praesensus
Product

The unified operating picture for legal work.

One state per matter. One recommendation per decision. Every driver named. Every approval tier explicit. Every advisory replayable.

What Praesensus does

Praesensus ingests from every legal tool your team already uses, reasons across sources through a genuinely new Combining Layer, and produces:

  1. A live operational state for every matter — calibrated, debounced, explainable.
  2. A recommended next action with the named drivers behind it.
  3. An explicit authorization to proceed — approval tier + reason code + HALT path.
  4. A cross-source unified view — the conclusion no single tool could have reached.

The seven-stage spine

Every advisory travels the same seven-stage pipeline, lifted intact from the sibling Atlas family and battle-tested across four production verticals.

Stage 1

OBSERVE

Ingest from every source your firm already uses.

Stage 2

NORMALIZE

Extract clauses, obligations, entities, dates.

Stage 3

DETECT_STATE

Interpret the matter's operational state.

Stage 4

DETECT_TRANSITION

Debounce state changes. Catch drift early.

Stage 5

ESTIMATE_OUTCOME

Forecast deadlines, exposure, cycle time.

Stage 6

RECOMMEND

Synthesize fused inputs into one action.

Stage 7

VALIDATE

Deterministic replay + precision/recall metrics.

The Combining Layer

The genuinely new IP. Ten sub-modules operating over a typed knowledge graph.

The Combining Layer is what produces the conclusions no single point tool can reach. The typed knowledge graph spans Person · Organization · Matter · Contract · Clause · Obligation · Email · Document · Court · Judge · Statute · Regulation · Precedent · Filing · Deadline · Billing · Evidence · Approval.

Identity resolution

Person X in email = Person X in contract = Person X in matter file.

Knowledge graph

18 node types, 16 edge types, typed reasoning.

State fusion

Merge per-plugin state estimates into one matter state.

Evidence fusion

Coverage against a target evidence schema per matter type.

Timeline fusion

Chronological merge with source lineage.

Risk fusion

Merge per-plugin risk signals into one risk band.

Semantic linking

Clause ↔ regulation ↔ deadline ↔ policy.

Cross-plugin correlation

Co-occurrence + causal chains across sources.

Contradiction detection

Contradictions across sources, escalated to human review.

Unified recommender

One recommendation per decision, with named drivers.

Three posture invariants — present at every layer

Insight-only. Praesensus advises. It never actuates on legal matters. Every action is human-approved. Permanent, not a phase.

Fail-closed governance. Default-deny. Every advisory carries a tier requirement, a reason code, and a HALT path if governance conditions aren't met.

Federate patterns, not privileged content. The federation moat is built from anonymized derived patterns. Enforced architecturally, not by policy.

What Praesensus does NOT do

Posture, not roadmap:

  • Draft documents. Praesensus reads what drafting AI produces (Harvey, Spellbook, CoCounsel) and reconciles it with everything else.
  • Replace research platforms. Westlaw, Lexis, Bloomberg Law are consumed as plugins.
  • Replace point CLM. Ironclad, Icertis, ContractPodAi handle authoring; Praesensus reads their output.
  • Actuate autonomously. Every action requires human approval. The regulatory moat.
  • Practice law. Decision support to licensed professionals; not itself a licensed professional.

See it in action

Request a demo Calculate your ROI