An answer about a contract is worthless without the clause.
iLeaf builds contract lifecycle intelligence for enterprise teams: extracting obligations, dates and risk from executed agreements, and answering questions about them with the clause attached. The retrieval is graph-based rather than flat, because contracts refer to other contracts — an amendment three documents away can invert the answer, and a system that cannot follow that reference will be confidently wrong.
Ask Lia about legal & contract intelligenceWhat this includes
- Obligation & date extraction
- Renewal windows, notice periods, caps and covenants pulled out of executed agreements and put somewhere they can be watched.
- Clause-level retrieval
- Every answer carries the clause it came from. An assertion without a citation is not usable in legal work.
- Graph RAG across documents
- Following the chain from master agreement to amendment to side letter, because the operative term is often not in the document you opened.
- Obligation monitoring
- Alerting before an auto-renewal or notice deadline passes, rather than reporting it afterwards.
- Bulk portfolio review
- Answering one question across an entire contract estate — which of these has an uncapped indemnity — in an afternoon.
- Private deployment
- Contract estates rarely permit external processing. The stack runs inside your boundary where required.
How we run it
Every phase ships something usable on its own, so you are never holding a half-finished system waiting on the next milestone.
Cite or say nothing
The system quotes the clause or states that it cannot find one. It never summarises its way past a gap, because a plausible answer about an indemnity is worse than no answer.
Follow the amendments
Retrieval is graph-based specifically so a superseded term is not returned as current. Flat search over contracts fails on precisely the documents that matter most.
Lawyers decide
The system finds, extracts and flags. It does not advise, and it does not tell anyone what a clause means for their position.
Wrong is expensive here
Recall matters more than fluency. A missed uncapped liability costs more than a hundred over-cautious flags.
What we build it with
- Graph RAG
- Neo4j
- Python
- Document AI
- OCR
- Vector search
- PostgreSQL
- Open-weight models
- On-premise GPU
- Audit logging
- Docker
- OpenTelemetry
Questions we get asked
How is this different from contract search?
Search returns documents; this returns the operative clause and the chain that makes it operative. Contracts amend other contracts, so the current term is frequently not in the agreement you opened. Graph-based retrieval follows master agreement to amendment to side letter, which is the difference between a correct answer and a confidently superseded one.
Does it give legal advice?
No, and it is built not to. It extracts, cites and flags — obligations, dates, caps, unusual terms — and every answer carries the clause it came from. What a term means for your position is a judgement for your lawyers, and the system is instructed to say so rather than opine.
Can it run without our contracts leaving our network?
Yes. Contract estates are among the most sensitive document sets an organisation holds, and the full stack — retrieval, models and governance — can run inside your own infrastructure using open-weight models, with nothing transmitted to an external service.
Let’s talk about legal & contract intelligence.
Tell us what you are running and what it needs to do next. We will tell you honestly whether we are the right team for it.
