Balanced
Competing interpretations and credible counterarguments are surfaced instead of silently collapsed into one answer.
Lawyer-directed · Balanced · Verifiable
A legal AI team that shows its work
The Usus Harness coordinates specialized agents to research, challenge, verify, and draft under a lawyer's direction. It is being designed for consequential legal work, with matter context supplied by a local-first practice system. The same core Harness serves independent lawyers and larger legal teams, with workflow depth matched to the assignment rather than the size of the firm.
In active development · Early access will open in stages · Product direction and available capabilities are labeled separately
Legal AI founders program · Early invitation
Bring the legal work that deserves more than a chatbot answer
Tell us where research, evidence review, argument testing, or drafting consumes the most time. Your note can help determine which workflows the Usus Harness validates first.
A conversation, not a sales sequence. No spam or shared addresses.
The Usus Harness
Balance before confidence. Verification before delivery.
Instead of asking one model for one polished answer, the Harness is designed to coordinate independent roles, preserve disagreement, challenge weak reasoning, and connect material claims to sources a lawyer can inspect.
Lawyer-directed workflow
- 1Assign
The lawyer defines the question, scope, jurisdiction, and approved matter context.
- 2Investigate
Independent roles examine law, facts, evidence, and competing interpretations.
- 3Challenge
A review pass looks for contrary authority, missing facts, and unsupported conclusions.
- 4Verify
Material claims and citations remain linked to sources and unresolved issues stay visible.
- 5Decide
The human lawyer reviews, revises, and approves every consequential work product.
Verifiable
Authorities, factual support, uncertainty, and review status stay connected to the work product.
One standard
The same core verification process serves independent lawyers and larger legal teams. Workflow depth follows the assignment's risk and complexity.
These safeguards are designed to reduce error, not guarantee correctness. AI output remains draft work requiring professional review.
Why legal AI needs a better operating model
The standard of work should not change with firm size
01
One answer hides disagreement
A fluent response can conceal competing authority, alternative readings, and unresolved questions.
02
Confidence can outrun evidence
Hallucinated authority and unsupported facts often look polished enough to escape a quick review.
03
Verification becomes manual labor
Time saved drafting can disappear when the lawyer must reconstruct every source and factual premise.
04
Generic tools lack matter context
Copying fragments into a chatbot loses relationships, provenance, custom data, and the surrounding record.
05
Rigor should not require scale
Firm size should affect seats and usage, not the quality of the reasoning and verification process.
Legal work at the center
The Harness becomes more useful when it understands the practice
Legal AI is the focus. Practice and business management supply the structured matter context that makes research, analysis, and drafting more useful and more inspectable.
01 · The legal work
Coordinate a supervised AI legal team
Research, fact analysis, counterarguments, evidence review, verification, and drafting follow a lawyer-directed process.
Meet the Usus Harness →02 · The practice
Ground the work in the matter record
Matters, relationships, custom fields, documents, deadlines, and evidence give the Harness approved context it can trace.
See the practice foundation →03 · The business
Keep the firm operational around the work
Intake, capacity, time, billing, and financial activity remain connected instead of becoming another set of isolated tools.
See the business layer →Control by architecture
Your firm should remain operational without our permission
1 · Firm deviceWork with readable local data
Records and files remain useful on the device.
2 · Before uploadEncrypt with the firm-held key
Protection happens locally, before synchronization.
3 · Cloud serviceStore and relay ciphertext
The service is designed not to hold the decryption key.
4 · Authorized deviceDecrypt locally for firm use
Readable content returns only inside the firm boundary.
Planned collaboration architecture · Security properties remain subject to implementation, testing, and independent review.
Work locally
Core records live in a database on your device, and documents remain in ordinary folders you control.
Shape the data
Custom and meta fields are designed to participate in forms, search, workflows, reporting, and documents as real data.
Collaborate privately
Planned team sync encrypts data and files on the device with firm-held keys before ciphertext reaches the service.
Keep an exit
Open documents, documented exports, and a tested migration path are intended to keep the firm from becoming captive.
Migration is part of the product
Bring your firm's history with you
A better system is not useful if moving feels unsafe. Usus is developing an assisted migration process for exports from existing platforms, including ambiguous custom data and document relationships.
The standard we are designing toward
- Preserve the untouched source export
- Use deterministic handling for identifiers, dates, money, and relationships
- Use AI to suggest ambiguous mappings—not silently decide them
- Show confidence, exceptions, duplicates, and unresolved records
- Reconcile totals and produce an import report
- Never silently discard a source record
A different economic relationship
Use the standalone system free. Pay when collaboration creates value.
The local edition is intended to remain a real, useful product—not a temporary trial. Paid service begins when a firm chooses encrypted synchronization and team capabilities that create ongoing service costs.
Standalone
Free
Core local practice and business tools, with no Usus cloud account required.
Firm collaboration
Paid
Optional encrypted sync, coordinated access, and services with continuing infrastructure costs.
Useful before launch
Practical guidance for consequential technology decisions
The Usus Journal examines legal software economics, security, professional responsibility, AI verification, and law-firm operations.
A practical way to measure effort, scope drift, staff cost, effective hourly rate, and contribution margin without turning flat-fee work back into hourly billing.
A practical review method for checking legal authority, factual claims, quotations, omissions, and reasoning before AI-assisted work leaves a lawyer's desk.
How to distinguish encryption in transit, provider-managed encryption, customer-held keys, and provider-blind designs when evaluating legal software.
Founders program
Help build the system your firm would choose to keep
We are inviting a small group of solo lawyers and small firms to shape workflows, practice-area structures, migration, and the AI review process through candid conversations and staged product access.