Factum AI™ | Artificial Intelligence Layer

Factum AI is the proprietary intelligence layer of Factum Software, embedded across the platform to convert clinical, financial, payer, and contract data into structured insights, litigation merit scores, and automated workflow actions - all within your sovereign infrastructure.

See Factum AI LiveTry the Interactive Tool

How Factum AI Processes Every Claim

Six sovereign AI modules work in sequence from intake through litigation - each purpose-built for a specific phase of the revenue cycle.

Medical Necessity Detection

NLP scans clinical notes, ABG values, and treatment documentation to extract structured medical necessity indicators aligned with payer LCD/NCD policies.

ICD-10 ExtractionLab Value ParsingSeverity Scoring

Litigation Merit Scoring

A weighted multi-factor model computes a 0–100 merit score based on clinical evidence strength, payer friction history, and contract variance analysis.

Weighted ML ModelPayer Pattern DBContract NLP

Automated Document Review

Processes hundreds of chart pages in seconds - highlighting key passages, flagging contradictions, and structuring evidence into court-ready format.

OCR + NLP PipelineEvidence BundlingContradiction Flags

Sovereign AI Inference

All clinical data stays inside your private infrastructure. No PHI transmitted to external AI APIs. Models trained exclusively on your case history.

Private LLMZero Data EgressCustom Training

Payer Friction Intelligence

Tracks denial patterns by payer, facility, and diagnosis code. Surfaces statistical evidence of systematic underpayment for litigation leverage.

Denial Pattern DBStatistical AnalysisTrend Detection

Workflow Automation Triggers

When a claim exceeds the merit threshold, Factum AI automatically routes it to the correct workflow phase - no manual triage required.

Rule EnginePhase RoutingSLA Enforcement
Interactive Tool

Litigation Intelligence Control Room

Adjust the three scoring inputs to see the merit score update in real time, then run the AI document parser to watch clinical data get extracted and classified.

Scoring Inputs
Clinical Evidence Strength(ICD + labs + notes)
42
LowHigh
Historical Payer Friction(denial pattern score)
38
LowHigh
Contract Variance Value(underpayment delta)
29
LowHigh

Score Breakdown

Clinical (45%)
19
Payer (35%)
13
Contract (20%)
6
Merit ScoreHOLD
025507510038MERIT SCORE

Automated Action

Hold - Insufficient Evidence for Litigation

Clinical documentation does not meet minimum merit threshold. Case should be returned to billing for additional documentation or closed.

  1. 1Return to billing team
  2. 2Request complete medical record
  3. 3Document denial rationale
  4. 4Consider administrative appeal
Score 38/100 · Factum AI v4.2 · 4:32:38 PM
Unstructured Chart Feed
READY
ADMPatient admitted 03/14/2024 – acute respiratory failure, O2 sat 84% on room air.
DXPrimary: J96.01 Acute respiratory failure w/ hypoxia. Secondary: J18.9 Pneumonia, unspecified.
TXInitiated BiPAP therapy, IV Levaquin 750mg QD, methylprednisolone 125mg IV Q6H.
NOTEAttending physician note: Patient deteriorating, ICU admission warranted per severity scoring.
LABABG: pH 7.28 / PaCO2 58 / PaO2 52 / HCO3 26.4 - consistent with acute hypercapnic failure.
DENYPayer denial: UHC EOB dated 04/01/2024 - "Level of care not medically necessary, outpatient equivalent available."
POLUHC Medical Policy MP-025.19: ICU admission criteria requires PaO2/FiO2 < 200 or acute organ failure.
CNTHospital contract §4.2(b): Payer must provide written rationale within 30 days of denial; none received.
HISTCase history: Payer denied 3/3 similar ICU admissions for this facility (FY2023). All appealed successfully.
OUTExpected recovery value: $47,200. Contract rate multiplier: 2.3x per amendment dated 01/2023.
NLP Extraction Engine
Zero PHI egress
ICD-10 CodeJ96.01-
O2 Saturation84% (room air)-
ABG pHJ96.01-
Level of CareICU admission-
Denial CodeMP-025.19-
Contract Clause§4.2(b) - 30d-
Payer Friction3/3 denials-
Recovery Value$47,200-

Sovereign AI Architecture

Factum AI is engineered to run entirely inside your organization's security perimeter - no PHI ever leaves your infrastructure, every decision is traceable to its source data, and models improve continuously from your own case outcomes.

01

Private Inference Engine

All NLP and scoring models run on infrastructure you control. No clinical, financial, or contract data is ever transmitted to external AI APIs or third-party model endpoints.

02

Zero Data Egress

Network-level egress controls guarantee that PHI cannot leave your perimeter. Every inference call is logged, audited, and tied to an authenticated user session.

03

Custom Model Training

Models are fine-tuned exclusively on your closed case history - learning your payer mix, contract language, and clinical documentation patterns for progressively sharper accuracy.

04

Explainable Decisions

Every merit score, denial classification, and workflow action includes a full audit trail tracing back to the exact clinical passages, policy clauses, and contract terms.

05

Model Governance

Versioned models with staged rollout and instant rollback. Administrators control which model version is live, when it deploys, and which case types it applies to.

Continuous Learning Loop

Closed case outcomes - won, lost, settled - feed back into the model automatically, so every new case benefits from the accumulated experience.