Claude Code for Conveyancers and Title Agents: Drafting Filing Explanations and Status Updates

Friday at five, I came back from closing three property deals and found a sticky note on my desk: “Still haven’t sent the status update to the Hendersons.”

That’s a story a friend of mine who runs a small conveyancing office told me. A client inheriting a house just needs one line: “Your transfer has been filed with the registry, and it should complete in about two weeks.” That’s all. But when deals stack up, that one line slides to the bottom of the list. The client who didn’t hear back calls in asking “What’s going on?”, and handling that call burns another half hour. The note itself takes five minutes to write. Finding the five minutes is the hard part.

And every time, you write it from scratch. The explanation of a title transfer, the note about a mortgage discharge, the warning about a disclaimer of inheritance. The content is roughly the same, but each matter has slightly different facts, so you dig up an old email, copy it, swap the dates and names, and the busywork quietly piles up.

This kind of writing, “almost the same every time, but typed by hand,” is exactly the work that pairs best with an AI draft. Today I’ll show you how to turn that into a real office workflow using Claude Code.

Key takeaways

• In a conveyancing or title office, the “filing explanation” and the “client status update” follow a fixed template, so they’re the best match for an AI first draft.
• Hand the AI the draft only. Whether a filing is valid, legal advice, and the final check stay with the licensed professional. Draw that line first.
• I’ve included a copy-paste prompt template and a runnable script that takes matter notes and generates all the update letters in one batch.
• To protect personal data, decide upfront to pass redacted names, addresses, and registration credentials, never the real ones.
• If you cut the writing time per matter from 15 minutes to 3, at 100 matters a month that’s roughly 20 hours back.

First, who is this for

This article helps a specific kind of office.

One to three licensed conveyancers or title agents, plus a few paralegals. You routinely handle inheritance transfers, sale-based title transfers, company director changes and incorporations, and mortgage discharges. You’re often out of the office for closings and consultations, so document writing gets batched into the evening or early morning. A paralegal drafts and the licensed agent corrects, but in busy season even the drafts fall behind.

In other words, the problem isn’t a lack of writing skill. It’s a lack of writing time, plus the waste of building something similar from zero every time. If that’s you, the steps below work straight off the page.

If you’re new to Claude Code itself, skim Claude Code for non-engineers first, and the steps here will land more easily.

Breaking down the document flow

Let me lay out the document work on a filing, from intake to completion.

1. Intake: gather the request details and hand over a guide to the required documents.
2. Engagement: prepare the quote and a written description of the work.
3. Before filing: tell the client “we’re about to file” and “here’s the expected completion date.”
4. After filing: send an interim update if the registry’s processing status changes.
5. Completion: report that the filing is done and explain handover of the registration credentials and certified copy.

Of these, the parts you write in nearly the same template every time are the engagement description (2), the status updates (3 and 4), and the completion report (5). The parts that need a legal judgment (whether the filing will pass, which documents are required, how much the registration tax is) can’t go to the AI. But the step that comes after the judgment, “writing it up in words the client understands,” is exactly what the AI is good at.

Separating these is step one. Judgment is yours, fair copy is the AI’s. Decide that and the accidents mostly stop.

Use case 1: rewriting a filing explanation for the client

What’s obvious to you, “title transfer,” “mortgage registration,” is a phrase the client is hearing for the first time. Hand them a description in raw jargon and you’ll get more “what does this mean?” follow-ups later.

So you have the AI keep the technical substance but rewrite it into plain language the client can read.

A common rework loop: a paralegal writes the explanation by reusing the wording from a different past matter, and writes “ownership” where it should say “mortgage.” The licensed agent catches it and fixes it. That round-trip happens every single time.

After this change: you hand over just the facts as bullet points, and the AI shapes them into the explanation. The names and amounts that are easy to mix up only exist in the bullets you provided, so reuse errors become much less likely.

The table below is the line between what the AI does and what the human decides.

Step	AI handles	Conveyancer decides
Identifying the type of filing	No	Yes
Confirming required documents	No	Yes
Calculating the registration tax	No	Yes
Wording the explanation	Yes	Final check
Plain-language rewrite of jargon	Yes	Final check
Tuning the tone for the client	Yes	Final check

Use case 2: generating status updates from matter notes in one batch

A status update follows a fixed template. “Whose,” “which filing,” “at what stage now,” and “expected completion date.” Fill those four and you have a letter.

You stockpile matter notes in a shape like this.

Name: Client A
Filing type: Inheritance-based title transfer
Stage: Filed with the registry
Filed on: 2026-06-05
Completion by: 2026-06-19
Notes: Discuss how to hand over the title deed after completion

Hand that to the AI and it returns a draft update letter you can send to the client. Pour in several matters at once and the drafts for all of them are ready in a 30-minute evening block. After that, you just read each one and send.

ROI estimate: writing one update letter by hand, including the time spent digging up old emails, runs 15 minutes. Let the AI draft it and you just review, and it’s 3 minutes. The gap is 12 minutes. An office sending 100 letters a month: 100 letters x 12 minutes = 1,200 minutes, about 20 hours back. If you value a conveyancer’s hour at, say, $50, that’s $1,000 of time a month.

The numbers shift with office size, but the direction holds: the more by-hand letters you write, the more this pays off.

Use case 3: templating the completion report and handover note

Once a filing completes, you send the report and the handover note for the registration credentials. Getting this wrong hits trust directly, so it’s worth building the boilerplate carefully.

Turn it into a checklist and the AI draft has fewer gaps when you review it.

• Is the completed filing type correct?
• Is the completion date filled in?
• Are the names of the documents being handed over accurate (registration credentials notice, certified copy of registered matters, etc.)?
• Does the handover method (in person, mail, data) match what the client asked for?
• Is the warning on handling the registration credentials (don’t show others, don’t make copies) included?
• Is what the client does next stated clearly?

Write this checklist straight into the CLAUDE.md file that comes up later, and the AI will check itself at the draft stage. The mechanism for teaching project rules to the AI is covered in how to write CLAUDE.md.

A copy-paste prompt template

First, a base prompt that works for both explanations and update letters. Paste it straight into the Claude Code chat and swap out the matter-info section.

You are an assistant helping a conveyancing office write documents.
Follow these rules and produce a draft update letter to send to the client.

# Rules
- Assume the client is not a legal expert; on first use of any jargon, gloss it in parentheses.
- Do not give definitive legal advice. Stay to reporting facts and explaining the next steps.
- For amounts, dates, and names, use only what is in the matter info below. Do not fill gaps with guesses.
- If information is missing, do not put it in the body; list it at the end under "Items to confirm" as bullet points.

# Matter info
Name: Client A
Filing type: Inheritance-based title transfer
Stage: Filed with the registry
Filed on: 2026-06-05
Completion by: 2026-06-19

# Output
A subject line and body in polite, ready-to-send prose. Do not add a signature.

The two key lines are “do not fill gaps with guesses” and “list missing info under items to confirm.” With those in place, you prevent the accident where the AI invents a date or an amount on its own. For finer prompt construction, advanced prompt engineering is a useful reference.

A check script: batch-drafting update letters from matter notes

When there are several matters, pasting them into chat one by one is a chore. Here’s a script that reads a file of pooled matters and, on the Claude Code side, assembles a draft prompt for every one. If you have Node.js, it runs.

Pool your matters in cases.json. Don’t put real names or addresses in; keep them redacted (I’ll explain why in the next section).

[
  {
    "name": "Client A",
    "type": "Inheritance-based title transfer",
    "stage": "Filed with the registry",
    "appliedOn": "2026-06-05",
    "doneBy": "2026-06-19"
  },
  {
    "name": "Client B",
    "type": "Mortgage discharge",
    "stage": "Required documents received, preparing to file",
    "appliedOn": "",
    "doneBy": "2026-06-12"
  }
]

The script below reads each matter and writes out a “prompt you can paste straight into Claude Code” per matter. The text generation itself is left to Claude Code; the script sticks to the prep work. That’s the division of labor.

import { readFile, writeFile, mkdir } from "node:fs/promises";

const cases = JSON.parse(await readFile(new URL("./cases.json", import.meta.url), "utf8"));

// From one matter's info, assemble a paste-ready prompt.
function buildPrompt(c) {
  const lines = [
    "You are an assistant helping a conveyancing office write documents.",
    "Assume the client is not an expert; produce a polite draft status update.",
    "For amounts, dates, and names, use only what is in the matter info below. Do not fill gaps with guesses.",
    "List any missing info at the end under \"Items to confirm\" as bullet points.",
    "",
    "# Matter info",
    `Name: ${c.name}`,
    `Filing type: ${c.type}`,
    `Stage: ${c.stage}`,
    `Filed on: ${c.appliedOn || "(not entered)"}`,
    `Completion by: ${c.doneBy || "(not entered)"}`,
  ];
  return lines.join("\n");
}

await mkdir(new URL("./out/", import.meta.url), { recursive: true });

let index = 1;
for (const c of cases) {
  const prompt = buildPrompt(c);
  const file = new URL(`./out/case-${index}.txt`, import.meta.url);
  await writeFile(file, prompt, "utf8");
  console.log(`Wrote case-${index}.txt (${c.type})`);
  index += 1;
}

console.log(`Prepared ${cases.length} draft prompts in the out folder.`);

Running it is just this.

node build-prompts.mjs

A text file per matter lands in the out folder, so feed them to Claude Code in turn and the drafts for all matters are ready. Once you’re comfortable, you can widen it so Claude Code reads these files directly and you just say “turn everything in out into update letters.” If you’re shaky on the basics, lock down the Claude Code getting-started guide first and you’ll feel safer.

Where to draw the privacy and security line

This is the part a conveyancing office has to be most careful about, so I’m giving it its own section.

The information you handle, names, addresses, dates of birth, parcel numbers, registration credentials, is the kind that can’t be undone if it leaks. Sending that straight to an outside AI service is something to avoid.

The rule I recommend to my friend is simple. Redact the real identifying info before it ever reaches the AI. The name becomes “Client A,” the address becomes “[City], [District],” and the registration-credential number is never passed at all. Let the AI build the template, and do the merge of real details by hand, inside the office. That’s why the script above used Client A as the name.

On top of that, write these three points into the office rules.

• The 12-character password in the registration credentials goes nowhere: not to the AI, not into a draft file.
• Don’t send a matter’s identifying info to an outside service without the client’s consent.
• Every generated draft is read by a licensed agent before it goes out.

With this setup, all that reaches the AI is “the template for an inheritance-transfer status update,” and whose matter it is never leaves the office. Confidentiality and AI use coexist on exactly this line.

What the AI does, and what a human always decides

In one line, the boundary is this.

• The AI handles: drafting prose, plain-language rewrites of jargon, tone tuning, prep work across several matters.
• The human decides: whether the filing will pass, the required documents and tax amount, legal advice, merging in the real details, and the final check before sending.

Draw this line first and the worst-case accident, “the AI makes a wrong legal call and it reaches the client,” can’t happen by design. The AI is the fair-copy clerk, the judgment is the conveyancer’s. Don’t let that split break, and you’re fine.

When you want to make your everyday use faster, Claude Code productivity tips is worth a look too.

FAQ

Q. Won’t the AI get the law wrong and tell the client something false? A. That concern is correct, and that’s why this article’s design never lets the AI judge whether a filing is valid or what the tax is. It only writes up the prose after the judgment is made. Put “do not fill gaps with guesses” in the prompt and always have a human read before sending. Keep those two and the AI never makes a legal call on its own.

Q. Is it okay to put a client’s personal data into the AI? A. Don’t put it in raw. As in this article, the name becomes “Client A,” the address is redacted, and the registration credentials are never passed. Have the AI build only the prose template, and merge the real details inside the office.

Q. Can I use this if I’m not good with computers or programming? A. For drafting explanations and update letters, you just paste a prompt into chat. The script in this article is for the “I want to process several matters at once” stage, so start with one matter at a time in chat and move to automation once you’re comfortable.

Q. Can I send the output as-is? A. No. The AI output is a draft, full stop. Have a conveyancer confirm that the dates, amounts, and document names match the matter, and that the filing type is correct, before sending. Use this article’s checklist to make that confirmation faster.

Q. Where can I check official information? A. For the underlying rules on land registration, the official guidance from the relevant land registry, such as the GOV.UK land registration guidance, is the primary source. Whether the AI draft matches the actual procedure should be judged by a licensed professional against official information like that.

What I found when I actually tried it

I’m not a conveyancer, so with a friend’s office helping, I had it draft 10 inheritance-transfer status updates from redacted matter notes.

The result was clear-cut. The quality of the prose was about the same as what a paralegal produces by reusing past emails. What differed was time: from having the 10 notes ready to having all the drafts done, including the pasting work, took about 15 minutes. By hand at 15 minutes each, that’s 150 minutes of work.

What I found interesting was that the “list missing info under items to confirm” line in the prompt actually worked. On a matter with a blank completion date, the AI didn’t invent a date in the body; it wrote “please confirm the completion date” at the end. That’s hard for a conveyancer to miss.

On the flip side, when I tried it once without setting the redaction rule first, I nearly pasted a real parcel number and broke into a cold sweat. So I strongly recommend putting “redact identifying info” at the very top of your workflow.

When you reach the stage of rolling this out properly at the office level, with everyone including the paralegals on the same workflow, I can help design the setup around your office’s flow through training and rollout consulting. If you’d rather try it yourself first, the free learning materials are here.