---
title: Motex — Autonomous legal operations
url: https://motex.ai/en
description: Motex organizes the litigation portfolio of a company or a law firm: it pulls the court file from the Chilean judiciary, classifies every document, computes deadlines and makes the portfolio queryable.
---

# Autonomous legal operations

> Motex takes what the Chilean judiciary already publishes, organizes it case by case and makes
> it queryable. The point is that a legal team stops managing folders and starts deciding with
> data.

> Human-readable page: https://motex.ai/en

## What Motex is

- A platform for governing litigation. Used by in-house legal teams and by law firms.
- It downloads the documents the court publishes and stores them in their case, ordered by folio.
- It extracts what matters from each complaint: parties, prayers for relief, amounts, subject matters, deadlines and hearings.
- From that it builds a single view of the portfolio: exposure, deadlines, hearings and firm performance.
- It deliberately keeps the judiciary's own vocabulary: folio, cuaderno, stage and procedural step.

## AI detection in legal filings

- A lawyer uploads a filing — their own or the other side's — and gets the text marked paragraph by paragraph.
- Three categories: human, AI-assisted and AI-generated, plus an overall percentage of human text.
- Input is a PDF with a text layer, up to 60 pages, or pasted text. A scanned PDF with no text layer is rejected: this flow has no OCR.
- The analysis takes between 5 and 30 seconds.
- Every analysis is stored in "Reviewed documents" and reopens without re-running. The extracted text is stored, not the original PDF.
- Available to law firms and individual users.

## The problem it solves

- When an organization faces thousands of lawsuits, the portfolio becomes a black box.
- Cases live scattered across PDFs, email and spreadsheets.
- Nobody knows what each case costs, or which outside firm performs.
- Decisions are made blind and the legal budget only grows.

## How it works

- 01 · Intake — Cases are connected from wherever they live: PDFs, email, internal systems and the judiciary.
- 02 · Structure — AI extracts and organizes each case into queryable data: amounts, deadlines, subject matter, firm and status.
- 03 · Monitoring — Every case stays tracked: deadline alerts, detection of new documents and fraud prevention.
- 04 · Recommendations and drafts — Strategy recommendations and draft filings for the lawyers to file and follow up.

## What Motex extracts from a complaint

- Parties: plaintiff, defendant, tax ID, counsel.
- Prayers for relief: every claim, one by one.
- Amounts: the exposure of the case, in pesos.
- Subject matters: dismissal, fundamental-rights protection, debt collection, nullity.
- Deadlines: answer, appeal, remedies.
- Hearings: into the team calendar.

## Capabilities

- Structured data — Every case ordered and queryable at once.
- Automatic tracking — Each case monitored, with its status always current.
- Deadline alerts — No missed hearing and no default from a lapsed deadline.
- Fraud detection — Anomalies and patterns by cost center, seasonality and history.
- Amount prediction — An estimate of the likely outcome of each case, for better provisioning.
- Executive dashboard — The whole litigation portfolio in a single view.
- Firm performance — Which outside firm wins by case type, and who misses hearings.
- Fee management — Control over outside counsel fees.

## The case file

- It lives inside the case record, under the "Movimientos" tab ("Historia" in appeals and collections).
- A table grouped by cuaderno, one row per folio: folio, document, attachments, stage, procedural step, date and status.
- Every document carries a type label from 32 classes. The label can be corrected by hand.
- A "show only relevant documents" filter.
- A fixed side viewer, not a modal: the document is read without losing the table, and toggles between PDF and text.
- A paragraph can be highlighted and shared by link: whoever opens it lands on that same passage.
- The PDF text layer is read first; 58% of documents resolve that way. When it is not enough, Amazon Textract runs.

## Scale measured in production

- 447,283 court documents downloaded and stored in their case.
- 331,741 docket movements recorded.
- 10,099 active cases, about 44 documents per case on average.
- 267,410 document classifications.
- 84% of the corpus has extracted text. The remaining 16% does not, and the interface says so.
- 204,088 pages run through OCR in the first 12 days of August 2026.

## What Motex does not do

- There is no full-text search across the whole case file. Search is by RIT, case caption or client.
- There is no bulk download or ZIP of a complete case. Documents are downloaded one at a time.
- 16% of the corpus has no extracted text.

## Portfolio queries

- Chat over the cases. It answers by outside firm, by business unit, by procedure or about one specific case.
- It reads a document from the case file and returns the folios of a case.
- It finds comparable cases in the same portfolio, with a score weighing court, procedure, subject matter, opposing counsel, amount and location.
- It shows the agenda: pending hearings and deadlines, overdue first.
- It answers on Chilean law with the verbatim article. Nine codes and statutes are indexed.
- Every statement carries its source number, and a side panel opens the case, the document or the statute next to the conversation.
- Every number comes from a system tool, not from the model's memory.
- If the model paraphrases instead of quoting verbatim, the system answers "quote not found" instead of inventing a highlight.
- Enabled per organization.

## Security

- Isolation per company: each organization runs in a sealed environment.
- Client information is never used to train third-party models.
- Encrypted in transit and at rest.
- The data belongs to the client: it can be exported or deleted on request.

## Related machine surfaces

- [/m/en/docs/quick-start.md](/m/en/docs/quick-start.md): what a client needs so their cases enter Motex
- [/m/en/docs/mcp.md](/m/en/docs/mcp.md): query the portfolio from the client's own model provider
- [/en/llms.txt](/en/llms.txt): machine site map