The interesting line in a property file is not the one between a person and a machine. It is the one between a document that has been drafted and a document that has been executed. Everything on the drafting side is production: reading, extracting, reconciling, arranging, quoting, filling fields from a file. Everything on the other side is an act that only a person with capacity can perform, and it is the act the whole file exists to reach.
The distinction matters because a draft can be complete, accurate and beautifully made, and still be incapable of doing anything. A contract that has been drafted to the last schedule but not signed transfers nothing, obliges nobody and can be abandoned by either party without consequence. A transfer that is signed and never registered may leave the buyer owning an instrument rather than a title. The document and the effect are two different things, and the gap between them is a signature, an attestation and, in many places, a formality that has nothing to do with either party’s intentions.
This piece is not legal advice, and none of it is a statement of what any particular jurisdiction requires. It is a description of where the boundary sits in this family of work: what an agent may draft, what a qualified person must sign, and why the signature is not a formality. The position differs by jurisdiction, by instrument and by the capacity of the signer, and where the file does not name its jurisdiction, the correct entry in the column is a question rather than an answer.
The line, stated once
An agent may draft. It may produce a document from the office’s approved template for the jurisdiction in the file, fill its fields from the documents supplied, mark every field it could not fill, assemble the schedules and annexes, and state in the document itself what it is, what it rests on and what remains unknown.
An agent may not do four things that look like drafting and are not. It may not invent operative language: the clause that creates an obligation is the lawyer’s, drawn from a template that the office has had approved, and an agent that writes a new clause has written something nobody has reviewed. It may not supply a legal conclusion, which includes every sentence of the form “this clause permits”, “this notice is valid” or “this document satisfies the requirement”. It may not recall a rule, a period or a statutory reference from general knowledge, because a near-miss on a notice period reads exactly like a correct one. And it may not mark anything final, approved, executed or ready to sign, because those words are claims about an act that has not happened.
The distinction is not modesty and it is not a safety wrapper around a capable system. It is a description of the arrangement. The agent’s output is a draft; the office’s qualified person reads it, corrects it and signs it; and the signature is where the office takes the document on as its own. An agent that drafted flawlessly would still not have signed, and the file would still not be executable until somebody did.
What drafting is, and why it is not advising
Drafting well is a technical skill and it is the one an agent can genuinely hold. It is the ability to take a set of documents and produce a paper that is complete against its own checklist, consistent in its defined terms, accurate in its quotations and explicit about its gaps. That is most of the labour in a property transaction and almost none of the judgement.
Advising is the other part, and it is distinguishable in a way that matters in practice: advice is a statement about what the client should do, or about what a document will do when tested. Whether a break is operable on the date the tenant intends, whether a review mechanism produces the increase the landlord claims, whether a covenant binds a successor, whether a signature in a particular form is effective — each of those is a conclusion, drawn by a person who is answerable for it, and each depends on the whole instrument and the law that applies to it.
The two are easy to confuse because they arrive in the same paragraph. A draft reply to a pre-contract enquiry contains the seller’s answers, which are the seller’s statements, and it also contains a sentence about whether the property benefits from a right of way, which is a conclusion the lawyer has to draw. An agent can assemble the first from the file; the second belongs to somebody else, and the document should show which is which rather than smoothing them into one voice.
That is why the family’s documents carry their status on their face. A draft is marked as a draft, with the date and the fields that remain open. A quotation carries its clause reference and page. A conclusion, where one appears, is attributed to the person who drew it. The marking is not decoration: it is the mechanism by which the qualified reader knows exactly which sentences to check, and it is the difference between a draft that saves an hour and a draft that costs one.
The documents, and who signs each
The six artefacts of the family divide cleanly, and the division is the one the rest of the transaction inherits. Five of them are working documents that nobody signs; one is an instrument that two people do.
Table 1. The artefacts of a property file and the instruments around them: what an agent may produce, who must read it, who signs, and what the signature adds. Source: constructed for this piece; the signatory column is a shape, not a schedule of any jurisdiction’s requirements.
| Document | What an agent may produce | Who must read it before it leaves | Who signs | What the signature adds |
|---|---|---|---|---|
| Comparable evidence table | The whole table, from comparables the office supplies | The person defending the figure | Nobody | Nothing: it is evidence, not a statement of value |
| Market assessment | The whole document, with its range, weights and gaps | The principal, before it reaches a client | Nobody | Nothing, provided it says on its face that it is not a valuation |
| Lease abstract | Quotations with clause references and pages | The adviser reading the lease | Nobody | Nothing: interpretation is the adviser’s |
| Pre-exchange schedule | The schedule, with statuses, owners and dates | The person running the transaction | Nobody | Nothing: it records a process |
| Vendor presentation deck | Slides, each claim traced to the file | The seller, who approves the facts | Nobody | Nothing, but the seller issues it, and issuing is a statement |
| Client proposal | A draft from the approved template, fields from the file | The principal | Client and principal | Creates the engagement and the fee obligation |
| Heads of terms | A record of what the parties said, from the file | A lawyer | Buyer and seller | Records intent; binds only to the extent the document says so |
| Replies to pre-contract enquiries | Draft answers assembled from the file | The seller and the lawyer | The seller | Converts an assembly into the seller’s own statement |
| Contract of sale | The schedule and the parties’ details; not the operative clauses | The lawyer | Buyer and seller | Binds, and attaches a remedy to breach |
| Transfer or deed | Schedule insertions from the file; nothing else | The lawyer or notary | Both parties, with whatever attestation the jurisdiction requires | Passes the interest, if the formality is complete |
| Statutory declaration or certificate | Nothing substantive | The declarant’s adviser | The declarant, before a commissioner or notary | Attests a fact on oath or affirmation, with consequences for a false one |
| Valuation report | Nothing: it is not the agent’s work at any point | — | A qualified valuer | Carries the standing the licence confers and the liability that goes with it |
| Lender’s or landlord’s consent | The request pack and the schedule entry | — | The third party | Removes an obstacle that no drafting can remove |
Two things follow from the table and they hold whatever the jurisdiction. The first is that the signature column is short: five of the thirteen rows are signed by nobody, and of the eight that are signed, seven involve somebody other than the office — the eighth is the engagement, which the office signs alongside the client. The second is that the agent’s contribution is concentrated where no signature is required — which is not a comment on its value but on the kind of act signing is.
What a signature does
A signature is treated as a formality because it is usually quick, and it is not a formality because of what it does. Four things are carried by it, and they are separable.
Attestation. The signer adopts the content as their own statement. A reply to a pre-contract enquiry signed by a seller is not the office’s assessment forwarded to a buyer; it is the seller’s answer, and a buyer who relied on a wrong one has a remedy against the seller. This is the effect that an unsigned draft cannot produce however accurate it is, and it is why a draft that has been assembled well is still only a proposal about what the signer might say.
Authority. The signature represents that the person had the capacity to bind the party. A company signs through a director or an attorney acting under a power; co-owners sign where title is held jointly; an executor signs where the seller is an estate; a person signing under a power of attorney signs in that capacity and the instrument has to show it. A document signed by somebody without authority is not a weaker document; it is a document that may bind nobody, and the defect is invisible on its face.
Formality. In many jurisdictions the signature alone is not enough for certain instruments. A deed may require a witness, or a notary, or both. A transfer may have to be registered before the interest passes, so that a completed and signed instrument leaves the buyer with a right to be registered rather than a title. Some instruments require a particular form of words, a certificate, an apostille or a stamp, and the requirement is about validity rather than proof. Where a formality is missing the document may be perfectly genuine and simply ineffective, which is the most expensive kind of defect because nothing about it looks wrong.
Recourse. The signature identifies who answers for the statement. This is the one that changes how the rest of the file behaves. A valuation is signed by a qualified person because the liability, the insurance and often the regulatory standing attach to that person; a certificate is signed by the professional who certifies; a completion statement is signed by the lawyer who prepared it. The signature is how a document acquires an owner, and a document without one is a document nobody is answerable for.
The four are worth separating because a file can have three of them and fail on the fourth. A contract signed by both parties, with no authority behind one of the signatures, has attestation and formality and no binding effect. A transfer signed, witnessed and never registered has everything except the step that actually moves the interest.
Where the position differs by jurisdiction
This is the part of the boundary that no agent can generalise, and the honest statement of it is short: the same document can be effective in one jurisdiction and ineffective in another, and the difference is not a matter of degree.
Some jurisdictions require a transfer of land to be notarised; others accept a signature before witnesses; others require neither for the instrument but treat registration as the moment the interest passes. Some permit an electronic signature on a commercial contract and not on a deed, or on a deed only with a particular kind of attestation. Some require a statutory declaration to be sworn before a commissioner or a notary, and a declaration signed without that step is not a declaration. Some regulate who may produce a valuation for a lender, some regulate who may act as an agent for reward at all, and some require a lawyer or a notary to prepare the instrument rather than merely to review it. Consumer-protection rules in some places impose cooling-off periods and prescribed disclosures on particular transactions, which change what a signature means rather than merely how it is obtained.
The consequence for this work is a rule rather than a list. Every document states its jurisdiction, its instrument and the date the requirement was checked, and where the file does not supply a source, the row records the question and who must answer it. An agent that fills that gap from general knowledge has produced the one kind of error that a careful reader cannot catch, because a recalled requirement is indistinguishable from a verified one on the page.
The ledger, worked
The way to hold the boundary in a live file is a ledger with one row per document and four columns: who drafts it, who reads it before it leaves, who signs it, and what formality the signature requires. The ledger is not the contract and it is not advice; it is the record that shows, at any point, which documents are drafts and which are executed.
The example below is a constructed sale of a constructed asset: sixteen documents, of which eleven carry a signature and five do not.
Table 2. An execution ledger for a constructed sale: sixteen documents, the signatures each requires, and the attestations beyond a signature. Source: constructed for this piece; the formality column describes the shape of a requirement, not the requirement of any named jurisdiction.
| Document | Who signs | Party signatures | Attestations | What the signature does |
|---|---|---|---|---|
| Client proposal | Client and principal | 2 | 0 | Creates the engagement |
| Heads of terms | Buyer and seller | 2 | 0 | Records the parties’ intent |
| Contract of sale | Buyer and seller | 2 | 0 | Binds both parties |
| Transfer or deed | Both parties | 2 | 3 | Passes the interest, subject to formality and registration |
| Replies to pre-contract enquiries | Seller | 1 | 0 | Becomes the seller’s statement |
| Statutory declaration | Declarant | 1 | 1 | Attests a fact on oath or affirmation |
| Lender’s consent | Lender | 1 | 0 | Releases an obstacle |
| Valuation report | Qualified valuer | 1 | 0 | Carries the valuer’s standing and liability |
| Energy performance certificate | Assessor | 1 | 0 | Certifies the assessment |
| Service charge certificate | Managing agent | 1 | 0 | Certifies the apportionment |
| Completion statement | Seller’s lawyer | 1 | 0 | States the sums due |
| Comparable table, market assessment, lease abstract, vendor deck, pre-exchange schedule | Nobody | 0 | 0 | Nothing: five working documents |
| Totals | 16 documents, 11 executed, 5 unsigned | 15 | 4 | 19 execution marks |
The arithmetic is the point of the ledger. Sixteen documents; eleven executed and five not, which is 69 per cent executed by count. Fifteen party signatures across the eleven instruments, and four attestations beyond a signature — two witnesses to the deed, a notary to the same deed, and a commissioner for the statutory declaration — for nineteen execution marks in total. Nineteen marks over eleven instruments is 1.73 each, and the distribution is not even: the deed alone carries five of the nineteen, 26.3 per cent, because it is the one instrument that both transfers the interest and attracts a formality. The other ten instruments average 1.4 marks each and none of them needs more than a signature.
The same division runs through the cost. The agent drafts three of the eleven executed instruments — the proposal, the heads of terms as a record of what the parties said, and the draft replies to enquiries — and it drafts all five working documents. The other eight executed instruments are the lawyer’s to draft or to obtain: the contract, the transfer, the statutory declaration and the completion statement on the conveyancing side, and the valuation, the two certificates and the lender’s consent from the professionals and third parties who produce them. The lawyer reviews all eleven. That is not a division of labour the agent chose; it is the division the four effects of a signature impose.
The cost of the boundary
The metered spend sits almost entirely on the unsigned side of the line, and the arithmetic says so plainly. One deal’s document set is fifty paid model calls at the measured rate, and the six artefacts divide by which side of the line they fall on. Five of them — comparable table, market assessment, lease abstract, pre-exchange schedule and vendor deck — are working documents that nobody signs: eight, eight, twelve, six and ten calls, forty-four in total. The sixth is the client proposal, which two people sign, at six calls.
So six of the fifty calls, 12 per cent, produce a document that will be signed, and the other forty-four, 88 per cent, produce drafts of documents that will not. At the measured DeepSeek rate on Base — $0.00700 a short-prompt call under exact — the signed side is $0.04200 and the unsigned side $0.30800, which together are $0.35000. At the metered ceiling of $0.00955 the same split is $0.05730 and $0.42020, and $0.47750 in total.
That ratio is the most useful thing this piece can say about cost, and it is not an argument for spending less. It says that the metered work sits almost entirely on the drafting side of the line, and that the executed side of a transaction is paid for in somebody’s hours rather than in model calls. The figure that circulates as DeepSeek’s price, $0.00875, is Qwen’s measured rate and not DeepSeek’s: at it the same fifty calls would read $0.43750, and the forty-four unsigned ones $0.38500. No part of that difference touches the cost of a signature.
What the signature cannot do
The limitation that belongs beside this argument is that a signature attests and does not verify. A person can sign a document that is wrong, having read it, and the signature makes the document the signer’s statement without making it true. Everything the office relies on — that the fields were filled from the file, that the quotations are accurate, that the arithmetic reconciles — is a property of the drafting and the review, and none of it is created by the act of signing.
That has a practical edge in this family of work, because a well-built draft is more consistent than a tired junior’s, and consistency invites trust. A reader who has seen twenty accurate abstracts from the same office will read the twenty-first quickly, and the signature at the end of it does not slow them down. The four effects described above are why the signature still matters; they are not a substitute for reading the document before it is signed, and no ledger column records whether anybody did.
The second limitation is that the boundary is described here as an arrangement, not enacted by anything. An agent’s ceiling on drafting is a property of its specification and of the office’s procedure; it is not a legal restriction and it does not bind a person using the same tools differently. Nothing in this piece regulates anybody.
The third is the one this publication keeps returning to. Everything above is jurisdiction-neutral by construction, which means it is also incomplete by construction. The ledger’s sixteen rows, its eleven executed documents and its nineteen marks are one constructed transaction’s; another jurisdiction’s transaction would have a different count of documents, a different set of formalities and possibly a different set of signatories altogether. Where a live file’s ledger differs from this one, the live file is right and this piece is a shape.
The decision at the end
The decision is a gate, and it is applied to every document rather than to the pack. No document leaves the office described as final, and no instrument is presented for signature, until four things are true of it: it names its jurisdiction and the instrument it was drawn from; it names the person who will sign and the capacity in which they sign; it states the formality the jurisdiction requires for that instrument, or records that the requirement is unverified; and it says on its face whether it is a draft, a statement attributed to a named person, or an instrument awaiting execution.
A document that fails any of the four stays a draft. That is the whole test, and it is deliberately unfashionable: it makes the file slower, it produces a page of open questions where a confident document might have gone, and it is the only version of the process in which nobody has to guess later who was answerable for a sentence.
Where this goes
- The pillar: x402.review/2026/10/31/property/ — the six documents and the five boundaries this piece works inside
- The role: studio.x402.press/roles/real_estate_assistant — the agent, and what it is specified not to do
- The service it buys: x402.press/services/deepseek/ — the canonical price
- The review: x402.review/2026/10/31/deepseek-review/ — the measured handshake
- The schedule these sixteen documents come from: sale-agreement-document-checklist — who produces each document and what stalls without it
- The abstract that must not interpret: lease-abstraction-checklist — the line between extracting and concluding
Method note
The price was read from the live 402 handshake at the DeepSeek endpoint on Base on 31 October 2026: $0.00700 for a short prompt under exact, $0.00955 for a long prompt under exact and as the upto ceiling at max_tokens: 1000. The ledger reconciles in both directions: 11 executed plus 5 unsigned is 16 documents; 15 party signatures plus 4 attestations is 19 marks; the deed’s five marks divided by 19 is 26.3 per cent; and 6 calls at $0.00700 is $0.04200, 44 calls is $0.30800, and the two are $0.35000. The documents, the parties and the transaction are constructed so the arithmetic can be followed; no party, property or jurisdiction is described, and none is real. Nothing in this piece is legal advice, and no jurisdiction’s requirements are stated as fact: the four effects of a signature are described in general terms, the differences between jurisdictions are described as differences, and the instrument and its jurisdiction govern in every live file. The agent drafts and assembles; a qualified person reads, concludes and signs.
x402.review is written and edited by N.P. Vincent. The publication is connected to x402.press, which sells the services reviewed here; that interest is disclosed on every piece that touches a service, and it is stated here rather than left to inference.
Filed under Jobs · Tags: property, agent-limits, signing, documents
