The problem
One of the most valuable things a mortgage broker does is match a specific client to the most suitable lender, taking into account the full picture of that client's circumstances. But doing that matching well requires holding a significant amount of criteria knowledge in your head, and applying it accurately to a set of facts that rarely fits neatly into a single box. Most brokers develop strong intuitions over time, but intuition can be wrong, and the criteria landscape changes constantly.
The practical problem is one of scale. When you have a client with, say, a recent default, a relatively low deposit, self-employed income, and a property with a non-standard construction, you are not looking at a single criteria question. You are looking at four or five intersecting criteria questions simultaneously, and you need to find lenders who score acceptably on all of them. Doing that manually across the full panel takes time even for experienced brokers, and there is always a risk that a viable option is missed.
Beyond the research challenge, there is a client communication challenge. Clients do not understand why one lender will consider them and another will not. Explaining criteria differences clearly and in plain English, without going into so much detail that you lose the client, is a skill in itself. Having a well-structured comparison to hand makes those conversations significantly easier.
The system
Step 1: Build a precise client scenario summary (Claude)
Start by using Claude to help you articulate the client scenario in precise terms. This is worth doing because it forces clarity before the research begins, and produces a clear brief you can use in subsequent prompts.
Prompt example: "I have a mortgage client with the following details. Please summarise this as a clear scenario description that identifies all the criteria considerations a lender would need to assess: Age 34, employed full-time for 2 years at current employer, annual salary £41,000. Previous employer for 4 years before that. Credit history includes one default registered 22 months ago for £400 (satisfied 18 months ago). No other adverse credit. 10% deposit. Property is a 1960s ex-council flat, concrete construction, in a large urban local authority block. Looking to borrow £190,000."
Claude will identify the relevant criteria categories: income calculation, employment history, adverse credit, deposit level, property type, and any others specific to the scenario. This gives you a clear checklist for the comparison.
Step 2: Research lender appetite by criteria category (Perplexity)
For each significant criteria category, use Perplexity to research current lender positions.
Prompt example: "Which UK mortgage lenders currently accept applications with a single satisfied default registered within the last 24 months, where the default was under £500? What are their typical conditions, such as minimum time elapsed since registration or satisfaction, and any LTV restrictions?"
Follow-up prompt: "For the property type, which UK residential mortgage lenders currently lend on ex-local authority flats with concrete construction? Are there restrictions on floor level, block size, or local authority concentration?"
Cross-reference key findings against your sourcing system or current lender documentation. AI research is a powerful starting point but criteria change frequently and require verification.
Step 3: Synthesise into a lender shortlist (Claude)
Bring your research together and ask Claude to identify the overlap: lenders who could potentially accommodate all the criteria considerations simultaneously.
Prompt example: "Based on the following criteria research, please identify which lenders appear on the acceptable list for all of the following criteria: adverse credit with satisfied default under 24 months, 10% deposit, ex-council concrete construction property, employed income. Where lenders appear on multiple lists, flag them as potentially suitable. Where a lender appears acceptable on some criteria but has a question mark on others, flag them as needing further investigation. [Paste your research notes]"
This overlap analysis is where the time saving is most dramatic. What used to require mental cross-referencing of multiple criteria notes is done systematically in seconds.
Step 4: Draft a client-facing summary (Claude)
For the client conversation, use Claude to produce a plain-English explanation of why certain lenders are suitable and others are not.
Prompt example: "Based on this scenario and criteria comparison, please draft a short client-friendly summary explaining: why this case requires careful lender selection, which types of lender are potentially suitable and why, which types are likely to decline and the main reasons, and what I as their broker have done to identify the best options. Write in plain English, no jargon, reassuring tone. Assume the client is worried about their credit history and has not bought a property before."
The results
Before: Building a lender comparison for a complex case took experienced brokers 45 to 90 minutes of manual research, cross-referencing multiple criteria sources. The risk of missing a viable lender was real, particularly for less common criteria combinations.
After: The same comparison is structured in 20 to 30 minutes, with a clearer output and a ready-made client explanation. One broker reported finding a suitable lender for an adverse credit client with a non-standard property that their initial instinct had suggested was "probably unbrokerable". The systematic AI-assisted comparison identified two viable options. The time saving across a typical week of complex cases amounts to two to three hours, which is redirected into client-facing activity.