How Petria works

From exhibits to filing-ready output

Petria is built around matters and jobs, not open-ended chat. The tower on the right walks through an LMIA example, but the same verified pipeline (tokenize, translate, extract, draft, verify, redraft, deliver) applies to refugee claims and other narrative-heavy Canadian work. Your team always knows what stage a case is in and which files are ready to review.

LMIA and refugee: one engine

Labour market impact assessments and refugee claims (BOC) both depend on long-form narrative grounded in exhibits, not checkbox forms. Petria treats them as the same job architecture: labeled evidence in any language, translation to the matter working language, template-grounded drafting, citation verification, and lawyer-facing Word deliverables.

The core pipeline

1

Open a matter and upload exhibits

Create an LMIA or refugee matter. Upload supporting documents in any language and assign exhibit labels. Set the matter working language (English or French). PII is tokenized before translation and drafting.

2

Submit a drafting job

Petria runs extraction, LMIA labour-market research when applicable, template-grounded generation, citation verification, and redraft as one coordinated job. Progress updates stream to your team in real time.

3

Review deliverables

Download the application draft and verification artifacts as .docx files in your working language. Your attorneys review grounded output instead of assembling narrative from scratch.

Bilingual delivery (EN / FR)

Canadian firms routinely work across English and French. Each matter has a working language; exhibits can arrive in either official language or a third language. Petria normalizes evidence before drafting and can produce final deliverables in the firm's chosen working language.

  • Matter-level working language (English or French)
  • Per-exhibit translation policy: translate for drafting or preserve certified originals
  • Translation runs after PII tokenization so identifiers never reach the translation API
  • Drafting prompts and .docx output follow the matter working language

Express Entry and PNP (v2)

Permanent-residence filings (Express Entry, provincial nominee programs) are document-heavy and checklist-driven, not narration-heavy like LMIA or refugee work. They are parked for v2 on the same filing_package track as US family forms (I-130, I-485). Today's production templates cover LMIA, SUV, and refugee claims. Contact us if you need EE/PNP scoped for a future release.

PII tokenization at intake

Petria replaces names, dates, and other identifiers with placeholders before any exhibit or intake field reaches the drafting pipeline. Inference runs on tokenized text only. Final .docx deliverables are restored for attorney review. This is implemented in the product architecture, not only described in a privacy policy.

Most immigration AI tools rely on SOC 2, encryption, and no-retention contracts. Petria adds a pre-model tokenization step so client PII does not enter generation prompts.

Outcome-verified learning

Onboarding teaches your firm's voice. Outcome learning is what happens after you file: when an application is approved, receives an RFR, or is refused, that result is tied to the matter and adjusts emphasis on the next application. The signal stays inside your firm. Petria does not train a shared model across customers.

  • Link an IRCC application number after filing, or confirm the outcome manually
  • Approved, RFR, and refused results tune emphasis on future LMIA and refugee matters
  • Compounds with voice from onboarding so associates inherit partner judgment, not a generic template

Your firm's structure, not a default skeleton

Standard and Pro draft against researched templates. Enterprise is for firms whose winning work follows a specific outline: section order, headings, where exhibits land, and which arguments get priority in each rubric. Petria learns that structure from your samples, your team approves it once, and every production job drafts the way your partners already expect.

  • Associates open a first draft that already matches your firm's table of contents and emphasis patterns
  • Section notes and exhibit placement reflect how your team actually wins cases, not a generic immigration outline
  • Approved structure is versioned through support so production jobs stay stable when you refine the playbook
  • Bring-your-own S3 (BYOB) so exhibits and deliverables stay in your bucket

See the pipeline on a real application matter

In a guided demo we open a sample LMIA or refugee (BOC) matter, walk through bilingual exhibit handling, run one full draft job, and download the verification report and Word deliverables in your matter working language. The tower on this page shows LMIA; the same job path applies to refugee narrative work.

  • LMIA or refugee matter setup with EN/FR working language
  • Translation and exhibit policy on real intake shapes
  • Same verified job path and .docx deliverables for both matter types
Request a demo