Claude Cowork for Discovery Organisation: Summarising Documents at Scale

Process thousands of discovery documents automatically. Generate privilege logs, relevance tags, and chronologies without the manual drudgery.

Published 6 November 2025 12 min read

Discovery is the largest single cost centre in litigation. You know this: 200 documents arrive on Tuesday, your paralegal spends three days reading, tagging, and summarising. By Friday, they've barely finished and the privilege log is still a spreadsheet full of "undecided." Discovery organisation with Claude Cowork reduces that three-day sprint to four hours.

This article is part of our Claude Cowork for paralegals series. If you're new to Cowork, start with the Claude Cowork guide. You'll also want to explore 8 Claude Cowork tricks every paralegal should know and Claude Cowork for legal document preparation.

The Discovery Problem: Scale Without Headcount

A typical mid-market firm receives 5,000-50,000 documents per matter. Each document needs:

  • Privilege review: Attorney-client privilege, work product doctrine, attorney-client communications flagged for privilege log
  • Relevance assessment: Coded to custodian, date range, subject matter threads
  • Chronological tracking: Timeline of key communications, decisions, events
  • Narrative summaries: One-line and paragraph-level summaries for attorneys preparing for depo
  • Custodian mapping: Which document belongs to which custodian, export for Relativity or Everlaw

A paralegal working at full capacity processes 30-50 documents per day. A 200-document production takes 4-5 days. A 2,000-document production takes 5-6 weeks. And the firm is paying $65-100/hour for this work. You're not—you're burning hours that could go to client analysis, strategy, or motion drafting.

Claude Cowork solves this. The Discovery Sprint workflow automates the triage, categorisation, and summarisation layer. Your paralegal reviews Cowork's output, makes judgment calls on privilege and relevance, and exports to your case management system. Total time: hours, not days.

How Claude Cowork Processes Discovery Documents at Scale

Cowork uses two core features for discovery work: batch document upload and canvas-based workflows.

Batch upload: You load your export file (PDF, DOCX, TXT, or even OCR'd images) into a Cowork canvas. The system reads all documents in sequence, processes each one through your discovery prompts, and generates a structured output table.

Canvas workflows: A canvas is a shared, living document where you load documents, Claude processes them, and you review/edit the results in real time. No round-tripping. No manual copy-paste. Just load → process → review → export.

Here's the workflow in practice:

  1. Export your document set from Relativity, Everlaw, or your production folder (ideally with metadata: custodian, date, subject line)
  2. Create a new Cowork canvas titled "Discovery Sprint: [Matter Name]"
  3. Upload all documents to the canvas (Cowork supports batch uploads up to 100 documents per session)
  4. Load your discovery prompt template (see below)
  5. Claude processes each document and populates a summary table with fields: Document ID, Custodian, Date, Subject, Privilege Flag, Relevance Code, Summary, Notes
  6. Your paralegal reviews the table, makes final calls on privilege, adds attorney notes, and exports to CSV
  7. Import CSV into Relativity/Everlaw, or use it as your privilege log source of truth

Document Categorisation: Privilege Logs, Relevance Tags, Chronologies

The Discovery Sprint generates three key deliverables:

1. Privilege Log

Cowork flags documents that appear to contain attorney-client communication, work product, or joint defence. Your prompt tells Claude what legal standards your jurisdiction applies (Rule 26(b)(3), Hickman, etc.). Claude flags candidates; your attorney makes the final call. This reduces privilege review from 40% of your timeline to 10%.

2. Relevance Tags

You define your relevance codes (e.g., "Contract Negotiation," "Regulatory Compliance," "Breach of Warranty"). Claude reads each document and assigns one or more codes. Export to your case management system with these tags pre-populated. Attorneys can search and filter by relevance in seconds.

3. Chronologies

Cowork extracts key dates, signatories, and decision points. You get a structured timeline of critical communications. This is gold in deposition prep—your attorney can walk through the case chronologically without reading 200 pages first.

See the Claude Cowork for lawyers article for how attorneys use these timelines during trial prep.

The Discovery Sprint Workflow: Load → Categorise → Summarise → Flag → Report

Here's the named workflow your team will run every discovery cycle:

The Discovery Sprint Workflow

Frequency: Per production, typically daily or weekly
Time commitment: 2-4 hours for 100-200 documents (upload 30 min, processing 30-60 min, review 60-90 min)
Output: Privilege log draft, relevance-coded summary table, chronology, export-ready CSV

Step 1: Load

Export documents from your discovery platform with metadata columns (if available): Document ID, Custodian, Date Sent, Subject, File Size. Upload to Cowork canvas. Cowork reads all documents and extracts text (handles PDFs, scans, Word docs, images).

Step 2: Categorise

Claude assigns each document to:

  • Privilege category: Attorney-client, work product, privileged opinion, non-privileged
  • Relevance code: Your custom codes (e.g., "Contract Formation," "Breach Notice," "Mitigation")
  • Custodian: If metadata is incomplete, Claude infers from To:/From: headers

Step 3: Summarise

Claude generates:

  • One-line summary: Subject line replacement (40 words max)
  • Full summary: Paragraph-level narrative of key points (100-200 words)
  • Key dates: Any dates mentioned (signing, notice, breach, response)
  • Signatories/parties: Who is communicating

Step 4: Flag

Claude flags for attorney review:

  • Documents with privilege language ("attorney," "legal advice," "confidential," "privileged communication")
  • Documents outside your custodian/date range
  • Duplicate or near-duplicate documents
  • Documents with unusual metadata (zero dates, corrupted text, suspicious file types)

Step 5: Report

Export the summary table as CSV. Your paralegal adds a final column: "Privilege Decision" (yes/no/under review). This becomes your privilege log. Upload to Relativity/Everlaw, or keep as a master spreadsheet.

Ready-to-Use Prompt Templates

Here are three templates you can copy directly into your Cowork canvas:

Template 1: Privilege Log Generator

You are a paralegal assistant specializing in privilege log generation for litigation. For each document, analyse the content for attorney-client privileged communication, work product doctrine, and attorney opinion work product. Assess the document against Rule 26(b)(3) of the Federal Rules of Civil Procedure and any applicable state-specific privilege laws. Output a structured analysis: 1. Document ID: [assign sequentially] 2. Privilege Status: "Attorney-Client Privilege", "Work Product", "Attorney Opinion", "Non-Privileged" 3. Legal Basis: Cite the specific rule or doctrine (e.g., "Rule 26(b)(3) - prepared in anticipation of litigation") 4. Privilege Holder: [Name of privilege holder, typically the company] 5. Assertion Strength: "Strong", "Moderate", "Weak", "Non-Privileged" 6. Recommendation: "Include in Log", "Review Further", "Not Privileged" 7. Notes for Attorney: Key language supporting your assessment Be conservative: when in doubt, flag for attorney review rather than exclude.

Template 2: Document Summary and Relevance Coder

You are a discovery paralegal specialist. For each document, extract and code the following: 1. Document Metadata: - Document ID (unique identifier from metadata) - Custodian (person who held the document) - Date Sent/Created - From / To / CC (full names and titles) - Subject Line 2. Document Type: Email, Letter, Memo, Contract, Report, Deposition, Other 3. Relevance Code (select from: Contract_Formation, Breach_Notice, Remediation, Regulatory_Compliance, Executive_Decision, Other) 4. One-Line Summary: [40 words max; attorney-ready summary] 5. Full Summary: [100-150 words; key facts, action items, decisions] 6. Key Dates: [Any dates mentioned that affect the matter timeline] 7. Critical Facts: [3-5 bullet points of litigation-critical information] Format output as a CSV-ready row: ID | Custodian | Date | Type | Relevance | One-Line | Full Summary | Key Dates | Critical Facts

Template 3: Chronology Builder

You are a litigation chronology specialist. Extract all date-specific events from each document. Create a chronological entry for: - Date (YYYY-MM-DD) - Event Type: Communication, Contract Event, Breach, Notice, Remediation, Other - Parties Involved: [Names and titles] - Summary: [One sentence describing the event] - Document Reference: [Document ID and page number] Build a master timeline. If multiple documents reference the same event, consolidate them under one chronology entry. Output as: Date | Event Type | Parties | Summary | Document IDs

Integration with Relativity, Everlaw, and Export Formats

Once Cowork generates your summary table, you have several export paths:

Relativity Integration

Export the Cowork table as CSV. Map columns to Relativity fields: Document ID → Control Number, Privilege Flag → Privileged (Yes/No), Relevance Code → Custom Field 1, Summary → Document Summary. Bulk import into your Relativity workspace. Your privilege log is now queryable in Relativity; no manual spreadsheet required.

Everlaw Integration

Everlaw supports CSV import with custom metadata. Export from Cowork, map your columns to Everlaw's document fields, and bulk-upload. Everlaw then automatically indexes and searches your summaries and tags.

Case Management Systems (Clio, LawLabs, etc.)

Most legal CMS systems support CSV import. Your Cowork export becomes the source of truth for document metadata. Your document review becomes tag-based filtering rather than manual searching.

Local Export (No Integration)

Download the summary table as Excel. Share with your attorneys. They use it as a reference during deposition prep, trial, and settlement negotiations. Your privilege log is in column C; your chronology is sorted by date; your relevance codes are in column D.

Before and After: A Real-World Example

Scenario: A contract dispute. 200 documents spanning two years. One paralegal. Seven-day deadline for discovery production.

Before Claude Cowork

  • Day 1-2: Paralegal manually reads and summarises 50 documents, fills in privilege log manually (paper-based notes, then spreadsheet)
  • Day 3-4: Paralegal continues reading, starts categorising by relevance (no consistent codes yet)
  • Day 5-6: Privilege review by junior attorney (re-reads documents, second-guesses paralegal, marks 15 for further review)
  • Day 7: Privilege log still incomplete, 10 documents still "under review," production delayed
  • Cost: 35-40 billable hours of paralegal time, 10 billable hours of attorney review

After Claude Cowork

  • Day 1: Paralegal exports 200 documents, uploads to Cowork canvas (30 minutes). Sets up Discovery Sprint workflow with privilege and relevance templates (15 minutes)
  • Day 1 afternoon: Cowork processes all 200 documents (1 hour). Generates privilege log candidates, relevance codes, chronology (automated)
  • Day 2: Paralegal reviews Cowork's output, makes final privilege calls, edits 3-5 summaries that need attorney voice (2 hours)
  • Day 2 afternoon: Attorney spot-checks 20 random documents for privilege accuracy and summary quality (1 hour)
  • Day 3: Export to CSV, import to Relativity, production ready. Bonus: Chronology already built and indexed
  • Cost: 6 billable hours of paralegal time, 1 billable hour of attorney time
  • Savings: 38 billable hours, 80% time reduction

See Claude Cowork for legal practice management for more on time tracking and efficiency gains.

Getting Started: Your First Discovery Sprint

You don't need special setup. Here's your first-run checklist:

  1. Create a Claude Cowork deployment in your firm (handled by your IT team; takes 2-3 hours)
  2. Onboard your discovery paralegal with access to Cowork and a sample canvas
  3. Run a pilot: Choose a small production (50-100 documents) from a closed matter
  4. Use Template 2 (relevance coder) on the pilot batch
  5. Have your attorney review the output, give feedback on summary tone and privilege calls
  6. Refine your templates based on feedback (add custom relevance codes, adjust privilege language)
  7. Deploy to your next active matter discovery production

Most firms run their first full Discovery Sprint within 1-2 weeks of Cowork deployment. By month 2, it's routine.

FAQ

Can Claude Cowork handle OCR'd documents and images?

Yes. Cowork uses Claude's vision model to read handwritten notes, scanned PDFs, and image-based documents. OCR quality varies by document age and scan resolution, but Cowork handles most discovery-quality scans. For heavily degraded images, Cowork will flag the document and note that manual review is needed.

What if our privilege rules differ from federal Rule 26(b)(3)?

Edit Template 1 to reflect your jurisdiction's privilege rules. If you're litigating in California, substitute state-specific privacy law and attorney-client privilege doctrine. If you're in the UK, substitute LPP (Legal Professional Privilege) rules. Your attorney should review the template before running the workflow. Cowork follows your custom prompt exactly.

How many documents can we process in one batch?

Cowork can process 100-200 documents in a single session without performance issues. For larger productions (2,000+ documents), split into batches of 100 and run the Discovery Sprint workflow four times. This also gives you checkpoint moments to review and refine your templates mid-production.

Can Cowork detect attorney-client privilege automatically?

Cowork can flag documents that contain privilege language and key indicators (attorney involvement, confidentiality markings, "legal advice," etc.). However, privilege is a legal conclusion. Final privilege determinations must be made by an attorney. Cowork's role is to reduce review time by 80% by pre-flagging candidates and generating a first-pass analysis. Your attorney makes the final call.

Next Steps

Discovery organisation is the entry point for most firms adopting Claude Cowork. Once your paralegals see the time savings—4 hours instead of 3 days—they'll start asking what else Cowork can automate. Billing narratives. Client reporting. Motion drafting. It starts with discovery.

Ready to deploy? Book a free strategy call with one of our Claude Certified Architects. We'll walk through your discovery workflow, design your custom templates, and schedule your first Discovery Sprint. You'll see time savings in your first week.

Also explore Claude Cowork for billing narratives and how Claude Cowork helps paralegals handle more matters without burning out.

Ready to Scale Your Discovery Workflow?

Let our Claude Certified Architects design a Discovery Sprint workflow tailored to your jurisdiction, case types, and discovery volume.