Ed Molyneux has spent recent weeks explaining what AI can do, where it fails, and why provenance and shared data matter. Now he walks through a day in the life of a buyer’s conveyancer working with an AI agent on a live transaction — and invites the profession to consider what changes and what stays the same.
Over the past few articles I have set out a framework for thinking about AI in conveyancing. The jagged frontier of capability. The judgment line between research and professional discretion. Why provenance matters. Why shared, structured data is the foundation everything else depends on.
All of that is abstract until you can picture it in practice. So today I want to try something different — and I’d like your help with it.
I am going to walk through what a single transaction could look like when a buyer’s conveyancer works with an AI agent — not a one-off report on title generator, but an autonomous assistant connected to verified property data, the lender’s handbook, and a diligence engine that has already analysed every piece of information on the file.
This is fiction, but only just. Every capability I describe either works today or is in active development. The property, the people, and the conversation are invented. The technology is not. But I am not presenting this as a finished answer. I am presenting it as a conversation starter, and I genuinely want to hear where you think it breaks down, what would never work in your practice, and what you wish it could also do.
The file
Sophie is a buyer’s conveyancer at Jones LLP. Her client Priya is a first-time buyer, approved in principle for a 25-year mortgage with Nationwide at 85% LTV.
The property is 14 Oakwood Terrace, Bristol. A semi-detached Victorian house, leasehold, with 68 years remaining on the lease. When instructions arrive and the property data is loaded, the platform’s diligence engine — a deterministic system that evaluates every transaction against the full range of transactional risks — runs automatically. By the time Sophie opens the file, the analysis is already done.
Monday morning: the briefing
Sophie opens her AI assistant. It presents the file summary:
> 14 Oakwood Terrace. Four risk flags, ordered by severity. The lease has 68 years remaining — critical. No building regulations certificate for a loft conversion — high. The property is in Flood Zone 2 — high. A rear extension with no planning permission on record — moderate.
In a traditional workflow, Sophie might have spent 30 to 45 minutes reviewing the title register, the property information form, the lease, and the search results to reach the same conclusions. The diligence engine reached them in seconds, and every finding links back to specific evidence — the title register entry, the TA6 response, the Environment Agency flood map.
But the assistant has flagged something else. The combination of a critical lease issue and a lender notification requirement means this file carries elevated professional liability risk. Jones LLP has configured their system to flag this: not as a gate, but as a prompt. Files above a certain liability threshold are surfaced for senior review at key milestones — report on title, exchange, and lender correspondence.
Sophie is two years qualified. She handles the file herself, but she knows that her supervisor will review her lender notification before it goes out. The system does not replace supervision. It makes supervision targeted. Instead of a senior conveyancer sampling files at random, the platform directs their attention to the files that most need it. For Sophie, it is a safety net that builds confidence. For her supervisor, it is a way of trusting junior staff with complex files without losing sleep.
The lease: where judgment begins
The lease term is the critical flag. The assistant has already cross-referenced it against Nationwide’s Part 2 handbook requirements: with a 25-year mortgage, the remaining term at redemption would be 43 years, well below the 70 years Nationwide requires.
That cross-referencing is research. Reading the lease, checking the handbook, performing the arithmetic. Accurate, necessary, and exactly the kind of work AI does well when it has access to the right data.
What happens next is judgment. Sophie considers the options. The seller has owned the property for 12 years, making her eligible for a statutory lease extension. There are three routes: complete the extension before completion, assign the benefit of a Section 42 notice to the buyer, or negotiate informally with the freeholder. Sophie decides the assignment route is best — it allows the transaction to proceed without waiting for the extension to complete, and Nationwide’s handbook confirms they will accept it.
The assistant drafts an enquiry to the seller’s conveyancer. Sophie amends the wording and sends it. The enquiry is linked to the risk flag, to the lease data, and to the handbook section. When the response comes back, the diligence engine reassesses the flag automatically.
Five minutes. In the old workflow, the same five minutes to draft — but only after the 30 minutes of reading that identified the issue.
The loft conversion
A loft conversion with no building regulations certificate. The assistant has already checked Nationwide’s handbook position on indemnity insurance as an alternative and confirms the lender will accept it. It has also flagged that the seller’s conveyancer has been asked to obtain a quote.
Sophie notes this and moves on. When the policy arrives, the assistant will check the wording against the lender’s requirements and flag anything non-standard. The research is handled. The judgment — whether indemnity is appropriate for this property, this client, this lender — remains with Sophie.
Flood Zone 2
The assistant presents the Environment Agency classification, the seller’s disclosure of no flooding history during 12 years of ownership, a non-return valve on the drains, and existing buildings insurance that includes flood cover.
Sophie considers whether a specialist flood risk assessment is warranted, or whether the existing evidence is sufficient. She decides it is. A less experienced conveyancer might have defaulted to ordering the report, adding cost and weeks. A more cautious one might have raised an enquiry asking for information already on the file. Sophie, freed from the overhead of compiling the evidence, has the bandwidth to make a proportionate decision.
The rear extension
The fourth flag resolves quietly. The extension was built in 1987 — well outside the four-year limitation period. When the local authority search returns three days later, the assistant checks it, confirms no enforcement action, and updates the flag to resolved. Sophie gets a one-line notification. She does not need to reread the search results to find this one point.
Lender correspondence
Sophie needs to notify Nationwide about the lease, the flood zone, and the resolution plan. The assistant drafts the notification with references to specific handbook sections and supporting evidence.
Because this file was flagged for elevated liability risk, her supervisor reviews the draft before it goes. He suggests a minor amendment to the lease wording. The whole review takes three minutes — targeted, efficient, and focused on the file that needs it most rather than spread thinly across the entire caseload.
The real value is not the time saved on this one letter. It is consistency. Every lender notification across Sophie’s caseload follows the same structure, addresses the same requirements, references the same provisions. Nothing is missed because she was tired at 4pm on a Friday or because this was her 15th notification that week.
Three weeks later
The seller serves the Section 42 notice. The indemnity insurance arrives. Nationwide confirms their position. The local authority search returns clean.
Sophie’s assistant tracks each development and presents a morning summary:
> 14 Oakwood Terrace — all four risk flags resolved. Lease: Section 42 notice served, assignment clause drafted. Loft: indemnity policy received, lender requirements met. Flood: insurance confirmed, lender satisfied. Extension: immune from enforcement per local authority search.
>
> Remaining before exchange: mortgage offer, expected this week. File is on track.
Three weeks from instruction to exchange-ready. The traditional timeline for a property with these issues would be eight to twelve weeks, driven by the time it takes to identify the problems, raise the right enquiries, chase responses, and coordinate between parties working from different information.
What changed and what did not
What changed is the research. Issue identification happened on day one, not week three. Handbook cross-referencing was instant. Enquiries were linked to evidence. Document tracking was automatic. Lender correspondence was consistent. The professional liability risk was flagged and supervision was directed to where it mattered most.
What did not change is the judgment. Sophie decided the assignment route was best. Sophie decided indemnity was appropriate. Sophie decided a specialist flood report was unnecessary. Her supervisor reviewed the lender notification on the file that warranted it. Every one of those decisions drew on professional experience, knowledge of the client, and a feel for proportionality that no system can replicate.
The technology did not make Sophie less important. It made her more effective. It took the 70% of her work that sits on the research side of the judgment line and handled it faster, more consistently, and with better traceability. And it left the 30% that actually requires her — the advice, the judgment, the professional discretion — exactly where it should be.
A question for working conveyancers
If you are reading this, think about your current caseload. How many files have issues you have not yet identified because you have not had time to read everything? How many enquiries are you waiting on that could have been raised on day one? How many times this month have you manually checked a lender’s handbook when a system could have flagged the issue in seconds?
The agentic diligence model described here is not about replacing your expertise. It is about removing the obstacles that prevent you from deploying it — and giving firms better tools to support their people, direct supervision where it counts, and build the confidence of the next generation of conveyancers.
This is already being built by several companies, including my own, Moverly. Some of it works today. Some is a few months away. All of it is grounded in the same principle: verified data, clear provenance, deterministic rules for what must be right every time, and professional judgment for everything else.
But I meant what I said at the start: this is a conversation, not a pitch. If you read Sophie’s morning and thought “that would never work because…” — I want to hear the end of that sentence. If you thought “but what if it could also do this for me?” — I want to hear that too. The best version of this technology will not be designed by technologists working in isolation. It will be shaped by practitioners who know what actually happens at half past four on a Friday with six completions to get through.
Tell me where I am wrong. Tell me what I have missed. That is how we build something worth using.
Next: the agent-ready firm, and what it practically means to prepare your practice for this shift.
About the author
Ed Molyneux is co-founder and CTO of Moverly, the property intelligence platform working with LMS, Connells Group, and Lloyds Banking Group to bring structured, verified data to property transactions. He is the original author of the Property Data Trust Framework (PDTF), the open standard for machine-readable property data now adopted across the industry.
