Saturday afternoon, someone views a house they like. By nine that evening they are working out whether they can afford it, and looking for a broker.
Monday morning, three brokers call them back. Two are too late.
Mortgage enquiries cluster around events that do not happen during office hours: viewings, offers accepted, rate announcements, and the moment a fixed term appears on a bank statement. Every one of those produces a burst of searching from people who want an answer that evening.
The intake problem
A broker's first conversation with a new enquiry is mostly fact-finding. Buying or remortgaging. What price. What deposit. Employed or self-employed. Any credit issues. When do they need it.
It is a twenty-minute call, it is not chargeable, and a meaningful share of them end in "we can't help with that at the moment". Adverse credit, a deposit that is not there yet, self-employed with one year of accounts, or a timeline six months out.
None of that is a criticism of the enquirer. It is just work that could have happened before anyone picked up a phone.
What a bot can safely establish
The intake set translates well into a conversation:
Purchase or remortgage. Different processes, often different advisers.
Property price and deposit amount. The two numbers that shape everything.
Employment type. Employed, self-employed, contractor, or company director. This is the field most brokers do not ask for until the call, and it changes which lenders are viable more than almost anything else.
Whether an offer has been accepted. Distinguishes someone with a deadline from someone exploring.
Timeline. Completing in six weeks and thinking about next year are not the same lead.
Any known credit issues. Ask neutrally and let people volunteer it. Many will, because they would rather know now.
That arrives as a record an adviser can prioritise, and it means the first call starts at the useful part.
The line that cannot be crossed
Mortgage advice is regulated, and this is where a chatbot in this sector either behaves or creates a problem.
Fine: how the process works, what documents are typically needed, what a decision in principle is, what valuation and survey mean, roughly how long each stage takes, what your fees are and when they are payable, what happens after an offer.
Not fine: how much someone can borrow, whether a product suits them, whether to fix or track, whether they will be accepted, what rate they will get, or any comment on affordability.
The pull toward the last group is strong, because those are the questions people actually ask. "How much could I get on £45k?" is the first thing many enquirers type, and a model will want to be helpful.
Configure the refusal carefully, and make it useful rather than blank. Explaining that borrowing depends on income, commitments, credit history, and each lender's criteria, and that an adviser can give a real figure quickly, is a better answer than a number and it converts perfectly well.
Process questions are the genuine value
Where a bot earns its keep here is the enormous volume of procedural confusion around mortgages.
First-time buyers in particular do not know what a decision in principle is, whether it affects their credit file, what a solicitor does versus a conveyancer, what happens between offer and exchange, why a valuation came back low, or what porting means.
None of that is advice. All of it is documented, or should be. And a broker who explains it clearly at nine on a Saturday evening looks considerably more helpful than the three who send a callback request form.
Remortgage timing
A specific and valuable use. People start looking roughly six months before a fixed rate ends, and they often do not know when that is or how early they can lock something in.
A bot that explains the timing, captures when the current deal ends, and flags anyone inside the window they should be acting in turns a vague enquiry into a diarised opportunity. Anyone further out gets a note for later rather than being lost.
Data caution
Keep it proportionate. Ask for enough to route and prioritise, not a full financial picture. No account numbers, no full credit history, no identity documents in a chat window. Those belong in your secure fact-find.
Say plainly at the start that the exchange is an enquiry, that it is not advice, and that an adviser will follow up. That framing protects everyone and costs one sentence.
Configuration and testing
Point it at your process pages, fee information, and any published guides. Set the intake fields. Write the advice boundary first, then test it hard, in the phrasings borrowers actually use, which are casual and direct.
Flag anything with a deadline for a first callback, because in this business the person who rings back first on a live purchase usually keeps the client.
The number that matters
Enquiries captured outside office hours, and what share of them your advisers can call already knowing the situation.
The second one is where the time comes back. A first call that begins with the facts already gathered is a shorter call and a better one, and it happens with someone who was still awake when your competitors were not.