AI-native practice management · drafting · vault
The legal AI that ships with its reasons.
Pollenix Law is a deployable AI-native legal platform. Every answer carries citations into your firm's Vault. Every AI output ships with a Why trace — model, retrieval breakdown, per-event cost. Files stay in your cloud.
Built for transactional, litigation, and tax practices. Practice agents tune the Assistant per posture. Custom workflows + pre-canned playbooks for the work attorneys do every day. Matter-team ACLs enforce ethical walls at the SQL layer of every retrieval — not just at the application layer. Pricing is coverage-based, not consumption-based.
Six tightly-coupled surfaces.
Each cites the others. Each writes to the same Vault. None of the screens require AI to be useful, but every screen is better because AI augments it — and the Why trace is auditable end-to-end.
- I. assistant
Assistant
- Citation-grounded Q&A across the firm Vault. Inline [#N] markers link to evidence cards.
- Hybrid retrieval — pgvector cosine + tsvector ts_rank_cd → Reciprocal Rank Fusion.
- Per-conversation source picker — Vault, firm templates, federal caselaw (CourtListener).
- Voice input via the browser's native SpeechRecognition — no STT round-trip.
- II. agents
Practice agents
- Three posture-tuned variants of the Assistant — M&A, Litigation, Tax.
- Each agent overlays a system-prompt extra + scopes Vault retrieval to relevant document classifications.
- Starter prompts demonstrate the agent's posture; recommended workflows pair alongside.
- Transparent broaden-when-empty so a strict scope never starves an answer.
- III. workflows
Workflows
- Pre-built — Contract Q&A, Deposition summary, Document compare, Discovery triage.
- Document Compare exports redline .docx with red strikethrough deletes + green underline inserts.
- Custom workflow builder — linear chains up to 8 steps with {{input}} + {{step_N}} placeholders.
- Playbooks library — NDA review, due-diligence request list, 50-state survey, negotiation review.
- IV. drafter
Drafter
- AI document drafter scoped to a matter — drafts brief / motion / contract from matter context.
- CourtListener citation verifier runs advisory on every citation in the output.
- Bluebook normalization pass — v./vs., U.S.C. spacing, F.2d, Fed. R. Civ. P., pinpoint commas.
- Why trace persists per draft — model, retrieval, verification.
- V. templates
Templates
- {{matter.title}}, {{client.name}}, {{today_long}} mechanically substituted from matter data.
- {{narrative: ...}} blocks filled by the AI with the substituted draft as context.
- Optional per-template LLM system prompt for firm voice.
- One-click 'Generate for matter' — finished prose, not just a fill-in form.
- VI. analytics
Analytics
- Criterion questions across every matter you can read — 'which NDAs > 5 years?'.
- Hybrid retrieval across all Vault chunks (no matter scoping), ACL-gated by your team membership.
- Per-matter findings: matches yes/no, factual 1-2 sentence finding, evidence chunk citations.
- Green badges for matched matters; click to jump to the matter detail.
Meet attorneys where they actually work.
Lawyers live in Word, Outlook, Slack, Teams, and their firm's DMS. Pollenix Law surfaces in all of them — same backend, same Vault, same matter-team ACL on every call.
-
Word add-in
Office.js task pane. Six actions — improve / redraft / continue / summarize / askvault / cite. Insert-at-cursor + replace-selection.
-
Outlook add-in
Office.js task pane. Five actions — draft reply / summarize / tone check / askvault / cite. Reply-form pre-fill via Office.context.mailbox.
-
Slack bot
/pollenix slash command + @mention event subscriptions. Block Kit responses with citation lines. OAuth v2 install with signing-secret request verification.
-
Microsoft Teams bot
Bot Framework Activity receiver. JWT verification against Microsoft's JWKS. Markdown reply with citation links. AAD tenant id maps to Pollenix tenant.
-
Email-to-Vault gateway
matter-<NUMBER>+<tag>@vault.<tenant>.pollenix.ai routes inbound mail to per-matter Vault docs. SES / SendGrid Inbound Parse / Mailgun Routes — shared-secret bearer auth.
-
Westlaw KeyCite
BYO API key. Signal flags (green / yellow / red) normalized onto a five-value treatment scale. Configurable priority order.
-
Lexis+ Shepard's
BYO API key. Signal letters (A / F / D / C / O / R) normalized onto the same five-value treatment scale as KeyCite.
-
CourtListener
Free Law Project. Always-on free fallback for citation existence checks; integrates into the citator aggregator and as a 'federal caselaw' knowledge source.
-
iManage Work
REST API v2. OAuth bearer or service-credential API key. Folder browse + bulk import. Stamped with source='dms_imanage' and connection label.
-
NetDocuments
REST API v1. Cabinets + folders. OAuth bearer. Stamped with source='dms_netdocs'. Same import limits as iManage.
Auditable AI for regulated work.
The same governance contract as the rest of Pollenix — applied to the unique demands of legal practice.
- Matter-team ACLs at the SQL layer — pgvector + tsvector retrieval both apply EXISTS clauses against legal_matter_team.
- Per-call AuthContext capability tokens (≤30s TTL, single-use, audit-logged).
- Why trace on every AI output — model, retrieval breakdown, by-source counts, fine-tune flag, per-event cost + latency.
- Conflict-of-interest checks persisted on every new client (ABA Rule 1.7 audit trail).
- secret_box (Fernet) envelope encryption on every credential blob — citators, DMS, fine-tune, bots.
- BAA available. Files stay in the customer's cloud — Pollenix is deployable, not SaaS.
- Citation verification is advisory, never blocking — model output is auditable but not gated by the verifier.
Coverage-based, not consumption-based.
Flat fee per active attorney seat. No per-query metering.
Deploy through AWS Marketplace or Azure Marketplace, or talk to us about a managed install. Files stay in your cloud. BAA available.