Conveyancing: enter the details once, generate every document

A property lawyer once said: "I didn't train for years to copy-paste names and addresses."

Yet a substantial part of conveyancing consists of exactly that. The client's details — the names of the parties, the property address and title number, the purchase price, the completion date — get typed into the purchase contract. Then the same details get typed again into the completion statement. Then the land-registry submission. Then the letters to the client.

The same data, entered four or five times. One typo away from a real problem.

Why this is expensive

Assembling the document pack for a typical matter takes around three hours. Not because it is complicated, but because it is repetitive: the same operation, page after page.

The risk, though, isn't in the time. A mistyped title number or a wrong completion date doesn't appear in one place — it appears everywhere it was copied to. If the error is in the contract, it is also in the completion statement and the land-registry submission. That is not a minor inconvenience; it is a liability question.

And meanwhile the buyers are waiting. A house purchase is the largest transaction most people ever make, and patience runs short when the reason for delay is paperwork.

Entered once, reused everywhere

The alternative isn't AI writing contracts. It is the client's details being captured a single time, with every document then produced from those details, using the firm's own approved templates.

In practice: the parties' names, the property details, the price and the completion date go into the system once. AI drafts the purchase contract from the firm's template, then automatically generates the follow-on documents — the completion statement, the land-registry submission, the letter to the client.

It isn't writing generic legal text. It is populating the firm's own templates, developed over years, with the structure and the clauses the firm already uses.

The firm's template stays the firm's template

This is where most legal AI tools fall down. A general-purpose generator of "legal-sounding" text is unusable, because every firm has developed its own structure, clauses and phrasing — often over decades, often as a direct result of specific matters that went wrong.

So the system doesn't compose. It populates the firm's existing, approved templates. If a template carries a clause the firm added after a past dispute, that clause stays. If a phrasing is deliberate, it doesn't get tidied into something "better".

The practical consequence is that implementation begins by reviewing templates, not by discussing technology. Anything imprecise in a template propagates faster once automated — so it is worth putting that right first.

The lawyer approves

This part isn't optional, and it shouldn't be.

What comes out stays a draft until a qualified lawyer has read it, reviewed it and signed it off. The system sends nothing, signs nothing, and makes no legal judgement. AI drafts; the lawyer is accountable.

That isn't a constraint to be tolerated — it is the point of the design. Legal judgement is what a client comes to a lawyer for. Automation doesn't take that away. It takes away the three hours that come before it.

What changes

• Time per matter: from about three hours to roughly fifteen minutes
• Follow-on documents: generated automatically, every one
• Typos: details are entered once, so there is nothing left to get wrong repeatedly

Non-standard matters

A substantial share of conveyancing is standard: two private individuals, unencumbered title, a mortgage, a conventional timetable. The system carries those end to end.

The rest aren't like that. Shared undivided ownership, a life interest, property from a probate matter, a foreign buyer, an amendment to the deed of a block — in those cases automation doesn't produce the whole pack, only the repetitive part of it.

That isn't a limitation of the system so much as a correct division of labour. A non-standard matter is non-standard because it needs legal judgement, and a system that "handled" that part too would be a risk rather than an advantage.

The question of liability

Firms reasonably ask: who is responsible if there's an error in a generated document?

The answer doesn't change from how things work today. Today a paralegal or trainee prepares the document, and the supervising lawyer is responsible, because they are the one who signs. Automation doesn't alter that — it simply replaces who does the preparing.

Which is why the approval step isn't negotiable. A system that sent documents out without a lawyer's review wouldn't be more efficient; it would be unusable.

What it doesn't solve

Automation speeds up repetitive document production. It does not replace reading the title, investigating charges, assessing the client's position, or negotiating. Those stay exactly where they are.

The difference is that the firm's time is available for them. Firms that adopt this won't be smarter than the ones that don't. They will simply spend far less of the day on admin, and far more of it with clients.