A drafting team,
over your firm's precedents.
Composed from your house templates and your past executed documents. Every claim cited back to the source. Every correction becomes a proposed improvement to the firm's standing instructions, approved by a partner. Your precedents stay in your DMS.
The partner sends one message. The team goes to work.
A partner opens the chat, picks the Senior Attorney agent, and types: “Acme Corp deal — SaaS evaluation. Mutual NDA, 2-year term, standard carve-outs. Get it done.” The senior plans the work in-chat so the partner can redirect at any moment. The drafter composes v1 against the firm's template, retrieving the most relevant clauses from precedent (ranked by relevance × lock-score so signed precedents outweigh drafts). The reviewer walks the firm's checklist and flags two issues. The senior resolves them, re-delegates a revision, validates the result, and returns a clean draft with per-clause citations. Save to matter; the redline the next day picks up the same conversation.
Who’s on it.
One published agent — the one your users talk to. The rest are internal team members with their own skills, retrieval scopes, and models.
The only agent your partners talk to. Plans the work, delegates to drafter and reviewer, resolves conflicts, does the final review, returns the result. Escalates to a partner if the reviewer flags more than 3 issues.
Composes v1 against the firm's house template. Retrieves relevant clauses from precedent, cites each. Follows the firm's standing positions from its drafting skill.
Walks the firm's review checklist. Verifies party identification, carve-outs, precedent citations (must cite to lock-score ≥ 0.9). Flags anything off. Escalates to the senior.
Plan → Delegate → Review → Return.
Plan.
The senior reads the request and writes a short plan in chat so the partner can see it: draft v1 → review → synthesize → return.
Delegate to the drafter.
Sub-thread spawned. Drafter retrieves 4-5 most relevant precedent clauses, composes v1 against the house template, returns with citations.
Delegate to the reviewer.
Another sub-thread. Reviewer walks the checklist, flags issues (e.g. “residual-knowledge carve-out is narrower than firm's standing position”). Returns a memo.
Synthesize.
Senior reads both, decides which issues need revision, re-delegates to drafter for revision, validates against the review memo. Partner watches a streamed narrative: “drafter v1 → reviewer flagged 2 issues → drafter v2 → ready.”
Return.
Senior returns the final draft to the partner with citations and lock-score badges. Save-to-matter action fires: writes draft to `/matters/Acme-NDA-2026/`, notifies the supervising partner via Slack, logs a billing entry.
Every correction becomes a proposed skill update.
When the reviewer flagged an issue on the last conversation, the auto-checkpoint pipeline emitted a skill_refinement item targeting the NDA Drafting Skill (“tighten verbatim-quotation rule to require lock_score ≥ 0.9”). When the partner remarked in chat “we should also check for an export-control clause on US-counterparty NDAs,” the system emitted refinements against both the Senior Attorney Skill and the NDA Review Skill. Once a week, a partner reviews the queue on a dashboard — each refinement shows the original skill, the proposed change, the trigger, and a one-line summary. Approve or reject. On approve, the skill's content is updated, versioned, and becomes the team's institutional voice on the next conversation. Nothing changes without a partner signing off.
Works over the tools you already run.
Data mapped from where it already lives. Nothing uploaded, nothing migrated. Common connections for this team:
Want to see this team on your workflow?
30-minute demo. Bring one specialist task your team keeps re-doing. We’ll show how the team topology fits.