THE PLATFORM
Every layer built for legal validity. Every output independently verified.
Regumint is structured around five principles: authority, jurisdiction, temporality, retrieval intelligence, and traceability. Here’s how each layer works.
HOW IT WORKS
Five layers. One defensible answer.
Upload your documents once. Regumint ingests PDF, DOCX, Markdown, and URLs, then enriches every file with deterministic legal metadata — jurisdiction, authority level, document type, date, and language. Quality scoring flags low-confidence documents. Deduplication prevents redundant sources. Structure-aware parsing handles articles, chapters, and sections. Definitions are extracted automatically. Multi-language support: English, French, and Italian.
RegumintEvery query triggers hybrid keyword and semantic search across your Firm Bank. Cross-language retrieval lets you ask in French and find English sources — or any combination. Results are jurisdiction-matched and authority-weighted, boosted across 10 metadata factors including recency, relevance, and document quality. Diversity selection ensures breadth across source types. Context expansion pulls surrounding passages so answers are never based on fragments in isolation.
RegumintEvery source is classified into a 6-tier authority hierarchy: Constitutional & Statutory law, Binding Case Law, Regulatory Guidance, Administrative Decisions, Interpretative Materials, and Commentary & Analysis. Authority weighting applies in both retrieval ranking and answer generation. Reliability scoring assesses source quality. Regime normalisation ensures consistent treatment across different legal systems.
RegumintPhase 1 produces a fully cited answer from your ranked sources with authority distribution, jurisdiction relevance, and temporal scope. Phase 2 independently audits that answer across 8 structured reasoning dimensions: temporal intelligence, legal ambiguity, interpretive method, authoritative silence, adversarial framing, source reliability, element coverage, and jurisdictional drift. Separate confidence scores for sources and answers. Query recommendations when confidence is low.
RegumintThe traceability layer enforces direct citation for every claim, visibility of source strength and authority tier, explicit identification of interpretive versus direct reasoning, and flagging of any unsupported claims. Follow-up query suggestions help close detected gaps. The system surfaces what it doesn’t know — not just what it does.
RegumintREASONING INTELLIGENCE
Eight dimensions. Every answer audited.
Phase 2 independently evaluates every answer across eight structured reasoning dimensions.
Temporal Intelligence
Detects superseded sources and mixed regulatory regimes across time periods.
Legal Ambiguity
Identifies open-textured language and discretionary standards that limit certainty.
Interpretive Method
Determines the applicable interpretation approach for each legal provision.
Authoritative Silence
Detects gaps where primary source authority is absent or insufficient.
Adversarial Framing
Classifies whether the law clearly applies, clearly does not, or remains genuinely unclear.
Source Reliability
Ranks each source by authority tier and jurisdictional alignment with the query.
Element Coverage
Measures how completely the retrieved sources cover each required legal element.
Jurisdictional Drift
Flags when retrieved sources come from a different jurisdiction than the query requires.
WITHOUT REGUMINT
Plausible answer. No accountability.
Generic AI search bar. No citations. No authority ranking. No gap detection. You don’t know what you don’t know.
- Black-box output
- No source hierarchy
- Hidden uncertainty
WITH REGUMINT
Defensible answer. Fully cited.
Authority-ranked. Gaps surfaced explicitly. Every claim traceable to its source fragment.
- Claim-to-source traceability
- Authority tier ranking
- Explicit uncertainty flags
- Cross-language retrieval
- 8-dimension reasoning audit
- Follow-up query guidance
WHAT’S INSIDE
Every component built for legal validity.
Firm Bank
Documents ingested, quality-scored, deduplicated, and enriched with legal metadata at ingestion.
Intelligent Retrieval
Hybrid keyword and semantic search with cross-language support and 10 metadata boost factors.
Authority Model
6-tier binding hierarchy with authority weighting in both retrieval and answer generation.
Query Decomposition
Jurisdiction, temporality, and authority scope identified before retrieval.
Two-Phase Reasoning
Answer generated then independently audited across 8 structured reasoning dimensions.
Traceability Layer
Every claim linked to its exact source fragment. Gaps flagged. Follow-up queries suggested.
Multi-Language Support
Ingest and search across English, French, and Italian documents. Ask in one language, find sources in another.
Learning Engine
Retrieval improves with your usage patterns and feedback.
DATA SOVEREIGNTY
Your data stays where you need it.
Cloud-hosted
Default. Fully managed, secure, isolated environments. Data encrypted in transit and at rest with role-based access controls.
Hybrid
On-premise storage components with cloud reasoning layer. For firms with data residency requirements or strict compliance mandates.
Specific configurations discussed during onboarding.
Ready to see it reason?
Book a walkthrough or start a 14-day free trial. No credit card required.