Platform
How the AI layeractually works.
A page for your CTO, IT lead, and compliance reviewer. The three-layer model that runs your engagement, the data flywheel that compounds, the integrations that close the loop, and the security architecture that lets you write the BAA.
The three-layer model
AI · Human · Data. In that order.
AI Layer
80%
Voice intake · document intelligence · case workspace · argument simulation · client portal · performance intelligence.
Human Layer
20%
Nofal implementation specialists + your attorneys. Every AI output below confidence threshold escalates.
Data Layer
∞
Every case generates outcome data that retrains the platform across the network — the compounding moat.
The data flywheel
Every case retrains the network.
Captured intakes feed the scoring model. Settled cases feed the damages model. Demand outcomes feed the negotiation simulator. The result is an asset that compounds: each firm's outcomes improve every other firm's recommendations.
Languages
English. Spanish. Portuguese. At launch — more on the way.
Real-time language detection on chat. Voice intake adapts to the caller's language mid-call. Additional languages added based on observed intake demand, not vendor roadmaps.
Integrations
Lands where the case lives. Lifts what the firm already runs.
CMS
- Litify
- SmartAdvocate
Documents
- Supio
Voice
- Twilio
Compliance / Observability
- Vanta
- Datadog
Inbound channels
- Toll-free
- Embedded chat
- Web forms
- SMS
- Google / Meta campaigns
Security architecture
Built for the BAA. Reviewed by your IT.
Encryption
AES-256 at rest. TLS 1.3 in transit. Per-tenant key wrapping.
PHI isolation
Healthcare data lives in a partitioned tenant with audit-logged access.
Audit logging
Every read, write, and AI inference is logged and retainable.
Regional residency
US-only by default; EU residency available on request.
AI quality controls
Confidence, calibration, and an escape hatch.
Confidence thresholds
Every AI output carries a confidence score; below threshold, a human reviews before the output ships.
Human-in-loop
Implementation specialists oversee initial cases; thresholds tighten as the model calibrates.
Continuous calibration
Outcome data from settled cases retrains the merit and damages models on a monthly cadence.
Compliance