The claim, proven
Every claim, traceable to the source.
Citation ledger · live sample
WHAT IT DOES
Pick the job. We'll build the system.
Two modes. Process inbound documents, or query your existing library. Pick one or both.
Process inbound
Contract review
- Clause extraction
- Risk flagging
- Signoff trail
- Audit log
- Citation back
Compliance and policy checks
- Policy library
- Inbound validation
- Auto-flag gaps
- Reviewer queue
- Signoff
Query your corpus
Internal search / RAG
- Vector index
- Citation-verified
- Permission-aware
- Multi-source
- Slack/Teams
AI-powered research
- Parallel queries
- Citation-verified
- DD memos
- Source weighting
- Word export
WHY IT HOLDS UP
Citations on every output. Audit trails included.
Four capabilities that decide whether a document system holds up in production.
PLUGS INTO YOUR STACK
We plug into the stack you already paid for.
- Pinecone
- Qdrant
- Weaviate
- pgvector
- S3
- SharePoint
- Box
- GDrive
- OpenAI
- Anthropic
- xAI
- Okta
- Azure AD
- Google Workspace
- Auth0
- CSV
- API
- Salesforce
- Your CRM
Or anything else with an API. Yours included.
Portable by design
Your VPC. Your endpoints. No license server.
HOW IT STAYS ACCURATE
Four things you're worried about. Four built-in answers.
Every concern below has a mechanism behind it.
WHAT IF
“It hallucinates a clause.”
EVERY CITATION LINKED TO SOURCE SPAN
Every extracted clause carries a page-and-paragraph citation back to the original document. Production track record: 0 hallucinated citations across 200+ contracts/month over 12 months of legal deployment.
WHAT IF
“Confidential documents leave my VPC.”
PRIVATE-VPC / ON-PREM / AIR-GAPPED
Build runs in your cloud. Documents never touch our infrastructure. SOC 2 / HIPAA available; private model endpoints supported.
WHAT IF
“Regulators and partners want an audit trail.”
NAMED SIGNOFF · EXPORTABLE LOG
Every output document carries a named human reviewer signature line. Every flagging decision is logged and exportable as a signed audit trail for SOC 2, HIPAA, or regulatory review. The AI accelerates the review; the named human owns the work product.
WHAT IF
“The model gets out of date.”
WEEKLY EVAL · DRIFT DETECTION
Production systems run a weekly eval suite against a held-out corpus you control. If accuracy drifts, we tune the prompts, swap the model, or refresh the retrieval index. You see drift before it reaches your work product.
All outputs citable. All decisions reviewable. All data yours.
HONEST PRICING
Year-1 totals. Fixed at scope.
Pilot starts at $59,995 Year-1. Harvey's 20-seat floor is $288,000. Standard, Pro, and Bespoke scale from there.
Year-1 total, all-in. No per-seat license. No renewal cliff.
Document Pilot
A single document type. Extract or classify on a stable taxonomy.
from $59,995/yr
from $288,000 typical
Harvey 20-seat floor
BUILDS IN 3-5 DAYS
$24,995 build + 12 mo ops
- One document type (contract, invoice, claim, etc.)
- Citation-verified extraction
- Confidence-threshold handoff to human
- Web UI + CSV export
- One round of revisions
Document Standard
Multiple document types. Multi-format ingest. Audit trail for regulated environments.
from $119,995/yr
from $300,000 typical
Harvey + CoCounsel mid-market
BUILDS IN 7-10 DAYS
$44,995 build + 12 mo ops
- Up to 5 document types
- PDF, DOCX, scanned image (OCR), email ingest
- SOC 2 / HIPAA available
- Vector index of your own corpus (RAG)
- Citation + signoff trail per output
- API + CRM export
- 2 rounds of revisions
Document Pro
Mixed workflow: review + internal search + research. Regulated industries.
from $199,995/yr
from $400,000 typical
AmLaw legal-AI floor
BUILDS IN 10-14 DAYS
$89,995 build + 12 mo ops
- Everything in Standard
- Both inbound processing and corpus retrieval (RAG)
- AI-powered research and due diligence flows
- HIPAA / SOC 2 / GDPR compliant deployments
- Private VPC, no third-party log retention
- Multi-region data residency
- Dedicated build team
- Unlimited revisions during build window
Bespoke
Anything you can imagine.
from $279,995/yr
custom typical
Six-figure custom
BUILDS IN 14-30 DAYS
$129,995 build + 12 mo ops
- Everything in Pro
- Custom architecture (air-gapped, multimodal, large corpora)
- Fine-tuned models on your domain corpus
- Custom integrations (internal APIs, ERP, EHR, DMS)
- Direct line to the build team
- Dedicated SRE for production
Every Year-1 price above is fixed at scope agreement. Build-only price: $24,995 / $44,995 / $89,995 / $129,995+.
Real outputs · across four industries
The artifact you forward to procurement.
Switch the tab. Each output is a real review your team would receive: every claim paired to its exact citation, every line traceable.
HOW WE'D BUILD YOURS
From your document problem to live. In days, not quarters.
3 to 14 business days from discovery to production. Less than the typical vendor takes to send the first scope deck.
1
Discovery
0 to 1 dayWe map your document problem against 5-10 example documents. Which clauses or facts matter, what reviewer signoff looks like today, which tier fits.
2
Ingest + index
2 to 4 daysWe ingest your example corpus, model your taxonomy, build the citation-verified retrieval layer, and wire up the human-in-the-loop review UI.
3
Pilot
1 to 2 daysThe system goes live on a held-out batch of documents. Your reviewers see flagged sections with citations. We iterate on the taxonomy and thresholds within hours, not weeks.
4
Production
Day 0 onwardContinuous monitoring. Weekly review for the first month. We hand over the keys, you tell us when to come back.
What you receive · day by day
First 14 days · Pilot tier
Standard and Pro scale similarly. Bespoke timelines vary with scope.
The system caught clause conflicts our senior associates were missing. It's not replacing lawyers, it's making them faster.
Bring us your document problem. We'll show you the working version.
Tell us the document type, the constraint, and what success looks like. We'll reply with a real number and a real timeline.