A telecom operator reviews thousands of site leases with private Gen AI
Tower and rooftop leases were scanned PDFs in many formats and languages. A private Gen AI model extracted rents, escalations and renewal dates, with every value linked to its page, so the legal team checked rather than read.
An illustrative scenario. The company is fictional and the figures show how the method works; they are not a client result.
The situation
A mobile operator rents land and rooftops for thousands of tower sites. Each lease was negotiated locally, over many years, in different formats and languages. Most existed only as scanned PDFs.
Finance paid rent from a lease register that had drifted from the contracts. Escalations were applied inconsistently, renewal dates were missed and nobody could say with confidence what the contracts actually required.
What we did
- Deployed LawRence, our private Gen AI for legal documents, inside the operator’s own cloud environment. No document left it.
- Extracted the terms that drive cost and risk: rent, escalation rate and frequency, start and end dates, renewal and termination rights, and the landlord’s bank details.
- Linked every extracted value to the page and clause it came from, so a reviewer could check it in seconds.
- Compared the extracted terms with the lease register and two years of rent payments.
What we found
| Finding | What it meant |
|---|---|
| Rent paid above the contracted amount | Escalations applied early or at the wrong rate |
| Escalations paid on leases that had none | Overpayment |
| Leases past their end date, still paid on old terms | Renegotiation opportunity |
| Landlord bank details that differed from the contract | Payment risk to verify |
| Renewal and notice dates not in any calendar | Risk of unwanted automatic renewals |
How the legal team worked
The model proposed; lawyers decided. Each extraction came with its source, and reviewers accepted, corrected or rejected it. Corrections were fed back, and the share needing changes fell as the review went on. The team spent its time on the leases that mattered most, not on reading every page.
What changed
- A register built from the contracts, with a link to the source for every term.
- Payments checked against the contract before each rent run.
- A calendar of renewals and notices, so each decision is made on time.
Read Private Gen AI for legal teams: ten questions to ask before you buy, or see contract review.