How LEASE-iQ works
Built around the lease.
Built around accuracy.
You ask a question in ordinary language. Behind the answer is a process designed to read the document, connect its meaning, bring in relevant legislation and challenge the result. You can then see the evidence and decide what to do.
See it workingThe approach, in plain English
From a difficult document to a useful answer.
LEASE-iQ uses generative AI within a workflow designed for leases. The difference lies in how the document and legal context are prepared, how answers are checked and how you can inspect the sources.
Read the scan
Leases can be old, badly scanned and awkwardly laid out. The document is processed into text and structure that the system can work with. The original PDF remains available, so you can compare the reading with the actual page.
The technical term: parsing
Parsing turns the document into usable text and identifiable sections. An unclear character, missing page or poor scan can still create uncertainty. The original image is the reference when the reading matters.
Connect the parts of the lease
“Property” may have a special definition. A repair promise may have an insurance exception. A schedule may change who pays. Those connections matter to the answer.
The technical term: lease ontology
An ontology is a map of the things in a lease and how they relate: parties, property, definitions, obligations, permissions, schedules and cross-references. It gives the question its document context.
Bring in relevant legislation
The lease is part of the picture. Legislation can add conditions or affect how its wording operates. LEASE-iQ brings relevant legal sources into the analysis and gives you a route to open them.
Why legal context matters
A repair obligation and the right to recover its cost are separate questions. The applicable law, jurisdiction, dates and facts can affect the answer. Later deeds and other governing documents also need to be supplied and considered.
Choose the depth of checking
Sean describes different routes for simple, medium and hard questions. A straightforward pet question can take a shorter route. Questions about money owed or legal disputes are directed to the more intensive route.
The technical term: question classification
The system assesses the question and applies routing rules. The route changes the amount of answer comparison and challenge review. Classification itself can still make mistakes.
Test and challenge possible answers
For more involved questions, the “juror” process compares candidate answers. A separate challenge process looks for faults in the draft and can send it back for revision.
Jurors and adversarial review
These are AI processes, not a panel of people. In Sean’s 9 October explanation, candidate answers are compared for similarity before challenge review. The challenge starts by trying to find what is wrong. Review has limits; agreement does not prove that a shared interpretation is correct.
Check the quotation against the source
Sean also describes checks performed by code: does the quoted wording occur in the lease or legislation source? This provides a separate check on the words used to support an answer.
The technical term: deterministic checks
A source-text check follows a programmed rule rather than asking AI for another opinion. Finding the quoted words does not, by itself, prove the interpretation is complete or that the provision applies to your situation.
What you get
An answer you can follow back to the evidence.
Understand the explanation, open the source, ask a follow-up and prepare your response in the same conversation. Ask LEASE-iQ how a letter might be received, then refine it. If the documents do not support a firm answer, the useful result is knowing what is still missing.
See the poor 1963 scan and the uncertainty in the Newsteer example →
Trust and verify
Match your check to the decision.
A straightforward question
For a clear pet clause, read the answer and any permission conditions, then get on with the next step. A disputed pet restriction would deserve more attention.
Money, major work or a dispute
Open the original lease and relevant legislation. Check the conditions, figures and facts yourself before committing. You choose whether a remaining issue needs specialist advice; LEASE-iQ can prepare the brief if it does.
Robust checks reduce risk. They do not create certainty. No AI or lawyer can promise 100% accuracy. Personal validation is part of staying in control, rather than a compulsory handoff to a colleague.
Why this is different
A workflow built for lease questions.
Generative AI supplies language and reasoning capabilities. LEASE-iQ adds a lease-specific structure, legal context, answer-checking processes and a source-viewing workflow around those capabilities. That is the distinction we want you to inspect in the demonstration.
The architecture explains the approach. It is not a numerical accuracy score or a claim that LEASE-iQ beats every general-purpose AI on every question.
About this explanation
Based on Sean’s explanation of the system.
This page summarises the architecture described by Sean Gilley, co-founder and CTO, including his explanation of question routing, jurors, challenge review and source-text checks on 9 October 2026. The stages show the design approach; they are not a record of the checks run on an individual answer. The checking route varies with the question and product version.
Updated 13 October 2026. Read the practical source-checking guide or watch the real examples.
