Warnvo is in early access. Some features are in beta. See status →
Technical explainer · Watch both layers work

Two layers of analysis. Every violation found.

Other tools are one LLM. ShieldEngine™ is a deterministic rule engine and contextual reasoning model working on the same document — watch them both below.

warnvo.com / scan / dual-layer-demo
BOTH LAYERS ACTIVE
DOCUMENT
Ace Collections LLCNotice #38291FINAL NOTICE
We will pursue legal action if payment is not received within 10 days.
A judgment against you could result in wage garnishment.
This account is past due.
Payment may be made by certified check or online at acecolpay.com.
LAYER 1

Rule Engine

Deterministic pattern matching. Zero hallucination.

ACTIVE
§807(5) MATCHED

Legal action threat pattern

Deterministic · 100%

§807(4) MATCHED

Wage garnishment threat

Deterministic · 100%

CLEAN PARAGRAPH

No pattern match

Patterns checked

247

Matches found

0

LAYER 2

Contextual Reasoning

LLM trained on consumer protection law. Handles gray areas.

ACTIVE

Reasoning passes

4

Concerns flagged

0

Both layers run on every scan · 10.2s per document

— The architecture —

Neither layer is enough alone.

Rule engines miss context. LLMs hallucinate. ShieldEngine™ runs both — so every finding is grounded in either a matched pattern or reasoned analysis, and critical violations are caught by whichever layer spots them first.

LAYER 1

Rule Engine

Deterministic code that checks every document against known legal violation patterns. If it matches, it's a violation — no interpretation needed. Runs in milliseconds. Never hallucinates.

Catches what's black and white

  • FDCPA §807 threats (legal action, wage garnishment, arrest)
  • FDCPA §809 missing validation
  • No Surprises Act benchmarked rates
  • CMS bundling rules (unbundling detection)
  • Statute of limitations math by state
  • Document format validity checks
LAYER 2

Contextual Reasoning

An LLM trained on consumer protection law that handles the judgment calls rules can't make. Reads intent. Weighs context. Knows when a technically-legal letter is still deceptive.

Catches what's shades of gray

  • Intent and tone analysis in threats
  • Omissions and material misrepresentations
  • Statute of limitations context (when does it apply?)
  • Cross-document pattern recognition
  • Gray-area overcharge assessment
  • Translation of non-English contracts

— Run your document through both —

Two layers. One result. Every violation caught.

Your first scan runs both layers free. You'll see exactly what each layer flagged, with citations from the Rule Engine and reasoning from the LLM.