Live demo
Legal-Index
Legal research handed over to autonomous agents.
Statute and case law search, contract analysis, drafting support. Every answer points to the article and to the version in force on the date you care about.
Try it onlineIn law, an answer without a source is not an answer.
Search across statutes and case law, contract analysis, deadline and risk-clause tracking. The engine is tuned for legal corpora: version in force, article number, amendments attached to their contract.
- Version in force at a date
- Article and decision cited
- Amendments linked to the contract
What actually costs time in a legal department
Rarely the analysis. Almost always what comes before it.
01
Finding what has already been done
A similar clause was negotiated two years ago, in a contract nobody remembers the name of. Redoing it costs two hours; finding it would cost five minutes if the estate were queryable.
02
Knowing what is in force
The original contract, three amendments, two letters of agreement. Knowing what applies today means reading all five in order, and nothing signals that they all exist.
03
Seeing deadlines coming
Automatic renewal clauses do not remind you of themselves. They are discovered once the notice period has passed.
What we put in place
The legal tuning changes the search itself: exact wording governs, and the effective date is a constraint, not a detail.
Sourced legal research
Statutes, case law and commentary queryable in plain language, with the article or decision cited on every statement.
Contract portfolio analysis
Querying the whole portfolio at once: which penalty clauses, which caps, which competent jurisdictions.
Amendment linking
Contracts, amendments and letters linked together, with the version in force at a given date reconstructed.
Risk-clause detection
Spotting clauses that deviate from your standards, with the deviation highlighted rather than merely flagged.
Deadline tracking
Automatic renewals, notice periods, effective dates: monitored and flagged before it is too late.
Drafting support
Drafting from your own standard clauses and precedents, not from generic templates found elsewhere.
Three uses
01 · Context
The portfolio holds several hundred contracts in a shared space. No cross-cutting question is possible without manual reading.
02 · What we put in place
Document layer with the legal profile, a graph linking contracts, amendments and parties, and structured extraction of deadlines into a tracking table.
03 · Result
Cross-cutting questions are answered in minutes, with the articles cited. Automatic renewals stop being surprises.
01 · Context
Precedent research occupies junior associates, and the result depends on who is searching.
02 · What we put in place
Sourced search across the firm's own estate and external databases, with systematic citation and links back to the originating matter.
03 · Result
The precedent is found regardless of who searches, with the original deliverable and the matter context.
01 · Context
Regulatory change is tracked by manual watch, and the impact on internal procedures is assessed case by case.
02 · What we put in place
Watch over the tracked texts, mapping to the internal procedures concerned, flagging of gaps to address.
03 · Result
The gap between the text and the internal procedure is visible on publication, with the documents to update identified.
What we measure
On a legal corpus, the measurement is about the correctness of the cited source, not only the quality of the wording.
- time to find a precedent within the indexed estate
- -90 %
time to find a precedent within the indexed estate
- of answers citing the article, the decision or the paragraph
- 100 %
of answers citing the article, the decision or the paragraph
- forgotten amendments: linked documents surface with the main contract
- 0
forgotten amendments: linked documents surface with the main contract
These orders of magnitude come from what we measure at pilot on comparable perimeters. On your corpus they are measured before industrialisation, not promised before.
What you can see running today
These products are already built. Some are available online, the others are shown on request.
Live demo
Ultra RAG
The search engine that understands your documents and cites its sources.
Retrieval-augmented search coupled with a knowledge graph, tuned per theme: legal, corporate documentation, timelines. Every answer is traceable down to the original paragraph.
Try it onlineLive demo
CERFA
Administrative forms filled in from your own documents.
The agent reads your files, pre-fills the matching French CERFA form and flags inconsistencies between fields. You review, you sign, you got your morning back.
Try it onlineFrequently asked questions
No. It finds, connects, flags and prepares. Qualification and decision remain the lawyer's, and we turn down projects that claim otherwise.
Effective date, version and linked documents are captured at ingestion, and the legal profile turns them into a search constraint. “What applies on 1 January” is a question the engine understands.
Not necessarily. Deployment on your own infrastructure is common on this kind of corpus, including with self-hosted models.
Yes, via an OCR layer at ingestion. Scan quality has an effect, measured at pilot on your real documents.
Yes, within the limits of your subscriptions and their terms of use. The internal estate and external sources can be queried together.
Legal-Index is our dedicated agent-based legal research product. It is available online, and the document layer can be deployed on your own contract estate.
What this solution does not do
We do not provide legal advice, and an AI system does not replace a lawyer. It finds, connects, flags and prepares — qualification and decision remain human. We turn down uses where the system would decide in the lawyer's place, and we measure its error rate before any deployment rather than estimating it.
How many contracts can you not query today?
Bring an anonymised sample to the diagnostic. We will show you what the legal tuning changes on your estate.