FactumOS
Transparent reasoning that you can see, verify, and trust.
The reasoning layer that makes enterprise AI transparent, auditable, and defensible.
✓ Transparent reasoning, not a black box
- →See every reasoning step, from evidence to conclusion
- →Verify sources and rule applications with complete visibility
- →Replay any decision to show exactly how it was reached
🔗 Every decision stays tied to its sources, reviews, and timing.
📄 Receipts bundle what happened into a simple, shareable record.
Transparent vs Black Box
See the difference.
Black-Box LLMs
- ✗Hidden reasoning—you can't see how decisions are made
- ✗No source visibility—can't verify where answers come from
- ✗Inconsistent outputs—same input can produce different results
- ✗No audit trail—can't explain decisions to regulators
FactumOS Transparent Reasoning
- ✓Visible reasoning—see every step from evidence to conclusion
- ✓Source transparency—verify every input and rule application
- ✓Deterministic outputs—same inputs always produce same results
- ✓Complete audit trail—replay any decision with full transparency
Why AI projects stall under scrutiny
Teams need transparent reasoning.
Market analysis
Where buyers are spending in 2025.
Trust beats novelty
Executives want answers they can stand behind, not another flashy demo.
Audit pressure is rising
Regulators and customers now expect a clear record for every high-stakes choice.
Lean teams need repeatability
Small teams win when decisions can be reused without starting from scratch.
Receipts preview
Clear, audit-ready proof.
Example Question
What are the availability requirements for our EU region service, and what happens during a failover scenario?
Record hash
7dd1a2d9:ca7e9b1f6d2e
Locked after review
Receipt root
0x7f82d1c87f3b8b6f72c957aa…
Proof of what changed
Replay mode
deterministic
Replay stays consistent
Evidence score
0.0%
Sources included
Explanation score
0.0%
Readable, constrained narrative
Conflicts
0 warning / 0 info
Logged with levels
Replay assurance
iDeterministic receipt preview
Replay hash unchanged; the same evidence set returned and outcomes stayed consistent.
Evidence score
0.0%
Grounded signals used
Explanation score
0.0%
Clear, guided narrative
Conflicts + replay status
0 warning / 0 informational conflicts captured in this demo receipt.
Replay uses record SNAP-7dd1a2d9 and hash 7dd1a2d9:ca7e9b1f6d2e; outputs remain stable across replays.
Extracted Texts from Documents
5 documentsService agreement — lines 118-142
Clause 4.2: EU Region Availability Requirements. The service shall maintain 99.9% uptime availability within the European Union region, excluding scheduled maintenance windows. During failover scenarios, the system must complete transition within 5 minutes with zero data loss. All failover events must be logged and auditable.
Runbook: failover — lines 44-63
Failover Procedure Step 1: Verify primary region health status. Step 2: Initiate controlled shutdown of primary services. Step 3: Activate secondary region services. Step 4: Verify data consistency checks. Step 5: Route traffic to secondary region. All steps must complete within 5-minute window per SLA requirements.
Audit policy — lines 142-188
Policy 142: All data processing within EU region must comply with GDPR Article 6 (lawful basis) and Article 44 (transfers). User consent must be explicitly obtained and recorded before any data processing. Audit logs must capture consent timestamp, method, and scope.
Knowledge base — lines 201-224
Historical Incident KB-24: On 2024-11-15, a failover event occurred in EU-West region. Root cause: Network partition detected. Resolution time: 4 minutes 32 seconds. Data consistency verified post-failover. No customer impact reported. This incident validates current failover procedures.
Policy config — lines 08-24
Conflict Resolution Policy CFG-08: When multiple sources conflict, priority order: 1) Latest verified policy document, 2) Explicit user consent records, 3) Historical incident data. All conflicts must be logged with severity level (warning/info) and resolution method.
Applied Policies
4 policiesEU-GDPR Data Residency
Ensures all data processing complies with GDPR requirements for EU region operations
SLA Compliance Verification
Validates that service availability commitments meet contractual obligations
Failover Audit Trail
Requires complete logging of all failover events with timestamps and verification
Conflict Resolution Rules
Defines priority order and logging requirements for conflicting information sources
Graph Reasoning Steps
8 stepsExtracted intent: availability requirements + failover procedures for EU region
Inputs: QueryNode
Found 5 relevant documents: DOC-12, RNBK-03, POL-142, KB-24, CFG-08
Inputs: DocumentRetrieval
Extracted 5 text segments covering availability SLA, failover procedures, GDPR compliance
Inputs: TextExtraction
Applied 4 policies: EU-GDPR residency, SLA verification, failover audit, conflict resolution
Inputs: PolicyApplication
Detected 4 conflicts: 2 warnings (unverified source attempt, outdated info), 2 info (record drift, human override)
Inputs: ConflictDetection
Constructed graph with 42 nodes, 118 edges. Connected: SLA requirements → Failover procedures → GDPR compliance → Historical validation
Inputs: GraphReasoning
EU region requires 99.9% uptime. Failover completes in <5min with zero data loss. All events logged per GDPR audit requirements.
Inputs: AnswerGeneration
Receipt locked to SNAP-7dd1a2d9 with hash 7dd1a2d9:ca7e9b1f6d2e. All evidence, policies, and reasoning steps bound to record.
Repeatable decision trail
Every step bound to the locked record.
Source intake
checkedStep owner: Structured parser
Duration 142 msRecord SNAP-7dd1a2d9Context link
guidedStep owner: Guided assistant step
Duration 388 msRecord SNAP-7dd1a2d9Policy check
allowStep owner: Rule set v4.2
Duration 91 msRecord SNAP-7dd1a2d9Receipt
sealedStep owner: Receipt writer
Duration 27 msRecord SNAP-7dd1a2d9
Differentiation
Built for scrutiny without the jargon.
Use cases
Where teams need the clearest trail.
Compliance verification
Receipts that make audits straightforward, even across regions or teams.
Outcome: Clean proofs that satisfy examiners the first time.
Policy interpretation
Consistent rulings with clear, reviewable overrides.
Outcome: Consistent, explainable rulings across teams.
Risk assessment
Receipts that explain every signal, score, and override.
Outcome: Defensible risk calls you can show to risk and legal.
High-stakes internal knowledge decisions
Versioned answers with replayable history for edits and escalations.
Outcome: Versioned, accountable answers leadership can sign off on.
Final check