AI for Lawyers That Knows Your Cases

Updated January 2026 | 7 min read

You're prepping a motion for summary judgment. The opposing counsel filed late discovery requests last month. Your client's deposition revealed three key admissions. The judge in this district has a documented pattern of granting MSJ in cases with similar fact patterns.

You ask ChatGPT to help draft the argument section.

It gives you generic summary judgment language that could apply to any case in any jurisdiction. Useless.

So you paste the case background, the procedural history, the relevant admissions, the judge's previous rulings, the local rules for page limits and formatting. You're not using AI to practice law. You're teaching it your case file so it can give you something better than a form book.

Next week, different case, same problem. ChatGPT has forgotten everything.

Why Generic AI Fails Lawyers

Legal AI tools give you two options: pay $80/month for a research database that still doesn't know your cases, or use ChatGPT and re-explain everything every time.

Neither option gives you what you actually need — an AI that knows the details of your active matters, remembers your jurisdiction's procedural quirks, understands your practice area's terminology, and maintains context about opposing counsel, judges, and case strategy.

Want to draft a demand letter? You explain the client's position again. Need to prepare deposition questions? You re-enter the discovery responses and testimony from other witnesses. Writing a brief? You paste the standard of review, the relevant statutes, the case citations you want to distinguish.

The AI can write. But it can't remember that Judge Martinez hates verbose briefing, that opposing counsel in the Morrison case always files extensions, or that your client in the Patterson matter has a credibility problem you need to work around.

What Legal AI Actually Needs

A lawyer needs AI that knows the case file:

Case details. Parties, claims, defenses, procedural posture. Key facts, disputed facts, undisputed facts. Timeline of events. Evidence gathered. Witnesses identified. Expert reports filed.

Jurisdiction rules. Local court rules, standing orders, individual judge preferences. Filing deadlines, page limits, formatting requirements. Substantive law quirks in your district. Recent precedent that matters.

Client history. Prior matters, communication preferences, business context. Risk tolerance, litigation goals, settlement authority. Internal contacts, decision-makers, document custodians.

Opposing counsel patterns. Who you're up against, their typical strategies, their strengths and weaknesses. Motion practice history, discovery conduct, settlement posture. What works against them, what doesn't.

Document strategy. Your brief style, your motion templates, your standard argument structures. Citation format, headings hierarchy, fact section organization. Language that's worked in previous matters.

ChatGPT has none of this. Every drafting session starts with a 500-word context dump that you hope covers everything relevant.

How It Works

CLAUDE.md is a markdown file on your computer. It stores your practice context. Claude Code reads it every session.

You document once:

• Practice areas: commercial litigation, employment defense
• Jurisdictions: EDNC, WDNC, North Carolina state courts
• Common opposing firms and counsel tendencies
• Local rules and judge preferences
• Your brief writing style and citation preferences
• Standard templates for motions, discovery, and correspondence

Then you maintain case files in markdown. Morrison v. TechCorp gets a file. Patterson employment dispute gets a file. Each one tracks the details that matter.

Morrison's file says: "Commercial contract dispute, EDNC, Judge Martinez. Plaintiff claims breach of software license agreement dated Jan 2024. Key issue: whether modification clause requires written approval from CEO or any officer. Our client's position: email from VP sufficient under course of dealing. Opposing counsel: Davidson & Hughes (always file kitchen-sink discovery). Discovery closes March 15. MSJ deadline April 30."

When you ask Claude to draft discovery requests for Morrison, it knows the case theory, the key disputed issue, the jurisdiction, the deadline pressure, and the fact that Davidson & Hughes will object to everything so you need to front-load your strongest requests.

Your case files stay local. No client confidences in the cloud. No ethics violation. No malpractice exposure. Obsidian on your machine, Claude reads local files.

What Changes

Before: Drafting a response to a motion to dismiss takes three hours. You outline the arguments, write the fact section, insert case citations, format the brief, proofread. Most of that time is translating what's in your head into written argument.

After: "Draft response to MTD in Patterson case, focus on Twombly plausibility standard and our amended complaint's specific allegations." Claude knows Patterson's facts, the claims at issue, your previous briefing on this matter, Judge Thompson's standards. First draft done in 10 minutes. You spend an hour refining arguments and checking citations.

Before: Opposing counsel sends a meet-and-confer letter about discovery disputes. You read it, check the case file, review what you've produced, draft a response. Takes 45 minutes for what should be a quick turnaround.

After: "Draft response to Davidson & Hughes discovery meet-and-confer, Morrison case." Claude references what's been produced, the outstanding requests, your objections, the discovery deadline. Response ready in 2 minutes.

Before: Client emails asking for status update on three active matters. You pull up each file, check the docket, remember what's pending, write a summary for each. 30 minutes of context-switching.

After: "Draft client status email covering Morrison, Patterson, and Chen matters." Claude knows the current posture of each case, upcoming deadlines, pending motions. You review for accuracy and send.

Before: Preparing for a deposition means reviewing discovery responses, prior testimony, relevant documents. You make notes about topics to cover, inconsistencies to explore, admissions to lock down. Then you draft your question outline from scratch.

After: "Deposition outline for Morrison case, witness is plaintiff's VP of Operations." Claude knows the disputed facts, references the witness's interrogatory answers and document production, suggests topics based on case theory. You refine the strategy and add follow-up sequences.

Before: You need to file a motion in a new jurisdiction you don't practice in often. You spend 20 minutes on the court's website finding local rules, checking formatting requirements, figuring out filing procedures.

After: Context file tracks rules for every jurisdiction you've appeared in. Claude references EDNC local rules, Judge Martinez's standing order, the page limit for motions. You draft without the research detour.

What You Get

Claude Code + Obsidian setup. CLAUDE.md configured for legal practice. Case file templates. Document workflows. Citation and brief formatting standards. 90-minute setup session.

You'll know how to:

• Structure case files Claude can reference
• Track procedural deadlines and case developments
• Document jurisdiction-specific rules and preferences
• Maintain opposing counsel intelligence
• Store brief templates and argument frameworks
• Keep client matter context accessible

Everything local. Your case files, your strategy notes, your client information — stored in Obsidian on your machine. Claude reads local files, you keep control.

$997. One setup. AI that knows your docket.