Legal

Know what's proven and what you still need to collect

3 million files. 3 days. Zero human review. Your team walks in knowing what's substantiated, where the gaps are, and exactly what to collect next — while other tools are still returning search results.

Built for intelligence analysts, investigators, lawyers, journalists, and research teams.

You talk to ORCA. This is what it builds.

3M files 5 modalities ingested
22 sources Enrichment cross-referenced
9:41
48 findings From 496K facts
3 days Zero human involvement
Scroll to follow the reasoning

01

Monday. A voice note and a case number.

You leave a voice memo with the case number. That's the input — natural language, nothing else. ORCA downloads the full release, reasons through 3 million files across five modalities, and the entity graph on the right is what it produces. Not a progress bar. The transcripts, the extractions, the cross-references — all structured analytical output from a single conversation.

Audio transcript Maxwell proffer transcribed
Video extraction Interview frames extracted
94K observations 5 modalities, cross-referenced
3 days Autonomous processing

ORCA web app

ORCA web app
CO_CONSPIRATORFINANCIALASSOCIATERECRUITEDOWNSOWNSFOUNDEDTRANSFERRED_TOOWNSPROPERTY_OFFOUNDEDPASSENGER_ONPASSENGER_ONINVESTIGATED_BYINVESTIGATED_BYJ.E.G.M.L.W.J-L.B.Zorro TrustSouthern TrustThe 1953 TrustL.S.B. LeasingJ.E. & CoN-392EPSDNYPalm Beach PD
221 Entities
·
5,549 Edges
·
496K Facts
Graph Summary

22 enrichment sources cross-referenced. 3 trust entities flagged — Zorro Trust, Southern Trust, The 1953 Trust. L.S.B. Leasing linked to aircraft registration.

PERSON TRUST ORG AIRCRAFT

You ask

9:41

02

Thursday. What does the evidence actually prove.

You come back Thursday and ask about the case in plain language. ORCA reasons across the full corpus and surfaces findings, verdicts, and confidence scores. When you shift to a different angle, it re-reasons and completely different evidence surfaces — the COA subgraph updates to match. The conversation drives the analysis. The preview shows the result.

36 findings CORROBORATED at 82%
Score 0.90 Court filing extraction
INTERROGATORY #4 Routing gap — VALIDATED_GAP
Switch angles 18 findings, different evidence

ORCA web app

ORCA web app
COA: Financial Concealment via Trust StructureOWNS / FINANCIAL / TRANSFERRED_TO edges through Zorro, Southern, 1953 TrustCO_CONSPIRATORRECRUITEDASSOCIATEOWNSOWNSFOUNDEDTRANSFERRED_TOOWNSFINANCIALPASSENGER_ONJ.E.G.M.L.W.J-L.B.Zorro TrustSouthern TrustThe 1953 TrustL.S.B. LeasingN-392EPProof ObligationsCONTKNOWACTIINTE
4 Edges
·
5 Entities
·
2/4 Slots
COA Analysis

Trust structure links L.W. → The 1953 Trust → J.E. Zorro Trust and Southern Trust hold properties. Gap: intent behind transfers — documents alone can't prove it.

INTERROGATORY 2/4 filled KNOWLEDGE gap

You ask

9:41

03

The proffer. Not transcribed — watched.

The Maxwell proffer isn't just another file in the corpus. ORCA watches it the way an expert legal commentator would — tracking behavioral cues, qualifying language, evasion patterns, and contradictions against the existing entity graph in real time. When she says 'much more sophisticated than,' ORCA flags the qualifier and cross-references against SEC filings. The dashboard on the right is what your team sees: a live credibility analysis that no transcript alone could produce.

14 cues Behavioral analysis
Qualifier 'much more sophisticated than'
89% confidence Contradiction flagged
2 new edges Document analysis missed

ORCA web app

ORCA web app
OBSERVE STEP
Driver VLM LLM
LIVE Day 1, 11:07 AM
Seg:$0.016 VLM:$0.228 LLM:$0.529
Behavioral Analysis
CLASSIFY
testimony_credibility 89%
ANALYSIS
Qualifying language detected — "much more sophisticated than" avoids specificity
RECOMMEND
Cross-reference SEC filing F-2005-0847
Contradiction Tracker
3 contradictions flagged
Current: 11:07 conflicts with Day 2 sworn testimony (DEP-2019-0334)
VALIDATED against entity graph
11:02 GAZE_AVERSION
11:04 LIP_COMPRESSION
11:07 QUALIFIER
11:09 SWALLOW
11:11 HAND_WRING
11:06 …the arrangement was, well, it was [HESITATION]
11:07 "much more sophisticated than" typical portfolios… [QUALIFIER]
11:08 …I wouldn't say I was directly… [GAZE_AVERSION]
11:09 …involved in the, um, the structuring… [HESITATION]
Cost: Seg $0.016 VLM $0.228 LLM $0.529

You ask

9:41

04

Thursday afternoon. When it happened — reconstructed.

You ask when the money moved relative to the flights. ORCA reconstructs a timeline — 847 events across 17 years from 5 modalities. The visualization you see is the analytical output: temporal clusters where entities co-occur within days, financial flows correlated with movements. You asked a question. ORCA produced this.

847 events 5 modalities, 17 years
Sept cluster 3 events in 48 hours
Correlation 0.87 Flights ↔ Payments
Email spikes 14 instances before each flight

ORCA web app

ORCA web app
Reconstructed Timeline
847 dated events from 2002 to 2019 · 5 modalities
847 events LIVE
JUN 2005 1 event
2005-06-14 PDF A1
$50M: L.B. → J.E. via undisclosed trust
AUG 2005 1 event
2005-08-22 EMAIL B3
G.M. → J.E. re: "new arrivals"
SEP 2005 ▌cluster▐ 2 events
2005-09-03 PDF A1
N-392EP flight manifest: J.E., G.M., +2 unidentified
2005-09-05 AUDIO A2
A.D. deposition — describes recruitment pattern
NOV 2005 1 event
2005-11-18 PDF A1
$2.3M: Zorro Trust → Southern Trust wire, $2.3M
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019

You ask

9:41

05

Friday. Where do we stand.

You ask "where do we stand" and ORCA delivers a structured case assessment brief — each theory with its verdict, confidence score, and collection targets. A lawyer walks in on day one with exactly what they need. One question in. One complete analytical product out.

ORCA web app

ORCA web app
Case Assessment Brief 3M files · 496K facts · 221 entities · 5,549 edges · 3-day autonomous run
Assessment complete FINAL
18 U.S.C. § 1621 Perjury / Obstruction
CORROBORATED ACTIONABLE
Sworn statement PROVEN
Material falsity PROVEN
Willfulness PROVEN
Contradicting evidence PROVEN
18 U.S.C. § 1956 Money Laundering
CORROBORATED VIABLE_WITH_GAPS
Financial transaction PROVEN
Unlawful proceeds CIRC.
Knowledge of origin STRUCT.ABS.
Intent to conceal CIRC.
Gap: Knowledge of unlawful origin — KNOWLEDGE_STATE predicate 6 COAs
18 U.S.C. § 1591 Trafficking
NOT_SUBSTANTIATED VIABLE_WITH_GAPS
Recruitment CIRC.
Coercion PROVEN
Transportation PROVEN
Victim ID ABSENT
Knowledge STRUCT.ABS.
Gap: Victim identification — witness testimony required 4 COAs
1 theory ACTIONABLE. 2 VIABLE_WITH_GAPS — targeted collection from 16 validated discovery requests.
CORROBORATED Perjury — ACTIONABLE
NOT_SUBSTANTIATED Victim ID — UNFILLED
VIABLE_WITH_GAPS Intent is STRUCTURALLY_ABSENT
206 COAs Every gap mapped to collection

You ask

9:41

Methodology

How it thinks — Evidentiary Threshold Classification

ORCA classifies every piece of evidence against four evidentiary thresholds — from probable cause through beyond reasonable doubt. This determines not just what the evidence says, but what it can actually support in a legal proceeding.

ThresholdStandardUse Case
Probable CauseReasonable basis to believeCharging, indictment
PreponderanceMore likely than notCivil proceedings
Clear & ConvincingHighly probableFraud, punitive damages
Beyond Reasonable DoubtNo reasonable doubtCriminal conviction

No other legal tool classifies evidence against specific evidentiary thresholds. This is why ORCA can tell you where your case actually stands — and what to go after next.

When evidence overwhelms judgment

You just inherited a case. 3 million files across PDFs, emails, audio, video, and images. Your team needs to know what the evidence actually proves — not what it mentions. Other tools give you a search bar and a network diagram. You need a complete evidentiary picture: which theories hold up, where the gaps are, and what specific collection closes them.

Not what the documents say — what they prove

ORCA builds a typed entity-relationship graph, maps evidence to each theory of the case, and identifies exactly which discovery requests close which gaps.

Your team reviews, overrides, and annotates at every stage. The system recommends — your team decides.

Workflow Steps

01

Build typed knowledge graphs — 221 entities across 9 types, 5,549 edges across 35 relationship labels

02

Per-COA subgraphs — each Course of Action filters to the entities and edge types that fill its specific proof gap

03

Detect structural unprovability — flag KNOWLEDGE_STATE predicates that documents alone cannot satisfy

04

Generate ranked collection priorities with specific legal mechanisms (MLAT, subpoena, deposition)

9:41

Outputs

Entity-Relationship Graph

221 entities · 9 types (PERSON, ORG, SHELL_COMPANY, AIRCRAFT, ACCOUNT) · 5,549 edges · 35 relationship labels · 22 enrichment sources cross-referenced

COA-Specific Subgraph

COA-007: Trafficking network · RECRUITED/VICTIM_OF/COMMUNICATED edges highlighted · 3/5 proof slots filled · Gap: INTENT requires testimony

Case Assessment Brief

Perjury: CORROBORATED — ACTIONABLE · Financial concealment: CORROBORATED — VIABLE_WITH_GAPS · Trafficking: NOT_SUBSTANTIATED — VIABLE_WITH_GAPS · Per-theory proof obligations with collection targets

Theories assessed, gaps mapped, collection ready

Your team walks in knowing what the evidence supports, what needs more work, and exactly what to collect next.

Frequently asked questions

All three — but none of them the way existing tools handle it. eDiscovery tools search and organize documents. ORCA ingests everything — documents, video, audio, images — generates hypotheses, and maps evidence to specific legal theories. For indictability research, it triages which leads are actually pursuable. For journalism, it builds the evidentiary chain that survives editorial review. The workflow is different for each, but the engine is the same: ingest, reason, identify gaps. Other tools help you find documents. ORCA tells you which leads are worth pursuing.

Every claim ORCA makes traces back to source evidence with full provenance. The system runs adversarial analysis — a Blue Team builds the case, a Red Team attacks it. Circular evidence gets flagged and zeroed. Claims that require testimony rather than documents get marked structurally unprovable — ORCA won't pretend it proved something it can't. ORCA doesn't just find supporting evidence. It tells you when the evidence isn't enough.

No — and that's the point. ORCA is not an AI lawyer. It's a highly sophisticated legal researcher that saves your team enormous time on triage. It identifies qualified, indictable, pursuable leads and maps every gap in the evidentiary chain. Your lawyers make every strategic decision. ORCA makes sure they're not starting from scratch. ORCA doesn't practice law. It does the research that lets your team practice faster.

Dump everything into a Google Drive and send the link. ORCA handles OSINT enrichment from 22 external sources as usual — pulling whatever's publicly available. If you need specific sources scraped, give explicit instructions and ORCA handles the collection too. Custom setup or self-initiated, both work. Send a link. ORCA starts reading.

No. You don't need to log in to get results. ORCA produces structured outputs — entity graphs, case assessment briefs, gap analyses, evidence chains — and delivers them. The web app exists for visualization and deep exploration, but it's not required. The research shows up. You decide how deep you want to go.

Tenant-isolated infrastructure. No cross-client data sharing. Your evidence corpus stays within your perimeter. Deployment options from cloud to on-prem depending on sensitivity. Your evidence never leaves your perimeter.

Legal AI tools summarize documents and answer questions about them. ORCA generates hypotheses, searches across every modality, enriches from 22 external OSINT sources, runs adversarial analysis to catch confirmation bias, and flags when claims are structurally unprovable. It's not a better search engine — it's a research methodology. Legal AI tools summarize what they read. ORCA tells you what your case still needs.

Ready to walk into your next case with answers?

See how ORCA turns millions of files into a complete evidentiary picture — so your team knows where the case stands on day one.

Book a Demo