In the latest article in his series on AI in conveyancing, Ed Molyneux puts a practical example in front of the profession and invites them to test it themselves.
I have spent the past month writing about what AI can and cannot do for conveyancers. Provenance. The judgment line. The difference between intelligence work and professional discretion. These are important distinctions, but I am conscious that they remain abstract until you can sit down with an actual tool and see whether it holds up.
So today I am going to describe something you can test for yourself, for free, in about ten minutes.
The scenario
A buyer’s conveyancer receives instructions on a leasehold flat. The lease has 72 years unexpired. The buyer has been approved in principle for a 25-year mortgage.
The first question is straightforward: can the buyer’s lender actually lend on this property?
The second question is more interesting: if not, which lenders can?
These are not obscure questions. They arise in thousands of transactions every year. The answers are in the UK Finance Mortgage Lenders’ Handbook — Part 2, section 5.14, published individually for each lender. There are over 60 lenders on the register, each with their own thresholds, referral criteria, and end-of-term requirements.
In practice, most conveyancers know the thresholds for the handful of lenders they deal with regularly. For the rest, it means a trip to the UK Finance website, finding the right lender page, locating section 5.14, and interpreting what is often dense and conditional language.
It is not difficult work. But it is time-consuming, and it is exactly the sort of intelligence work that sits below the judgment line I described in an earlier article.
The test: ask a general-purpose AI
I asked a leading AI model – without any specialist guidance – to assess whether a buyer could get a mortgage on a 72-year lease.
The response: “Most lenders require a minimum unexpired lease term of around 70-80 years remaining at the point of completion.”
That answer is not wrong. It is also not useful. It does not name any lender. It does not distinguish between minimum term at completion and minimum term at end of mortgage. It does not mention that Nationwide’s threshold is 55 years – making it one of the most permissive mainstream lenders for short leases – while Virgin Money requires 85, effectively ruling out any lease under that figure entirely.
A conveyancer acting on “around 70-80 years” would have no basis for advising their client on whether to proceed, switch lender, or negotiate a lease extension.
You can verify this yourself. Open ChatGPT, Claude, or any AI tool and ask the question. You will get a version of the same answer – plausible, generic, unactionable.
The test: ask with specialist knowledge
The same model, equipped with a specialist skill containing the actual Part 2 lease requirements for every major UK lender, produced a materially different answer.
For a 72-year lease with a 25-year mortgage:
Nationwide: Eligible – 55-year minimum at completion.
NatWest: Eligible – requires mortgage term plus 30 years. At 72 minus 25, the remaining term at redemption is 47 years, exceeding the requirement.
Santander: Eligible with conditions – complex rules, referral required if the unexpired term is below 82 years.
Barclays, Halifax, Lloyds: Marginal – 70-year minimum at completion. At 72 years, the lease technically qualifies but is close enough to the threshold that a referral is likely.
HSBC: Ineligible – requires a minimum of 50 years remaining after the mortgage term. That means 75 years at completion for a 25-year mortgage.
Virgin Money: Ineligible – requires 85 years at completion.
Six eligible, five marginal, two ineligible out of thirteen major lenders – with the specific reasoning for each.
Why the difference matters
I want to be clear about what this is and what it is not.
This is not artificial intelligence making a professional judgment. It is not advising your client. It is an AI model doing what it does well – reading structured data and cross-referencing it accurately – because it has been given access to verified, current reference material instead of relying on whatever it absorbed during training.
The professional judgment remains yours. Deciding whether to recommend a lender switch, how to present a marginal lease to the valuation team, whether to negotiate an extension as a condition of purchase – these are decisions that require understanding of your client’s circumstances and priorities. No AI skill replaces that.
But the processing work – compiling, cross-referencing, and structuring the handbook data across multiple lenders simultaneously – is handled in seconds rather than the 15 to 20 minutes it might otherwise take. And the output can be traced back to specific handbook entries, not to “the AI thinks so”.
There is a second advantage that is easy to overlook. The handbook changes regularly. Last week, our automated refresh detected that Landbay Partners had updated their ground rent thresholds – introducing distinct limits for new-build and non-new-build leaseholds, and adding a requirement for a deed of variation where those limits are exceeded. That change was committed to the toolkit within 24 hours – the handbook is checked daily and any differences are automatically flagged and updated. Any conveyancer using the skill from that point on would receive advice based on the current requirements, not the version that was correct six months ago.
That is the difference between a tool that relies on a model’s training data and one that works from maintained, versioned reference material.
Try it yourself
We have open-sourced these skills. They are free, work in ChatGPT, Claude, or any AI tool that supports skills, and require no sign-up. Instructions on how to use the skills (easiest in Claude) are laid out clearly.
• SDLT Calculator – deterministic stamp duty calculations with live-updated rates
• Lease Impact Advisor – the lender eligibility analysis described above
• Lenders Handbook Pre-Screen – systematic Part 1 and Part 2 handbook checks across 67 lenders
The toolkit is on GitHub at github.com/MoverlyLtd/conveyancing-toolkit.
I would encourage any conveyancer reading this to try it. Pick a property from your current caseload – a leasehold with a short lease, a purchase where you are uncertain about a lender requirement — and see whether the output matches what you know. If it does, consider what that means for how you allocate your time. If it does not, tell us – we want to know.
The toolkit is permissively MIT-licensed and will remain free. We are publishing it because we believe that better tools for conveyancers should not be locked behind subscriptions, and because we think the profession is better placed than any technology company to decide what those tools should do.
We want to build 30 skills, not three. Lease extension calculators, search report analysers, protocol compliance checkers, building regulations advisors – practical tools for practical work. We would welcome contributions from practitioners who know what is actually needed.
About the author
Ed Molyneux is co-founder and CTO of Moverly, the property intelligence platform working with LMS and Connells Group to bring structured, verified data to property transactions. He is the architect of the Property Data Trust Framework (PDTF), the open standard for machine-readable property data now being adopted across the industry. Ed writes about AI, property data infrastructure, and the future of conveyancing.
