Can AI help us review a lease on a property matter?

Yes, for the mechanical half of the work. An AI assistant can read a long lease and the papers around it, pull out the provisions you ask for, point you at the page each one came from, and tell you where the documents say nothing. What it cannot do is decide whether a provision is acceptable for this client, read the register for you, or sign the report.

Split the job into extraction and judgement

Most lease reviews are two tasks wearing one coat.

The first is extraction and cross-checking. Who are the parties, what is the term and when did it start, what does the rent review clause say and when does it bite, how is the service charge apportioned, what are the repairing and reinstatement obligations, what does the alienation clause permit, is there a break and what conditions attach to it, what does the user clause allow, how does the insurance provision work, what does the lease say about forfeiture. Then the same provisions checked against the agreement for lease, the licence to assign, any deed of variation, the replies to enquiries and the title.

The second is judgement. Is this acceptable for this client, for this use, for this lender. Is anything here unusual enough that the client needs to be told plainly. What should be negotiated and what is not worth the fight.

AI is useful on the first task and unsafe on the second. Keeping them apart in your own head is what makes the tool worth having.

Where it genuinely saves time

• Finding every mention of a defined term across a long lease and its schedules, including the one buried in the eighth schedule that contradicts the fourth.
• Cross-referencing between documents: the plan reference in the lease against the transfer, the names and dates against the title, the rent in the lease against the rent in the licence.
• Flagging silence. A lease that says nothing about a point is harder to spot than a lease that says something awkward, because there is no text to catch your eye.
• Working through a thick bundle of enquiries and replies and listing which questions were answered, which were deflected and which were not answered at all.
• Producing a dated list of the time limits the documents create, with the clause behind each one set out so you can check it.
• Reading a variation or a supplemental deed against the original and telling you which clauses moved.

On a file with three hundred pages of lease, deeds and correspondence, that is an hour of work rather than a morning, and you start the judgement part already knowing where everything is.

What stays with you, without exception

The register and the plan. Official copies come from HM Land Registry and you check them yourself. An AI tool reading a copy of a document on your file is reading your file, not the register, and it has no way of knowing whether what you gave it is current.

Searches and their results. The same point applies. If it is not in the papers you supplied, it is not in the answer.

Anything that turns on market practice or on what the client wants. Whether a provision is onerous for this tenant, whether a lender will accept it, whether the client will live with it: that is professional judgement and the file needs your view on it, not a machine's.

Anything that turns on what the client was told. The AI sees documents. It does not know what was said on the phone in March.

The report itself. A report on a lease is advice to a client. A draft can be prepared, but a qualified person reads it, corrects it and sends it.

A routine that holds up afterwards

1. Give it the complete set and name each document as you go: the lease, every supplemental deed, the agreement for lease, the title, the replies, the plans. A review built on three of the five documents is worse than no review.
2. Ask for extraction first, with a page reference against every item. Do not ask for a view.
3. Ask what the papers do not deal with, as a separate question. The gaps are where the risk usually sits.
4. Open the page behind each extracted provision and read the clause yourself. This is quick when the reference is exact and it is the step that catches the errors.
5. Check every date and figure against your own diary entry and the source document. The tool's arithmetic is a prompt to check, not a substitute for checking.
6. Note on the file what you asked, what you checked and what you concluded. A reviewer six months later should be able to see where your conclusions came from.

If a page is illegible, a plan is a poor scan or a schedule is missing, you need to know that before you rely on the summary rather than after.

Where Alesis fits

Alesis is an AI assistant for UK law firms, made by L25 Limited, used through the web browser, with one conversation for a matter. It answers questions about a matter from that matter's own papers and names the page each answer came from; if the papers do not say, it says so, and any page it could not read is flagged rather than skipped. It counts key dates and figures with the working shown, each step carrying the rule that allows it, and says the firm's diary system still governs. It prepares drafts for a qualified person to review and sign off, and it assists qualified professionals rather than replacing them.