CASE INTELLIGENCE
AI Case Intelligence for Personal Injury Law Firms
Score every incoming lead from 0 to 180 across five tiers — tuned to your firm's carriers, counties, and venues. The AI reads the intake call, extracts structured facts, benchmarks against comparable verdicts, and ranks qualified cases first.
Five tiers, one signal
Clear liability, catastrophic or high-economic damages, capacity-ready defendant. Route to senior attorney immediately.
Strong case on multiple dimensions. Warrants full attorney review and standard case-development workflow.
Mixed signals. Some strengths, some concerns. Needs targeted follow-up to clarify liability, damages, or defendant capacity.
Weak on multiple dimensions. Review briefly before deciding; most cases here are candidates for referral out.
Does not meet minimum thresholds. Decline with a clean, respectful close — or refer to a firm better positioned for the case.
Three-layer scoring model
Every score is the combination of three independent layers. That's what lets Valryn encode your firm's specific knowledge about carriers, counties, and venues without retraining the entire model.
Global layer
Core factors that apply to every personal injury case: liability clarity, injury severity and documentation, damages provability, defendant capacity, and SOL proximity.
Case-type layer
Factors specific to the case type — standard of care for med-mal, notice for premises, serious-injury threshold for NY auto, paid-or-incurred for TX auto. The scoring engine automatically loads the right playbook.
Modifier layer
Firm-specific modifiers: carrier behavior profiles, county jury tendencies, venue-specific damages patterns. This is where your firm's institutional knowledge gets encoded into the AI.
How a case gets scored
Intake call arrives
AI receptionist handles the call (or your team does — intake source doesn't matter). Transcript, sentiment, and structured facts are extracted in real time.
Case facts structured
Incident date, location (city → county resolved), case type, injury details, medical treatment, insurance, and liability signals are extracted into structured fields.
Case scored across all three layers
Global, case-type, and modifier layers run in parallel. Each contributes to the final 0-180 score, with explainability for every factor.
Comparable verdicts surfaced
Hybrid search (pgvector semantic + keyword) pulls the most comparable verdicts from the same case type, state, and county. Damage estimates are grounded in these comparables.
Tier assigned + ranked
Case is assigned to one of five tiers (Prime / Qualified / Follow-up / Review / Rejected) and ranked in your pipeline. High-tier cases surface to senior attorneys first.
Under the hood
Scoring runs on a three-layer architecture backed by PostgreSQL and OpenAI embeddings. Comparable verdicts are indexed in pgvector with HNSW for fast semantic search. CourtListener integration provides keyword search in parallel. County-level resolution happens via a city+state → county lookup at intake so cases are compared to the venue where they'd actually be tried.
See case intelligence score your live leads
Book a demo and we'll run sample leads from your practice area through the live scoring engine — showing every factor contributing to the final tier.
Request Demo