🍔 Your Takeaways
Most lawyers prompt AI like a Google search — structured prompting turns the same tool into a completely different output
Claude has 5 features purpose-built for legal work — a prompting formula, persistent project workspaces, firm-specific playbook files, agentic task execution, and a legal plugin with slash commands
A midsize firm cut contract review time by an estimated 60% just by changing how they structured their prompts — no new software, just better technique
The honest part: roughly 5-10% hallucination rate means you still verify everything, but that's notably lower than some legal-specific platforms
You can test all of this before your next meeting — no firm-wide rollout, no IT approval, one person and one plan
This week I'm doing something different.
No trending headlines.
No scary adoption stats.
Instead, I want to show you five things you can actually try with Claude before your next meeting.
I put together a companion video walking through each feature on screen.
But the newsletter goes deeper on the "why it matters for your firm" part.
THE SETUP
🔍 Most Firms Use AI Like Google Search
You've probably tried this yourself.
You open Claude, type "summarize this contract," and get a wall of text that's technically accurate but practically useless.
Then you close the tab and decide AI isn't ready.
That conclusion is wrong.
The tool is fine.
The prompt is the problem.
A midsize litigation group cut their contract review time by an estimated 60% recently.
They didn't buy new software.
They changed how they asked.
THE FOUNDATION
🏗️ Five Words That Change Everything
The fix is a five-word framework: Context, Task, Standards, Tone, Format.
Context sets the scenario — who you are, what you're reviewing.
Task defines the specific output you want.
Standards give the AI a measuring stick — your firm's risk thresholds, jurisdiction norms.
Tone controls who it's writing for — a senior partner, a client, a trainee.
Format tells it how to present the output — table, memo, bullet list.
Same contract.
Same AI.
Completely different output.
You learn this once, use it every day, and it works on every AI platform.
Build Better AI Prompts in 60 Seconds
The 5-Part Legal AI Prompt Builder — pick a template, fill in 5 fields, and get a ready-to-use prompt for contract review, due diligence, NDA triage, and more. Works with Claude, ChatGPT, or any AI tool.
YOUR PLAYBOOK
📋 Stop Re-Explaining the Deal Every Morning
Every time you start a new conversation with Claude, it forgets everything.
You upload the documents, explain the deal, set the context — and then you do it all again tomorrow.
Two features fix this.
Projects give you one workspace per client matter.
Upload the deal documents once, set the context once — "Client is buyer, flag anything affecting purchase price" — and every conversation inside that project starts already informed.
Partners, associates, paralegals — all working off the same foundation.
Custom Skills go a step further.
A Skill is a plain text file that tells Claude how your firm does things.
Your NDA review rules, engagement letter checks, indemnification red lines.
A senior partner writes the NDA playbook once, and every associate in the firm uses it automatically.
Unlike ChatGPT's Custom GPTs, Skills are just files — copy them, email them, put them on SharePoint.
THE AGENT
⚡ Give Claude a Job, Not a Question
Most AI is back-and-forth.
You ask, it answers, you ask again.
Cowork is different.
Upload three contracts.
Tell Claude: "Compare these. Create a spreadsheet with key terms, expiration dates, renewal clauses, indemnification caps, and risk flags for each."
Claude shows you the plan first.
"Here's what I'm going to do, step by step."
You approve.
It reads each contract, extracts the terms, builds the spreadsheet, and presents it for review.
That's a task that takes a junior associate half a day.
The honest caveat: Cowork is still a research preview.
macOS only.
Not captured in audit logs yet.
But for internal work — due diligence prep, matter summaries, contract comparisons — it's already incredibly useful.
🔎 What It Can't Do (And Why That Matters)
Claude still hallucinates.
Roughly 5 to 10 percent of legal citations it generates don't exist or are misattributed.
For context, Stanford research found error rates of an estimated 17% for Lexis+ AI and 33% for Westlaw AI-Assisted Research.
Lower doesn't mean safe.
It means you verify everything.
That's non-negotiable.
Claude also doesn't browse the internet.
If you need this morning's regulatory guidance, use Perplexity for the research and Claude for the analysis.
No Westlaw, no LexisNexis, no paywalled databases.
And the best features — Cowork, Skills, the Legal Plugin — are Mac only right now.
Windows support is expected mid-2026.
YOUR NEXT STEP
🚀 Try This Before Your Next Meeting
Step 1: Use the 5-part formula on your next contract review — free, works today, works on any AI platform.
Step 2: Create one Project for an active matter — Claude Pro is $20 a month, and setup takes two minutes.
Step 3: Watch the full video walkthrough.
I recorded a screen-by-screen demo of all five features.
[Watch: Claude for Lawyers — Every Feature You Should Know in 2026]
We run workshops that walk your entire team through this setup in a single session — and they leave using AI the same day.
Related Legal AI News:
🛠️ 10 Second Explainers - AI Tools & Tech
Perplexity — Like having a research assistant who cites every source.
Useful for real-time legal research where Claude's training data has a cutoff.
NotebookLM — Google's tool that answers questions only from documents you upload.
Zero hallucination by design.
Think of it as a fact-locked research partner.
Spellbook — AI contract drafting that lives inside Microsoft Word.
No tab switching.
Suggests clauses, flags risks, and drafts language from your existing templates.
READER POLL
Which Claude feature would save your team the most time this week?
A) Structured prompting (better outputs from the same tool)
B) Persistent project workspaces (stop re-uploading documents)
C) Automated contract comparison (spreadsheets in minutes)
D) Firm-specific review rules (your playbook, applied automatically)
[Reply with your letter choice] - I'll share the results in the next edition.
My Final Take…
The firms getting real value from AI right now aren't doing anything exotic.
They're not running experimental pilots or building custom platforms.
They're using features that already exist — and using them well.
Five features.
One afternoon to set up.
And your team stops treating AI like a search engine.
Hit reply and tell me which feature you're trying first.
Liam Barnes
We help legal teams implement AI workflows that actually get used.
Grab some time to chat
(if you don’t see a suitable time, just shoot me an email [email protected])
How Did We Do?
Your feedback shapes what comes next.
Let us know if this edition hit the mark or missed.
Too vague? Too detailed? Too long? Too Short? Too pink?
Was this week’s newsletter forwarded to you?
Sign up, it’s free.
