BlitzLearnAI
1 / 9
Day 4 of 14 · AI for Law Firms

Your First Legal Task with Claude

Today you stop reading about AI and start using it. You'll run your first real legal task in Claude Chat — analyzing a contract provision and producing work product you could actually send to a client.

The gap between "AI is a toy" and "AI changed my practice" lives entirely in the quality of your instructions. Today you'll see the difference firsthand.

The prompt is the skill

Most lawyers who try AI write something like "review this contract" and get back something mediocre. Then they decide AI isn't useful for legal work.

The problem is not the AI. The problem is the input. Compare these two approaches:

Prompt A: "Review this contract."

Prompt B: "Review this services agreement from the vendor's perspective. Flag provisions where the customer shifted risk beyond market norms for this type of deal. Check for missing provisions that should be present, including limitation of liability, IP ownership, data handling, and termination for convenience. Produce a severity-rated summary with specific counter-language for each high-severity issue. Note that the vendor has limited negotiating leverage and wants to close the deal, so recommendations should focus on provisions worth fighting for versus provisions to concede gracefully."

Prompt A produces work product that requires extensive revision — if it's useful at all. Prompt B produces work product that is useful on the first pass. The entire gap lives in the quality of your instructions.

Knowledge Check
Why do most lawyers conclude that AI isn't useful for legal work?
A
The AI models are not capable enough for legal tasks
B
AI cannot understand legal terminology
C
They give vague instructions and get mediocre output — the problem is the input, not the AI
D
Legal work is too complex for any AI
The quality of AI output is directly proportional to the quality of your instructions. "Review this contract" produces generic results. A detailed prompt specifying perspective, risk focus, missing provisions checklist, severity ratings, and negotiating context produces work product that's useful on the first pass.

Building a great legal prompt

Every effective legal prompt to Claude includes these elements:

1. Context — What type of document is this? What's the deal? Who is your client?

"This is a Series A stock purchase agreement. My client is the company (seller side). The round is $5M led by a well-known VC fund."

2. Perspective — Whose interests are you protecting?

"Analyze from the company's perspective. Flag terms that are unfavorable to the company or its founders."

3. Scope — What specifically should Claude focus on?

"Focus on: protective provisions, board composition, anti-dilution, information rights, and drag-along provisions."

4. Standards — What should Claude measure against?

"Compare each term to standard market practice for Series A rounds in enterprise SaaS. Flag anything that deviates from NVCA model documents."

5. Output format — How should the analysis be structured?

"Produce a severity-rated table: Critical / High / Medium / Low. For each Critical and High item, include specific counter-language I can propose."

6. Constraints — What practical considerations matter?

"The company needs this funding to make payroll in 6 weeks. Recommendations should focus on must-haves versus nice-to-haves."

🤖
Claude
online
This is a Series A SPA for a $5M round in enterprise SaaS. My client is the company. Analyze the protective provisions from the company's perspective. Compare each to NVCA model terms. Flag deviations. Severity-rate each issue. For Critical and High items, draft counter-language. The company needs this funding within 6 weeks — focus on must-haves vs. nice-to-haves.
**Protective Provisions Analysis — Company Perspective** 🔴 **CRITICAL** **Section 3.4 — Consent Rights over Debt** Current: Investor consent required for *any* debt over $25K NVCA Standard: Typically $100K-$250K threshold for Series A Risk: Blocks routine bridge financing and equipment leases **Counter-language:** "...indebtedness in excess of $250,000 in the aggregate, excluding equipment financing up to $100,000" *Must-have. This will hamstring operations.* 🟠 **HIGH** **Section 3.7 — Board Observer Rights** Current: Two observer seats with full information rights NVCA Standard: One observer seat is standard; two is above market Risk: Governance overhead; observers with information rights create confidentiality exposure **Counter-language:** "One (1) observer seat, subject to exclusion from discussions involving conflicts of interest" *Worth pushing on. Concede if needed to close on timeline.* 4 more issues analyzed. Shall I continue?
↻ Replay conversation
Knowledge Check
Which element is most important in a legal prompt to Claude?
A
Keeping the prompt as short as possible
B
Asking Claude to act like a specific famous lawyer
C
Using formal legal language
D
Specifying context, perspective, scope, standards, output format, and practical constraints
Effective legal prompts give Claude the same briefing you'd give a senior associate: the deal context, whose interests to protect, what to focus on, what standards to measure against, how to format the output, and what practical constraints affect the recommendations. Detail in the prompt produces quality in the output.

Try it yourself

Here's your first exercise. Open Claude Chat and try this prompt (adapt it to a real matter you're working on, or use it as-is for practice):

Practice prompt:

"I represent a SaaS startup that just received a Master Services Agreement from a Fortune 500 prospect. The prospect sent their standard form — we have no leverage to demand our paper. Review the following provisions from my client's perspective and tell me: (1) which terms are standard for enterprise SaaS and I should accept, (2) which terms shift risk beyond market norms and I should push back on, and (3) which terms are missing entirely and I need to add. For each pushback item, suggest specific counter-language that a Fortune 500 legal team would find reasonable. Format as a table with columns: Provision | Status | Risk Level | Recommended Action | Counter-Language."

You don't need an actual MSA to try this. Paste any services agreement you have on hand — or describe the key terms and let Claude work with what you give it.

Notice the difference between this and "review this contract." Same AI. Dramatically different output.

What to expect

Your first few prompts won't be perfect. That's normal. The skill develops quickly because the feedback loop is immediate — you see exactly how your instructions affected the output, and you adjust.

After a few iterations, you'll develop a feel for:

- How much context Claude needs to produce useful analysis

- Where Claude excels (comprehensive issue-spotting, cross-referencing provisions, generating counter-language)

- Where you need to apply your own judgment (deal dynamics, relationship considerations, business strategy)

The goal is not to have Claude practice law. The goal is to have Claude handle the production so you can focus on the judgment.

⚖️
Day 4 Complete
"The gap between 'AI is a toy' and 'AI changed my practice' lives in the quality of your instructions. A detailed prompt with context, perspective, scope, standards, format, and constraints produces work product that's useful on the first pass."
Tomorrow — Day 5
Setting Up Cowork for Legal Work
Tomorrow you'll set up Claude Cowork with a dedicated legal workspace and run your first autonomous document task.
🔥1
1 day streak!