can draft and send status updates and reminders, but communications must be clear, fair and not misleading and must support customer understanding under the Consumer Duty. A message that confused or over-promised would breach that standard however it was generated.FCA
- Do not let an automated message give a lending decision, a guaranteed timescale or advice. That content is regulated and stays with the adviser under MCOB 4.7A, so the assistant reports factual status and routes anything regulated to a person.FCA Handbook
- Client contact details and case data are personal data under the UK GDPR. Secure the channels and keep an audit trail of what was sent, and remember that automated chasing can misfire on the wrong client or an outdated status, so check the live case state before any message goes out.ICO
- The Consumer Duty requires extra care for vulnerable customers, so flag signs of vulnerability to a human rather than automating around them. A distressed or confused borrower needs a person, not the next scripted reminder.FCA
Protection and GI prompts: the cover conversation raised at the right moment, never advised by the machine
The mortgage is only half of what a borrower needs, and it is the half everyone remembers to sort. The protection that keeps the roof over their head if income stops is the half that quietly gets skipped, usually because nobody raised it at the moment it mattered. More than 40% of mortgage holders have no life insurance at all, and take-up of income protection and critical illness cover is lower still. That is a real customer need going unmet, not a missed sale.
You are the natural person to close that gap, because you are already sitting across from the client discussing the mortgage. The Association of Mortgage Intermediaries frames the protection gap as an awareness problem more than an access problem, with brokers central to the solution. The trouble is timing and memory: at offer stage the case is busy, the client is elated, and the protection conversation slips off the end of a long to-do list.
An assistant built for your firm treats that as a workflow, not a hope. It watches for the offer, checks whether any cover is recorded against the client, and prompts you to open the conversation while the moment is right. It drafts a plain-English opener from what the fact-find already tells you, so you start warm rather than cold.
The line here is bright and it stays bright. Advising on protection or general insurance is a separate regulated activity, so the assistant reminds you to raise it, it never advises the client. The recommendation, and any suitability assessment for a protection or GI product, comes from a suitably authorised adviser and follows the same advised-sale discipline as the mortgage. The nudge serves the customer's need; it never becomes a push toward commission.
Every offer flagged for the protection conversation
As a case reaches offer, the assistant checks whether any life, income protection or critical illness cover is recorded against the client. Where the file shows none, it raises an internal prompt for you to open the conversation, with the client's dependants and mortgage size drawn from the fact-find so the reminder has context. It flags; it does not contact the client or advise.
The Bakers' residential offer comes through with two children on the fact-find and no protection recorded anywhere. The assistant surfaces the case on your review list with a one-line note that a protection discussion is outstanding, so it happens this week rather than being remembered, or not, three months on.
The conversation that most often gets skipped now gets raised, at the one moment the client is most receptive. A genuine customer need is met, and it is met because the workflow prompted you, not because anyone had to hold every offer in their head.
A tailored opener drafted for you, not advice sent to the client
From the facts already captured, the mortgage amount, the term, dependants, employment, the assistant drafts a short, plain-English opener you can use to introduce protection, framed around the client's own situation. It is a talking point for the adviser to deliver and adapt, never a recommendation, a quote or a product the assistant pushes at the client itself.
For a self-employed client with a young family and a 25-year term, the assistant drafts an opener noting that, with no employer sick pay behind them, it is worth reviewing what happens to the mortgage if illness stopped their income. You read it, make it your own and raise it in the review.
You start the protection conversation warm, with something relevant to the specific client, instead of reaching for the same generic line. The judgement and the advice stay entirely yours; the assistant just removes the blank page.
Declines logged and diarised for a fair later review
When a client declines cover, the assistant records that it was offered and declined, and diarises a proportionate later check-in rather than dropping it forever. Circumstances change, a new child, a new job, a bigger mortgage, and the assistant surfaces the case again at a sensible point, so a one-off no does not become a permanent gap nobody revisits.
A first-time buyer declines protection because money is tight after the deposit. The assistant logs the decline and diarises a review at the next annual touchpoint. A year on it resurfaces the case, you raise it again, and this time the client is ready to talk.
A decline is respected, not nagged, but it is not forgotten either. The client gets a second, well-timed opportunity to protect themselves, and you evidence that the need was addressed more than once.
The Consumer Duty evidence captured as you go
For each case, the assistant records that protection was raised and what happened, taken up, declined or referred to an authorised colleague, so the file shows the client's needs were genuinely considered. It never frames cover as a condition of the mortgage and never presents a nudge as advice, because conditional selling and misleading framing are exactly what the Consumer Duty forbids.
During a network file review, a case shows a clear line: protection discussed at offer, client declined, review diarised. The evidence that the conversation happened is already in the file, in order, rather than being reconstructed from memory after the event.
You can show, case by case, that protection was put in front of every client, which is precisely what a good-outcomes review looks for. The record builds itself as you work, and it never strays into presenting the mortgage and the cover as a package.
Prompting the protection conversation at the right moment is a straightforward build now, and it answers a documented customer need.
- An assistant that watches for the offer, checks whether cover is on file and prompts the adviser to raise protection can be built today. The need is real and measured: more than 40% of mortgage holders have no life insurance, and take-up of income protection and critical illness cover is lower still, largely because no one raised it at the right moment.Which
- The broker is the natural person to close the gap. The Association of Mortgage Intermediaries frames the protection gap as an awareness problem more than an access problem, with brokers central to the solution. Because you are already discussing the mortgage, a prompt that flags clients with no cover recorded is exactly the nudge that turns awareness into a conversation, without the assistant itself advising.AMI
- It is an internal nudge, and that framing is what keeps it clean. Advising on protection or general insurance is a separate regulated activity, so the recommendation and any suitability assessment come from a suitably authorised adviser. The assistant reminds and drafts; the advice, and the responsibility for it, stay with a qualified person.FCA
- Keep the scoreboard yours, not a vendor's. Any take-up or conversion uplift quoted for a cross-sell tool comes from another firm and from the party selling it, so read it as a direction of travel rather than a promise. The figures that decide it are your own: how many protection conversations get raised, how many clients end up properly covered, and whether the file evidences the discussion every time.
A cross-sell nudge is only safe if it stays a nudge, so these limits come first.
- Advising on protection or general insurance is itself regulated. An AI prompt can remind the adviser to raise it, but the recommendation must come from a suitably authorised adviser who has assessed the client's circumstances. The assistant never quotes, never recommends and never resolves the protection question in the chat.FCA
- The FCA's mortgage intermediary strategy singles out high-pressure selling and conflicts of interest, so a cross-sell nudge must serve the customer's need rather than commission. A prompt tuned to revenue instead of genuine cover gaps is exactly the conduct the regulator is watching for.FCA
- Do not present protection or GI as advice, or as a condition of the mortgage. Conditional selling and misleading framing breach Consumer Duty expectations, so the assistant states plainly that cover is optional and separate, and never packages it with the loan.FCA
- Health and financial information used to identify a protection need is personal data, and some of it is special category. Process it lawfully, only for that purpose, and hold it under a written processor contract with a defined retention period.ICO
Remortgage and product-transfer retention: every expiry surfaced in good time, so the client comes back to you
Every fixed rate you have ever arranged has an end date, and every end date is a decision the client will make with you or without you. Roughly 1.8 million fixed rates expire in 2026, on top of the 1.6 million that expired in 2025, which means your firm is sitting on a rolling pipeline of retention opportunities. The catch is that each one is a hard deadline, and a deadline nobody diarised is a client you lose by default.
The way you lose them is quiet. Product transfers grew 18% to 256bn pounds in 2025, and lenders market execution-only transfers straight to the borrower with a click-to-accept link. A client who taps that link at renewal, without a review, keeps their lender happy and leaves you out of the loop, and leaves themselves without an independent look at whether staying put was actually the right call.
This is precisely the diarised, repetitive tracking that software should own. An assistant built for your firm holds every expiry date across the whole client bank, surfaces each one early enough for you to act, and drafts the re-engagement so you re-open the conversation before the lender's letter lands. It works the calendar so you can work the advice.
Where it stops is the recommendation. Reminding a client that their deal is ending, and gathering an updated picture, is administrative and can be systematised. Whether to product-transfer, remortgage to a new lender or hold is regulated mortgage advice, and that stays with you. Letting a client drift onto a higher standard variable rate, or nudging them to an execution-only transfer without advice, is a foreseeable harm the Consumer Duty expects you to prevent, so the assistant re-engages, it never decides.
Every expiry tracked across the whole client bank
The assistant holds the rate-expiry date for every client and computes when each one needs attention, working back from the expiry so the review starts weeks ahead, not the day the deal ends. It maintains a live who-is-due-when list ordered by date, so no renewal depends on someone remembering it, and no client falls through the gap between one busy month and the next.
In a three-adviser firm the list shows, in June, the eleven clients whose fixes end in September and October. The team sees the autumn wave in good time and plans the re-engagements, instead of finding out a rate ended when the client calls to say their payment jumped.
The rolling pipeline of expiries becomes a managed calendar rather than a series of near-misses. Every renewal gets attention while there is still time to review and recommend, so the client comes back to you rather than defaulting to the lender.
Re-engagement drafted before the lender's letter arrives
As each expiry approaches, the assistant drafts a timely, plain-English message inviting the client to a review, and requests the handful of updated details a fresh assessment needs. It reaches the client ahead of the lender's own retention offer, so the first conversation about the renewal is with you. It books nothing binding and recommends nothing; it opens the door for advice.
Six weeks before a client's fix ends, the assistant sends a short note offering a review and asking whether income, plans or the property have changed. The client replies before the lender's click-to-accept text arrives, and the renewal starts as an advised conversation instead of a solo decision.
You get in front of the client before the execution-only link does. The relationship and the advised review are protected, and the client gets the independent look they would otherwise have missed.
The product-transfer versus remortgage picture laid out for your judgement
For a client at renewal, the assistant gathers the comparison an advised review rests on: the existing lender's retention deal against what other lenders' published criteria and pricing would allow, given the client's updated position. It presents this as research for you to weigh, clearly labelled as support, never as a recommendation to the client and never as a lending decision.
For a client whose fix is ending, the assistant lays out the incumbent's transfer rate beside two other lenders whose criteria the case appears to meet. You review the comparison, apply judgement to what the figures do not capture, and make the recommendation in the client meeting.
The legwork behind a good renewal recommendation is done before you sit down, so your time goes to judgement and the conversation. Product transfers dwarf external remortgages, so getting this comparison right, case by case, is where much of the client's value is decided.
The drift onto standard variable rate caught before it happens
The assistant flags any client heading toward the end of a deal with no review booked, so nobody quietly rolls onto a higher standard variable rate for want of a reminder. It escalates the at-risk cases to you rather than resolving them, and keeps a record that the client was contacted in good time, which is the evidence a good-outcomes review looks for.
A client who did not respond to the first re-engagement is surfaced again, more prominently, two weeks before their fix ends. You call them directly, book the review and avoid the client sliding onto the reversion rate unadvised and unhappy.
A foreseeable harm, the client stranded on an expensive default rate, is designed out of the process. The firm protects both the customer outcome and its own recurring revenue, and can show it acted in time.
Expiry tracking and re-engagement sit on firm, buildable ground, because the dates are fixed and the client bank already holds them.
- An assistant that tracks every rate-expiry date and surfaces each one in good time can be built now. The pipeline is large and dated: about 1.6 million fixed rates expired in 2025 and roughly 1.8 million expire in 2026, so a diarised, whole-client-bank view that nudges at the right moment is exactly the repetitive task software should own, leaving you the conversation and the recommendation.UK Finance
- The commercial risk it addresses is measurable. Product transfers grew 18% to 256bn pounds in 2025, and lenders market execution-only transfers direct to borrowers, which is how a firm quietly loses a client at renewal. A timely, advised re-engagement, prompted before the lender's own offer lands, protects both the client outcome and the relationship.UK Finance via Mortgage Solutions
- Done well, retention is a good-outcomes activity, not just revenue protection. Letting a client drift onto a higher standard variable rate, or nudging them to an unadvised transfer, is a foreseeable harm the Consumer Duty expects firms to prevent. The reminder and the review can be systematised; the recommendation stays with the adviser.FCA
- Set the target against your own book, not a headline. Any retention or conversion uplift attached to a tool comes from another firm and from the party selling it, so treat it as direction rather than a promise. The numbers that matter here are yours: how many expiries are re-engaged in good time, how many clients return for an advised review, and how few roll onto a reversion rate unnoticed.
Retention automation is trusted only if the advice stays with a person, so keep these limits in from the start.
- An expiry reminder is fine, but recommending whether to product-transfer, remortgage or stay put is regulated mortgage advice for the adviser. The assistant diarises, re-engages and gathers an updated picture; the personal recommendation, and the suitability behind it, are yours.FCA Handbook
- Letting a client roll onto a higher standard variable rate unprompted, or steering them to an execution-only transfer without advice, can be a foreseeable harm under the Consumer Duty. The workflow exists to prevent that, so at-risk cases are escalated to a person rather than quietly closed.FCA
- Client-bank data used to track expiries and send reminders is personal data, so keep any marketing within the client's consent and your lawful basis, and hold the data under a written processor contract with defined retention.ICO
- Automated expiry tracking is only as good as your data. A wrong expiry date or a stale product detail re-engages a client at the wrong moment and undermines trust, so verify the dates and product details against the live record before any reminder goes out.
Compliance file and suitability evidence: the case that stands up to a review, assembled and gap-checked before one lands
An advised mortgage sale is only as good as the file that evidences it. With intermediaries arranging close to 89% of all UK mortgage lending, the advised model is the mainstream, and MCOB 4.7A requires the firm to be able to demonstrate that its recommendation was suitable. The case file, the fact-find, the research, the suitability note and the disclosures, is the evidence the whole sale stands on, and a thin or inconsistent one is where advised cases come unstuck.
The pressure on that file is rising, not falling. The FCA's mortgage intermediary strategy says it will review whether firms carry out sufficient assessment of customer circumstances and give suitable advice, which makes file quality a live supervisory issue. A file that looks complete but is templated will not satisfy a Consumer Duty review, so the goal is substance you can show, not the appearance of it.
Much of the risk here is not judgement, it is completeness: a missing suitability note, a protection discussion not recorded, a disclosure absent. That is structured, checkable work. An assistant built for your firm can assemble each file and gap-check it against a required-contents list, flagging the cases that are missing something so you fix them before a review, not during one.
The boundary is firm. The suitability assessment and its reasoning are the adviser's, always. The assistant surfaces what is absent and orders what is present; it never writes the rationale after the event, never back-fills a note to make a file look complete, and never becomes a solely automated decision about a borrower. Since 5 February 2026, any such decision with significant effects runs under UK GDPR Articles 22A to 22D and needs meaningful human involvement, which in an advised model you already have.
Each file assembled and gap-checked against a required-contents list
The assistant pulls together the components of a case, fact-find, research, suitability note, disclosures and evidence of any protection discussion, and checks them against your required-contents list. It flags what is missing or inconsistent per case, so a gap is caught while it is still a quick fix, not discovered in a review. It orders and checks; it does not write the suitability reasoning.
Across a batch of recent completions, the assistant flags three files: one with no suitability note attached, one where the protection discussion is not recorded and one missing a disclosure. You close all three in an afternoon, well before the network's review, instead of finding them under scrutiny.
The completeness of every file is known in advance, not gambled on. The gaps that turn a routine review into a difficult one are closed early, and your time goes to the substance rather than to hunting for what is absent.
A pre-review sweep of the back book
Ahead of a file review, the assistant sweeps a defined set of completed cases and produces a single readiness view: which files are complete, which are missing a named item and which need a closer look. It ranks by risk so the principal sees where the real exposure is without opening every case, and it evidences the check itself as part of the trail.
Before a network audit the assistant reviews the last quarter's completions and returns a list: most complete, six missing one item each, two needing attention. The team works the short list rather than re-reading two hundred files blind, and walks into the review knowing exactly where it stands.
A file review stops being a scramble and becomes a managed exercise. The firm knows its own position before anyone else assesses it, and the effort concentrates on the handful of cases that genuinely need it.
A drafted suitability note the adviser owns and rewrites
From the fact-find and the research, the assistant can draft a first-pass suitability note for the adviser, structured around the required points. It is a starting draft only: the adviser reads it, corrects it and makes it reflect the reasoning they actually applied. The assistant never presents an unread template as the record, and the note is not evidence until a person has owned it.
For a completed case the assistant drafts a note setting out the client's circumstances and the shape of the recommendation. The adviser rewrites the reasoning in their own terms, adds the judgement the draft could not know, and signs it. The final note is the adviser's, informed by a draft, not generated in place of one.
The blank-page friction that leaves suitability notes thin or late is removed, while the substance stays genuinely the adviser's. You get a fuller, more consistent note in less time, without letting a template stand in for real reasoning.
Human involvement kept in, by design
The assistant is built so that no decision about a borrower is ever made solely by the automation. It assembles, checks and drafts, then routes every judgement to the adviser, and it records that a person reviewed and signed off. This keeps the process on the right side of the automated-decision safeguards and mirrors how an advised sale already works.
A file the assistant flagged as incomplete is not closed automatically; it is escalated to the adviser, who decides what is needed and signs off the fix. The record shows a human made the call, which is exactly what the safeguards, and a Consumer Duty review, expect to see.
You get the throughput of an assembled, gap-checked file without ceding any decision to a machine. The human sign-off that the rules require, and that good advice depends on, is designed into the workflow rather than bolted on.
The file work is structured and checkable, which makes it a strong fit for automation now, with the judgement left where it belongs.
- Assembling a case file and gap-checking it against a required-contents list is buildable today. MCOB 4.7A requires the firm to be able to demonstrate that its recommendation is suitable, so a system that flags cases missing a suitability note, a disclosure or a protection discussion lets you fix gaps before a review, not during one. The completeness check is mechanical; the reasoning it protects stays human.FCA Handbook
- The reason to do it now is that file quality is a live supervisory issue. The FCA's mortgage intermediary strategy says it will review whether firms carry out sufficient assessment of customer circumstances and give suitable advice, so a file that only looks complete will not pass a Consumer Duty review. Automation should surface substance, flagging what is genuinely missing, not manufacture the appearance of completeness.FCA
- The build stays compliant because the human is already there. Since 5 February 2026, any solely automated decision about a customer with significant effects runs under UK GDPR Articles 22A to 22D and requires safeguards including meaningful human involvement. In an advised model the adviser signs off every case, so the assistant assembles and checks while a person decides.legislation.gov.uk
- Judge it on your own review outcomes, not a vendor claim. Any compliance-efficiency figure a tool quotes comes from another firm and from the party selling it, so treat it as direction. The measures worth watching here are yours: how many files are complete first time, how few gaps a review finds, and how much less time a file check takes, each before and after.
The file is the evidence the advised sale stands on, so these limits are not optional.
- AI can assemble and gap-check the compliance file, but the suitability assessment and its evidence remain the adviser's responsibility under MCOB 4.7A. The assistant orders what is present and flags what is absent; it does not judge suitability.FCA Handbook
- If any decision about a customer is solely automated with significant effects, UK GDPR Articles 22A to 22D, in force since 5 February 2026, require safeguards including meaningful human involvement. The workflow keeps a person in every case for exactly this reason.legislation.gov.uk
- The FCA expects genuine assessment of each customer's circumstances, so a file that only looks complete will not pass a Consumer Duty review. The assistant must surface real substance, not templated filler that dresses up a thin case.FCA
- Never use AI to back-fill or invent suitability rationale after the event. The record must reflect the advice actually given, when it was given, so any note added later is dated honestly and never back-dated to look original.
Procuration fees, commission and reporting: what you earned matched to what completed, without a night on spreadsheets
Broker income is a patchwork, and reconciling it is one of the least loved jobs in the firm. Lender procuration fees, protection and general insurance commission and any client fees all arrive on their own timetables, from different payers, in different formats. Matching what was earned against what actually completed, and spotting what a lender still owes, is fiddly, repetitive and easy to get wrong, and it usually lands on the owner at month end.
The scale makes it worth fixing. With internal product transfers alone reaching 256bn pounds in 2025, and a busy purchase market on top, even a small firm processes a lot of completions, each carrying a procuration fee that has to be tracked to payment. An assistant built for your firm can match expected fees to completed cases, flag the payments a lender has not yet made and total protection commission by adviser, so you see the real picture without a late night in a spreadsheet.
Reporting is not only about revenue, though. Under the Consumer Duty, fees and commission must be disclosed and represent fair value, and remuneration must not drive unsuitable recommendations. Good reporting makes the fair-value and outcomes data easy to see, not just the volume, and reporting that celebrates volume over outcomes is exactly the conflict of interest the FCA is watching for.
There is a tax dimension arriving too. From 6 April 2026, self-employed brokers with qualifying income over 50,000 pounds fall under Making Tax Digital for Income Tax, with digital records and quarterly updates. Keeping fee and commission records in a clean digital form now makes that transition far less painful later. Throughout, the assistant compiles the management information; the fair-value judgement, and the numbers themselves, stay with the firm.
Expected fees matched to completed cases
The assistant matches the procuration fee expected on each completed case against what the lender has actually paid, so the reconciliation that used to be manual runs continuously. It works from your own records, presents the matched and unmatched items clearly, and leaves the figures for you to confirm rather than treating its match as final.
At month end the assistant shows that of forty completions, thirty-seven procuration fees have landed and match, while three do not yet appear. You review the three rather than reconciling all forty by hand against a lender statement.
The monthly reconciliation stops eating an evening and becomes a review of the exceptions. You see what has been earned and what has been paid at a glance, with the checking effort aimed only at the items that do not tie up.
Unpaid procuration fees flagged before they are forgotten
The assistant tracks which completed cases have not yet produced the procuration fee the lender owes, and flags them by age so nothing quietly falls off the radar. It surfaces the chase list; it does not contact the lender or write off anything itself, leaving the follow-up and any judgement to you.
The assistant flags two completions from six weeks ago where Nationwide's procuration fee has still not arrived. You chase them with the case references to hand, instead of discovering the shortfall only when the quarter's numbers look light.
Fees that would otherwise slip through unnoticed get chased while the trail is fresh. The firm collects what it is genuinely owed, and the owner has a live view of outstanding income rather than a vague sense that something is missing.
Commission and fair-value data, not just volume
The assistant compiles the management information the Consumer Duty expects you to see: fees and commission by adviser and by type, alongside the outcomes and fair-value context, so the report shows more than headline volume. It presents the data; the fair-value judgement, and any action on it, stay with the firm.
The monthly report totals protection commission by adviser and sits it next to case outcomes, so a pattern of high commission with thin suitability evidence would stand out. The owner reviews the picture as a fair-value question, not a league table.
Reporting supports the Consumer Duty rather than working against it. You can see remuneration and outcomes together, which is exactly what a fair-value assessment needs, and avoid the volume-only view the regulator treats as a red flag.
Clean digital records ready for Making Tax Digital
The assistant keeps fee and commission records in a structured, digital form as the money moves, rather than reconstructing them at year end. For the many advisers who are self-employed, that means the records that Making Tax Digital for Income Tax will require, and the quarterly updates it brings, start from an already-clean base. It organises the records; the tax filing and figures stay with you and your accountant.
A self-employed adviser over the threshold approaches the first Making Tax Digital quarter with fee and commission income already recorded digitally and categorised, so the quarterly update is a check rather than a reconstruction of a shoebox of statements.
The move to quarterly digital reporting becomes a formality instead of a project. The records are clean as you go, which serves the month-end picture now and the tax obligation that lands from April 2026.
Fee reconciliation is structured, repetitive work that an assistant can take on now, with the figures left for you to confirm.
- Matching expected procuration fees to completed cases, flagging what a lender has not yet paid and totalling protection commission by adviser is buildable today. It is a fiddly, error-prone monthly job that runs on data the firm already holds, so an assistant can carry the reconciliation and let the owner review the exceptions instead of the whole ledger.UK Finance
- Reporting should serve conduct, not just cash. Under the Consumer Duty, fees and commission must be disclosed and represent fair value, and remuneration must not drive unsuitable recommendations, so the management information needs to show outcomes and fair value beside the totals. An assistant can compile that picture; the fair-value judgement stays with the firm.FCA
- There is a timing reason to get the records clean now. From 6 April 2026, self-employed brokers with qualifying income over 50,000 pounds fall under Making Tax Digital for Income Tax, with digital records and quarterly updates. Keeping fee and commission data in a structured digital form as it arrives makes that transition, and the quarterly reporting, far less painful later.GOV.UK
- Measure it on your own books. Any time-saved figure quoted for a reconciliation tool comes from another firm and from the party selling it, so read it as a direction of travel. The numbers that decide it here are yours: how long month-end reconciliation takes, how many unpaid fees you recover, and whether the reporting actually surfaces fair value, each before and after.
Reporting touches both conduct and confidential data, so keep these limits in view.
- Fees and commission must be disclosed to the customer and represent fair value under the Consumer Duty, and remuneration must not drive unsuitable recommendations. Reporting that celebrates volume over outcomes is exactly the conflict of interest the FCA is watching for, so the management information shows fair value and outcomes, not just totals.FCA
- From 6 April 2026, self-employed brokers with qualifying income over 50,000 pounds fall under Making Tax Digital for Income Tax, with digital records and quarterly updates. Clean digital fee and commission records now make that obligation manageable, but the tax filing and the figures stay with you and your accountant.GOV.UK
- The FCA is scrutinising conflicts of interest and high-pressure selling tied to commission, so a report that ranks advisers on revenue alone can quietly reward the wrong behaviour. Present remuneration alongside suitability and outcomes, never as a bare league table.FCA
- Commission and client financial data are personal and commercially sensitive, so keep them secure and within your lawful basis, hold them under a written processor contract with defined retention, and keep internal fee and margin figures out of any client-facing message.ICO
We build them, on Claude
These AI flows do not stay on paper. Svennis Cloud Solutions builds and integrates them into your systems, with a team of certified Claude architects, on Anthropic technology, from the first WhatsApp message to the finished invoice.
See what this looks like in your brokerage
Describe your brokerage in one sentence and Claude will show you live where AI would make the biggest difference, using your own numbers.
Sources
- 1. Mortgage Solutions / IMLA - intermediary share forecast
- 2. FCA - portfolio letter to mortgage intermediaries (30 January 2025)
- 3. Mortgage Solutions - Paradigm and Smart Money People broker AI surveys 2025
- 4. FCA Handbook - MCOB 4.7A Advised sales
- 5. MoneyHelper - what happens when I get a mortgage in principle
- 6. ICO - Guidance on AI and data protection
- 7. UK Finance - how AI is transforming UK mortgage lending
- 8. FCA Handbook - MCOB 11 Responsible lending
- 9. UK Finance via Mortgage Solutions - product transfers and remortgaging 2025
- 10. FCA - PS25/11 Mortgage Rule Review
- 11. UK Finance - 2026 mortgage market forecast
- 12. FCA - Consumer Duty
- 13. Which - life insurance gap leaves mortgage holders exposed
- 14. AMI - Tackling barriers: the protection gap
- 15. legislation.gov.uk - Data (Use and Access) Act 2025, s.80
- 16. GOV.UK - Making Tax Digital for Income Tax for sole traders and landlords