Documents written over 100 years ago — can Hobson really read them?

The question was simple:

Not a clean ten-page lease signed last year. The difficult end of the archive.

Old scans. Different typefaces. Dense drafting. Language lawyers stopped using decades ago. Handwritten amendments. Plans and schedules. Parties whose names and legal forms have changed. A right granted in one document but qualified somewhere else. A lease that refers back to an earlier lease, a historic deed and a separate development agreement before you have even reached the operative provisions.

And age is only part of the problem.

One of the leases in the portfolio runs for more than 150 years. Its rent is not simply a number printed on the first page. The document contains a premium, formula-driven rent provisions, overage and clawback arrangements, redevelopment rights and detailed provisions governing things as specific as tower cranes and ground anchors.

Another generation of leases uses traditional drafting such as "NOW THIS DEED WITNESSETH", with repairing obligations that work on five- and seven-year cycles and detailed provisions about exactly how premises are to be handed back.

Then there are the documents behind the document: earlier leases, underleases, historic deeds, later agreements and plans — each capable of changing the answer.

That is the real challenge.

It isn't whether AI can recognise old words on a page. It is whether it can understand what kind of fact each word creates, what it relates to, whether something later changed it, and where the evidence for the answer actually sits.

For an experienced property team, "nearly right" would be immediately obvious — and no use at all.

Hobson reads these documents as property documents, not simply as pages of text.

It can take a long historic lease and separate the things that matter: the parties, the term, the rent, the premium, review machinery, repairing obligations, rights, restrictions, redevelopment provisions, notices, assignments, underletting and the documents the lease itself depends upon.

Crucially, it keeps those things distinct.

A £22 million premium is not rent. Overage is not an amount automatically due. A 153½-year contractual term is not evidence that occupation actually lasted 153½ years. A reference to an earlier deed is not background decoration if that deed governs a right the current lease still relies upon.

And where an answer runs through several places, Hobson follows the chain.

So instead of somebody knowing that "there's something about that in one of the schedules", they can ask:

• What's the term?
• How is the rent actually calculated?
• What can the tenant do on redevelopment?
• Are there restrictions on underletting?
• Who is responsible for this repair?
• What earlier document does this right depend on?

Hobson returns the answer with the wording and source behind it.

That matters enormously in an institutional portfolio. The people managing it already have expertise. Hobson is not there to replace that expertise; it is there to make a century of accumulated property knowledge immediately usable by the whole team, without every question having to find the one person who remembers where the answer lives.

And perhaps most importantly, age does not give a document permission to become vague.

If Hobson can establish the answer, it shows it.

If the document does not establish it, Hobson says that too.

Old law. Old paper. Same standard of proof.

Reading the archive is the foundation.

The bigger opportunity is what happens when generations of property history stop being an archive and become a working record.

For a portfolio like this, that means the knowledge held across leases, deeds, schedules and historic papers can ultimately help the team stay ahead of what comes next — rather than only being opened when somebody already knows there is a question to ask.