73+ capabilities, in eleven groups.

This is the unabridged Workers’ Comp list — the rating argument, the statutory clocks, the indemnity math for three states, the drafting, the named agents that work the book overnight and still have to ask, the trust ledger, search and deposition testimony cited to page and line, and the portal the injured worker signs in to. Every item below is reachable by an attorney today: a route you can click, or something you can ask Justine for. Where a capability has a real limit — a scope it does not cover, a file type it will not take, an end of it that has no screen yet — the limit is written into the entry rather than left for you to find. Code we have written and tested but that you cannot yet reach is named at the foot of the page rather than counted here. For the guided tour, start on the WC overview; for what shipped and when, see the WC release notes.

Decision support, bound by the ABA Rules. Justine reasons and drafts; the attorney reviews, attests and signs every output. Rating and settlement figures are projections the attorney verifies against the current schedule. Nothing here is legal advice, and nothing here decides on its own.

Win more comp cases — and stop leaving value behind.

Every other state, DC, and the federal programs (Longshore, FECA, and the Defense Base Act) are recognized by the jurisdiction resolver today — it identifies the governing forum and can ground and cite the controlling rules. Full benefit computation is built state by state, next.

1. 01Which forum governs

   More than one state can claim the same injury. Justine surfaces the candidate forums and how the benefits and deadlines differ — you choose. No silent guess about whose law applies.

2. 02The numbers, cited

   Benefits and indemnity computed to the governing statute, with the section cited. Every figure is marked either verified against the loaded schedule or flagged for confirmation — never a confident guess.

3. 03The deadline that ends the case

   State-specific docketing tracks the clocks that foreclose a claim — California’s §5405 SOL, Georgia’s one-year §34-9-82, Texas’s 90-day impairment-rating finality — before they pass.

4. 04Grounded, yours to verify

   Answers stand on statute- and case-cited rule bases, never invention. A rating is a rating aid, not a medical opinion; the analysis is work product; the attorney decides.

Not legal advice · not a law firm · a rating aid, not a medical opinion · the attorney decides.

• Benson apportionment

  It refuses the comparison the record cannot support.

  Benson v. WCAB requires a separate, separately-apportioned award for each industrial injury. Where the file does not carry what that comparison needs — a second recorded injury with its own date, or a finding that can be attributed to a claimed region — the comparison does not run and the screen names the missing evidence instead. It returns a reason, not an estimate.

• Deposition transcripts

  A transcript it cannot number is refused, not stored.

  Page-and-line numbering has to survive the export. A missing line number, a repeated page, or numbering that changes style mid-transcript is rejected with the page and line named — nothing is stored and nothing is inferred, because a guessed line number shifts every citation taken after it. That is why a page-and-line cite this product prints is quotable in a filing.

• The approvals inbox

  It keeps a record of what Justine declined to do.

  Declining is a titled section of the inbox, not a deletion: an attorney's refusal is stored with the written reason beside Justine's original draft, and a step the agent itself would not take is recorded as considered-and-declined. The inbox also distinguishes 'nothing is waiting for you' from 'we could not ask' — two states that most queues render identically, and that mean opposite things.

• Outbound send

  Approving is not sending.

  Outbound send is a per-firm capability that is off unless a firm deliberately turns it on, and it is on nowhere in the production deployment today. Approving a draft records the firm's decision and sends nothing — transmission stays a step a person takes. An agent cannot reach your client because a model decided it should.

• Reasoning streams

  An interrupted answer reports itself interrupted.

  If a reasoning run is cut short — a deploy, a dropped connection, a stop — the partial work is kept and labelled as partial. It does not render as a finished answer. This is the quietest failure mode in agentic software and the most dangerous one in legal work, because a truncated analysis looks exactly like a complete analysis that found less.

• Reading scanned records

  Where it cannot read, it says so.

  Optical character recognition runs on Azure Document Intelligence under managed identity. Where that route is not available to a tenant, the read returns an explicit not-configured result and the document is left unread. Nothing is inferred from a page that was never successfully read — which is the only version of this feature that is safe to put a medical record through.

1. Stage 01Intake & compensability

   Justine runs conversational intake against the workers’-comp schema for the matter’s state — date of injury, employer and carrier, body parts, compensability, prior claims, and the statute clock. She flags cumulative-trauma exposure windows and the post-termination defense before the Application goes out. Applicant-side; California, Georgia, or Texas.

2. Stage 02Application & filing

   Justine drafts the California WCAB documents that open and move the case — the Application for Adjudication of Claim and the Declaration of Readiness to Proceed — from the matter record, on the firm’s own letterhead. These are drafts, not the official state PDF forms: Justine fills no fillable form, and any field the record does not establish is left as a bracketed placeholder rather than invented. The DWC-1 claim form is not drafted at all — Justine recognises it on ingest and dockets the Labor Code §5402(b) compensability clock from its filing date. Every draft surfaces to the attorney for attestation before it is filed; nothing is auto-served. Client-facing letters draft in English or Spanish.

3. Stage 03Treatment & utilization review

   Justine tracks each authorized treatment request through the utilization-review process — flagging UR denials, the Independent Medical Review (IMR) appeal deadline, and out-of-MPN care. When a request is denied, she surfaces the reviewer’s stated basis against the treating physician’s report so the IMR appeal starts from the record itself.

4. Stage 04AME / PQME analysis

   Justine reads the AME or PQME report — “Justine’s read” — extracting whole-person impairment by body part, work restrictions, causation, apportionment opinions, and the P&S date. She flags where the evaluator departs from the record. Where the treating provider runs ChironAI™ Occupational Medicine, those assessments can feed the matter.

5. Stage 05MMI / P&S & impairment

   Justine watches every active matter for Maximum Medical Improvement / Permanent & Stationary signals across treating providers, then reads the impairment picture into the rating engine. She distinguishes the strict-AMA whole-person reading from a rebutted alternative before the attorney certifies anything.

6. Stage 06PD Value Engine

   In California, the Value Engine builds the rating string — WPI → the 1.4 future-earning-capacity factor → occupational variant → age modifier — and models permanent disability as a PD percent, a dollar value, and the swing between a strict-AMA rating and a rebutted one (Almaraz-Guzman / Ogilvie), with apportionment carved out under Escobedo. Georgia rates on the AMA Guides 5th and Texas on the AMA Guides 4th, each to its own benefit schedule. Figures are projections for the attorney to verify against the current schedule.

7. Stage 07Settlement

   Justine models Compromise & Release against Stipulations with Request for Award, applies the ~15% judge-approved contingency fee and reasonable lien values, and reports net-to-client under each path. The attorney sees the trade-off — a clean C&R buyout versus Stips that leave future medical open — before deciding the number and the posture.

8. Stage 08MSA flag & analysis

   Justine flags when a Workers’ Compensation Medicare Set-Aside is implicated before a C&R is papered — Medicare-beneficiary status, the CMS review thresholds, open future medical — and frames the analysis the attorney needs. She does not compute the set-aside dollar amount; that stays with a WCMSA vendor and the attorney’s judgment.

9. Stage 09Closeout & compliance

   On resolution, Justine delivers the final file, exports the tamper-evident audit log for ethics review, and closes the matter. The apportionment patterns and carrier behavior the firm learned inform its next case — inside the firm’s own tenant, never a shared model.

• Compensability sub-agent

  AOE/COE analysis. Cumulative-trauma date-of-injury under the last-injurious-exposure rule. Going-and-coming and personal-comfort exceptions. Post-termination and statute-of-limitations defenses. Employer/carrier identification. Concurrent-employment average weekly wage. Prior-claim apportionment exposure. Applicant-side.

  For example — A warehouse picker files for a cumulative lumbar injury after termination. You drop the personnel file, the panel QME report, and three years of pay records. The compensability sub-agent fixes the CT date-of-injury, flags the post-termination defense the carrier will assert, surfaces a prior 2019 lumbar claim that opens an apportionment argument, and computes the average weekly wage across two concurrent employers — each value traced to its source page for you to weigh.

• Medical sub-agent

  Treating-provider chronology. Utilization-review and IMR appeal tracking. AME/PQME impairment and apportionment review — “Justine’s read” of the report. MMI / P&S signal detection. Causation linkage. ChironAI™ OM upstream assessment ingestion where the provider runs it.

  For example — A 4,000-page file across an orthopedist, a pain-management group, and a panel QME. The medical sub-agent surfaces the UR denial of a lumbar fusion, the IMR appeal window closing in eleven days, and the contradiction between the PQME’s 8% whole-person impairment and the treating surgeon’s range-of-motion measurements supporting a higher rating — with the apportionment opinion isolated for the attorney’s review.

• Rating sub-agent

  The PD Value Engine — rating California (2005 PDRS), Georgia (AMA Guides 5th), and Texas (AMA Guides 4th). California rating-string construction — WPI, the 1.4 FEC factor, occupational variant, age modifier. Strict-AMA versus rebutted (Almaraz-Guzman / Ogilvie). Apportionment carve-out under Escobedo. PD percent, dollar value, and the PD/$ swing. Life-pension threshold flags.

  For example — A shoulder injury with a 12% whole-person impairment from the AME. The rating sub-agent builds the California rating string, models the permanent-disability award as a PD percent and dollar value, then carves out the share apportioned to a pre-existing rotator-cuff tear — showing the swing between the strict-AMA reading and an Ogilvie rebuttal, and flagging where the adjusted rating crosses the life-pension threshold. Projections the attorney verifies against the current schedule.

• Resolution sub-agent

  C&R versus Stipulations modeling. The ~15% judge-approved contingency fee and reasonable lien values. Net-to-client under each path. Open-future-medical trade-off. WCMSA-required flag against CMS thresholds — flag and analysis, not a dollar computation.

  For example — An accepted back-and-knee claim with disputed future medical. The resolution sub-agent models a Compromise & Release against Stips-with-open-medical, applies the ~15% fee and known lien values, and reports net-to-client under each. It flags that a WCMSA is implicated — the client is Medicare-eligible and future medical is open — and tells you to resolve the set-aside before papering the C&R, without inventing the set-aside figure.

• Client texting with a privacy relay

  Opt-in, STOP-compliant SMS keeps the client updated on their case, and attorney alerts route through a number-privacy relay so the client never sees a personal cell number. Transmitted over Azure Communication Services on a toll-free line.

• Per-case client email

  Real client messages threaded to the matter, sent through Microsoft 365 / Outlook — the correspondence lives with the file, not in a separate inbox.

• Bilingual EN / ES client letters

  Client-facing letters draft in English or Spanish. Filings stay in English; only the client-facing correspondence switches language — so the applicant reads their own case in the language they speak.

• English & Spanish voice

  Voice dictation and read-aloud honor either language — flip to Spanish once and both dictate and listen follow (es-MX / en-US). Push-to-talk in, natural speech out.

Not legal advice · not a law firm · ABA-conduct aligned · the attorney decides. Client communications are attorney-reviewed work product.

• Compensability analysis

  Justine surfaces — On a cumulative-trauma matter, Justine fixes the date of injury under the last-injurious-exposure rule, surfaces the going-and-coming defense the carrier is poised to raise, and flags a prior accepted claim to the same body part that opens an apportionment argument — each tied to the record page that grounds it, for the attorney to weigh.

• Utilization-review appeal

  Justine surfaces — Justine catches that a UR denial of a lumbar fusion rested on a misread of the requesting physician’s report, that the IMR appeal window closes in nine days, and that the report already contains the documentation the reviewer claimed was absent — the appeal half-drafted from the record itself.

• AME / PQME scrutiny

  Justine surfaces — Across a panel QME report, Justine surfaces that the evaluator apportioned 40% to non-industrial degeneration without substantial medical evidence explaining the “how and why” under Escobedo, that the impairment number understates the documented range-of-motion loss, and that a prior consistent treating opinion contradicts the apportionment — assembling the cross-examination the attorney can run.

• PD Value Engine

  Justine surfaces — On a shoulder injury rated at 12% whole-person, Justine builds the full California rating string — the 1.4 FEC factor, occupational variant, and age modifier applied — models the PD percent and dollar value, carves out the apportioned share, and shows the PD/$ swing between the strict-AMA rating and an Almaraz-Guzman or Ogilvie rebuttal, so the attorney sees what a rebuttal is worth before pursuing it.

• Settlement & net-to-client

  Justine surfaces — Justine models C&R against Stips, applies the ~15% judge-approved fee and the reasonable lien values, and reports net-to-client under each path — surfacing that leaving future medical open on Stips exposes the client to a WCMSA problem a clean C&R buyout avoids, so the number is chosen with its consequences visible.

• Citation discipline

  Justine surfaces — Justine grounds every statute and decision it cites against a verifiable authority — reasoned from real law, never invented — and self-labels its rating and settlement outputs as projections to verify against the current schedule and CMS thresholds. The attorney decides.

The AI reasons; the attorney decides.