1. Knowledge Ingestion
Accepts the firm’s existing archive—research memos, legal opinions, advice emails, pleadings, contracts, case analyses, due diligence reports, and internal notes—without requiring reorganisation; handles scanned documents and inconsistent naming.
2. Natural Language Querying
Lawyers describe problems in ordinary language, with no keyword syntax, folder navigation, or tagging vocabulary required.
3. Similar Matter Identification
Matches prior matters to the current problem based on legal issue, jurisdiction, factual pattern, contractual structure, and procedural posture rather than word overlap.
4. Matter Comparison
Systematically compares the previous matter with the current one, highlighting similarities and material differences in facts, clauses, jurisdiction, and procedural posture.
5. Reusability Mapping
Identifies which components of previous work (reasoning, research, drafting language, argument structure, authorities) are potentially reusable, recognising that these degrade at different rates.
6. Change Detection
Flags whether previous analysis may need updating due to later judgments, legislative amendments, authorities being distinguished/overruled, or changes in factual/commercial context.
7. Source-Linked Output
Every substantive statement links back to the exact document, page, paragraph, or clause it was drawn from; nothing is asserted without a traceable source.
8. Reliability Layer (Green/Amber/Red)
Provides graded reliability signals:
o Green: Reusable subject to normal review.
o Amber: Reasoning sound but facts/contract shifted; specific parts marked for rework.
o Red: Conclusion rests on disturbed authority or changed position; do not adopt without review.
9. Drafting Assistance from Past Work
Surfaces previously drafted clauses, notices, pleadings, and advice emails that are structurally similar to the current task, so lawyers can start from the firm’s own proven language instead of a blank page.
10. Clause and Document Reuse with Context
Shows not just the clause text but the surrounding reasoning: why the firm took that position, what risks were weighed, and whether that stance has since been overtaken by later developments.
11. Proofreading and Consistency Checking
Compares a new draft against the firm’s prior positions and documents to flag inconsistencies (e.g., different interpretations of the same clause, conflicting advice on the same point, or deviations from the firm’s usual risk posture).
12. Citation Verification
Cross-checks authorities cited in a draft against the live case-law feed to warn if a case has been distinguished, overruled, or treated differently in a later judgment, reducing the risk of relying on outdated precedent.
13. Confidentiality Architecture (Tiered Access)
Assigns each document a confidentiality state:
o Tier 1 (Client confidential): Matter-level access only; no cross-client retrieval.
o Tier 2 (Firm internal): De-identified reasoning retrievable across the firm.
o Tier 3 (Publishable): Cleared reasoning eligible for the public commons.
14. Public Legal Knowledge Commons
Enables firms to contribute de-identified, cleared legal reasoning to an open, searchable commons free for legal aid organisations, clinics, sole practitioners, and individual users.
15. Integration into Existing Workflows
Surfaces inside tools lawyers already use (e.g., Outlook, Word) at the moment a matter opens, rather than requiring a separate portal.
16. Matter-Closure Learning Loop
Prompts a short, structured note at the end of each matter to record what the firm learned, automatically improving institutional memory for future matters.
17. Audit Logging and Verification Controls
Logs every query, every document surfaced, and every reliability assessment with reasoning and sources; enforces mandatory human verification before any previous work is reused in client-facing advice.
18. Focused MVP Scope (Commercial Contracts under Indian Law)
Initially confined to commercial contract advisory work under Indian law; litigation strategy, tax, and regulatory work are planned as later expansions with their own similarity models.
