Atlas Lists: Reusable Agent Workflows for Every Matter

The Playbook Problem

Every AmLaw practice runs the same workflows over and over. M&A teams privilege-review 50 deals a year. Patent counsel diligence 200 patent portfolios. Litigation teams extract privilege from 100-document data rooms. Each time, the same steps: read documents, classify them, flag issues, generate findings, draft output.

But today, those workflows live inside individual brains, in chat histories, in ad-hoc prompts teams swap over Slack. A partner builds a privilege review prompt in Claude, runs it once, then next deal the new associate starts over. Or worse, the review process drifts because the prompt got lost or the precedent changed.

What if instead, teams could codify their workflows once and run them agentic-scale across every future matter, with the firm's own iManage and SharePoint content grounding every decision?

What We Built

Atlas Lists are reusable sequences of agentic actions. A practice team authors a list once: a series of steps agents will execute on documents staged in an Atlas Workspace. Each step is either an agentic action (extract clauses matching X pattern, flag documents with Y risk markers, generate markup from Z findings) or a human gate (require partner approval before drafting the final memo).

Once the list is built and saved, it becomes repeatable firm IP. Upload a new deal's documents to a Workspace. Run the list. Atlas agents execute every step in parallel across all 200 documents, classify, route, and extract, then draft new client documents (SPA markups, privilege logs, diligence memos) directly back into the workspace. No re-prompting. No drift. Same playbook, every time.

Lists are matter-scoped: they live inside Workspaces, operate on the documents staged there, and write their output back to the same workspace. But they're reusable across Workspaces: an M&A team builds one Deal Review list; 30 deals later, they run the same list again.

How It Works

Under the hood, Lists are grounded in the firm's curated knowledge graph. When a list executes, agents read the documents in the Workspace against the firm's iManage and SharePoint precedents, firm closings, and prior diligence findings, all already indexed into the graph by Atlas's curation agents. A privilege review list doesn't just flag docs with privilege language; it compares them to the firm's own prior redactions and counsel-sign-off patterns.

This week we shipped the prompt library redesign and live polling for shared Lists. Teams can now author lists collaboratively, refine them in real-time, and see agent execution update live as shared-workspace participants watch the same workflows run. A partner can tweak the diligence memo template mid-execution and see it apply to the next batch of 50 documents without stopping the agent.

In Practice

Consider a typical M&A close. The deal team has 180 seller-provided contracts. Before Lists, an associate would: manually load each contract into chat, prompt the model to extract key reps and warranties, flag indemnification caps, compare to the draft SPA, then generate redline comments. 180 documents, 180 manual prompts. 30 hours of work.

With Lists, the M&A partner built a "Deal Diligence" list six months ago: extract key reps/warranties, flag indemnification triggers, compare to firm SPA template, extract any survival periods, draft findings memo. That list is now run-once. Upload the 180 seller docs to the Workspace. Run the Deal Diligence list. Agents execute all three steps in parallel, return a 50-page diligence memo with granular findings, and auto-generate an indemnification schedule with flagged gaps. Same day. Same playbook.

Another example: IP counsel runs a "Patent Diligence" list on every portfolio acquisition: extract independent claim scope, cross-reference to prior prosecution history (from iManage), compare to known competing patents, flag any obviousness red flags based on prior art in the knowledge graph, draft diligence findings. Once built, the list runs identically on every new portfolio.

Why This Matters

Am law practices have spent a decade externalizing knowledge work: associate work moved to outside counsel, document review moved to review vendors, diligence moved to third-party data-room companies. But AI is collapsing the cost of in-house review. If diligence can run agentic, and agents can read your own firm's precedents at scale, the work is coming back inside.

Lists are how that happens. Not as a chatbot or a one-off prompt, but as repeatable, auditable, firm-branded workflows grounded in the firm's own data and embedded in the firm's own Workspaces. The same way Salesforce sales teams run playbooks, legal teams now run agentic workflows.

How to Try It

Lists are live in Atlas v6, inside every Workspace. If you're already staging documents in a Workspace, you can author your first list by clicking "New list" in the prompt library. Pick a workflow your team runs every deal or every matter (privilege review, due diligence, redaction, closing checklist). Author the steps. Save it. Next matter, upload docs, run the list, watch agents execute at scale.

Lists execute inside the same agent orchestration layer that curates your knowledge graph, so every agent decision is grounded in your firm's iManage and SharePoint precedents. And Lists integrate seamlessly with Agents (the execution layer), Workspaces (the matter scope), and MCP + API for Claude (if you want to run Lists outside the UI). One platform, one graph, reusable workflows.