The 9-step demand-package build.

The demand-package surface is the most consequential output JustineAI produces — it is what the firm sends to opposing counsel. The 9-step demand-package build assembles it through discrete, attorney-attestable steps. Each step is sub-agent-led; each step is attorney-reviewed; the final package leaves the platform only when the attorney signs off on the assembled whole. Two different nine-item pipelines used to share one name on this site, which was our error and is corrected here: the CURATION LAYER is nine modules that clean the record on the way IN (OCR Janitor, Provider Canonicalizer, code validator, and so on), before F5 reasons over anything. This is the nine steps that build the demand on the way OUT. They share a count and nothing else.

What the platform actually does.

1. 01

   9 build steps: liability narrative → mechanism of injury → medical chronology → ICD-10/CPT exhibit → economic damages → non-economic damages → future-care projection → settlement demand → final cover.

2. 02

   Each step has its own sub-agent, its own Eve-Genesis-trained reasoning posture, and its own attorney-attestation interface.

3. 03

   Inter-step consistency checks run automatically: the medical chronology must reference all and only the records in the exhibit; the damages calculations must trace to the records cited; the settlement demand must be consistent with the damages.

4. 04

   Citation provenance is maintained throughout: every claim in the demand traces to an underlying medical record, expert opinion, or computed value.

5. 05

   The package leaves the platform only when the attorney signs off on the assembled whole. The platform does not transmit.