Every term, with the clause behind it
Dates, rent steps, lease type, options, TI and guarantor language, pulled out of the executed lease and every amendment stacked on top of it — each one shown against the sentence it came from.
An extracted term you cannot trace is a term you have to re-read
Any tool can put a number in a box. The question a careful reader actually has is where it came from, and that is the question this answers by default rather than on request.
- The sentence, and the pageEvery value carries the clause it was read out of and where it sits in the document. Verifying one is a glance, not a hunt through forty pages.
- Amendments stackThe lease and its amendments are read together, so a rent step added by a third amendment supersedes the original schedule instead of sitting beside it.
- Long documents do not get truncatedA lease too long to read in one pass is split into passages and read across all of them. A clause on page fifty-one is found the same way one on page two is.
Cardinal Freight Systems — Lease + 2 amendments
42 pages · extracted 8 terms
- Commencement1 Jul 2021
“The Term shall commence on July 1, 2021 and continue for one hundred twenty (120) months.”
p.3 §2.1
- Expiration30 Jun 2031
“…expiring at 11:59 p.m. on June 30, 2031, unless sooner terminated.”
p.3 §2.1
- Base rent$7.85 / SF
“Annual Base Rent of Seven and 85/100 Dollars ($7.85) per rentable square foot.”
p.4 §3.1
- Lease typeNNN
“Tenant shall pay its Proportionate Share of Taxes, Insurance and Common Area Maintenance.”
p.7 §6.2
The terms are not the deliverable. The rent roll is.
Abstraction on its own produces a summary. What makes it worth doing is that the terms land in the rent roll, and the rent roll is what every number in the deal is derived from.
- Suite by suite, with its own scheduleEach suite keeps its own dates, stepped rent and lease type, so a mid-year increase is modelled as a mid-year increase rather than averaged away.
- NOI is derived, never enteredBecause the income comes from the table, a figure here reconciles to an operating statement instead of merely looking plausible.
Suites
Add Suite| Status | Suite | Tenant | SF | $/SF/Yr | Exp |
|---|---|---|---|---|---|
| Occupied | Suite 100 | Cardinal Freight Systems | 78,400 | $7.85 | Jun 2031 |
| Occupied | Suite 200 | Meridian Cold Storage | 52,000 | $8.40 | Mar 2029 |
| Occupied | Suite 250 | Halstead Components | 31,600 | $7.20 | Nov 2027 |
| Vacant | Suite 300 | — | 22,500 | — | — |
Every row came out of a lease. NOI is derived from this table, never entered.
What we do not claim
Extraction is AI, and it is not perfect. We would rather say that here than have you discover it on a lease that mattered.
There is no human review tier. Some platforms put a CRE analyst on every document and stand behind an accuracy figure. That is a real service and it is a reasonable thing to buy if being able to point at somebody else is the point — see how that compares.
What we do instead is refuse to hide the working. Every term shows its clause, so the check is cheap enough to actually perform — which is a different promise from “you will never need to check”, and one we can keep.
Questions people ask
What does lease abstraction extract?
Commencement and expiration dates, base rent and its stepped schedule, lease type, renewal and termination options, tenant improvement allowances, recovery terms and guarantor language. Each one is shown with the clause and page it came from.
Does it read amendments?
Yes, and that is most of the work. A lease and its amendment stack are read together, so a rent step added by a third amendment supersedes the original schedule rather than sitting beside it. Reading only the base lease is how a rent roll ends up confidently wrong.
How accurate is it?
Good, and not perfect — extraction is AI and we are not going to claim otherwise. What we do instead of claiming accuracy is show our working: every extracted term displays the sentence it came from and the page it was on, so verifying one takes a glance rather than a re-read. Platforms that put a human reviewer on every document exist and are a reasonable choice if somebody else being accountable for accuracy is the point.
What about scanned leases with no text layer?
They are read as images. Long documents are split into passages and extraction runs across all of them, so a clause on page fifty-one is found the same way one on page two is.
What happens to the extracted terms?
They become the rent roll, which is the system of record for the whole deal. NOI is derived from it rather than entered beside it, so every number downstream traces back to a clause in a document you uploaded.
Where this goes next
Upload & auto-build
How the documents get in, and what happens to each kind.
All capabilities
Everything else the product does.
What is a rent roll?
The table these terms build.
Lease guaranties
The clause behind the rent, and why a blank column is not a denial.
NNN, gross and modified gross
The lease type that decides who pays what.
Compared to Prophia
Where a human-audited abstraction service fits instead.
Drop in a lease and see what it finds.
Upload an executed lease and its amendments. DealWise reads them, shows you every term against the clause it came from, and builds the rent roll from what you approve. Free plan, no credit card.