Drop a ZIP. We classify it.
The biggest friction in REPSE compliance isn't the audit — it's the vendor uploading 16 documents into the right portal, with the right name, in the right quarter. Vigía Legal solves it with a single box: drag the ZIP, our AI classifies, extracts the critical fields and builds the file.
- constancia_repse_2026.pdfREPSEREPSE certificate99%
- opinion_32d_abril.pdfSAT32-D opinion · SAT98%
- imss_positiva_q1.pdfIMSSIMSS opinion98%
- infonavit_abril2026.pdfINFONINFONAVIT opinion96%
- sisub_t1_2026.xmlSISUBSISUB · quarterly99%
- ICSOE_REPSE_padron.pdfICSOEICSOE · quarterly97%
- nomina_quinc_05.xmlCFDIPayroll CFDI (stamped)99%
- DC3_curso_alturas.pdfDC-3DC-3 STPS training94%
- scan_005.jpgDC-3Likely DC-3 (low quality)71%
- anexo_3.pdf—Not classifiable12%
37 more files · classified in 26.3s
The real friction is in manual upload.
A typical vendor must upload between 14 and 28 documents every quarter, each to a different portal, in a different format, with different naming. For a vendor base of 150 vendors × 4 quarters × 16 documents, that's 9,600 files a year someone has to receive, classify, validate and archive.
Most existing REPSE systems solve the "after" (audit, alerts, reports) but leave the "before" — the upload — to Excel spreadsheets, emails with incoherent subjects and uncontrolled shared folders. That's where the process breaks.
Five steps, no forms.
Upload
The vendor drags a ZIP (or a folder), or forwards an email with attachments to the dedicated alias for their contract. No forms, no imposed naming.
AI classification
Each file passes through a model fine-tuned for Mexican regulatory documentation. Identifies the type among 16 critical REPSE + 12 optional documents, with confidence score per document.
Field extraction
Vigía extracts folio, RFC, issue date, expiry, issuing authority, SAT seal. Compliance certificates are validated against the STPS registry in real time.
Cross-validation
Confirmation against SAT, IMSS, INFONAVIT APIs. Detection of duplicates, expired items, false positives. What doesn't pass is flagged for human review — never fabricated.
Signed file
Validated documents are compiled into the quarterly vendor file, with NOM-151 seal and immutable chain of custody. Audit-ready.
Less friction for the vendor, less work for your team.
AI to accelerate. Humans to sign.
AI classifies and extracts
Identifies document types, extracts folios, dates, RFCs and expiries. Accelerates repetitive manual work.
AI does not decide nor sign
Every final legal decision (accept a file, escalate an exception, respond to STPS) is taken by your team. Vigía proposes, humans decide.
Your data does not train models
Your vendor documentation is never used to train public models. Contractual clause + per-tenant isolation.
Visible confidence score
Every classification comes with a confidence score. Below 85%, the document is queued for human review before entering the file.
Models are fine-tuned on a corpus of Mexican regulatory documentation and updated when official formats change (last update: STPS REPSE registry 2026). We don't use generic public models for production classification or validation decisions.
Send us a test ZIP. No setup.
We provide a temporary alias where you can send 5–10 sample REPSE documents. In under a minute you receive a classification, extraction and validation report. No contract signed, nothing installed.