The Lawyer's Claude Handbook · No. 1

The gap between lawyers who use AI and lawyers who leverage it is widening faster than most people realize. This is the first in a series on what that difference actually looks like in practice, specifically for AI, Web3, and fintech legal work.

At a tech law event last month, I asked a room full of lawyers whether they used Claude or GPT. Nearly every hand went up. When I asked how, the answers were mostly the same: summarize this document, clean up this draft, translate this clause.

That''s not leverage. That''s spell-check with extra steps.

The lawyers pulling real value out of these tools are doing something structurally different. They''re not asking Claude to finish sentences, they''re giving it a thinking framework and using it to compress hours of analytical groundwork into minutes. The writing, the judgment, the interpretation: that still belongs to them. But the scaffolding gets built in a fraction of the time.

What follows are five specific patterns I use in AI, Web3, and fintech legal work. Not prompting theory, actual structures, with examples, and an honest note on where the tool still falls short.

The lowest-value use of Claude is asking it to summarize something you could read yourself. The highest-value use is giving it a specific analytical frame and asking it to apply that frame to a problem.

In regulated tech, where the rules are new, contested, or simply absent, that frame is almost always the interesting part. What counts as a crypto-asset service under Turkish law? Is this AI system a high-risk system under the EU AI Act? Does this referral program look like an unlicensed financial promotion?

Claude can''t answer those definitively. But it can build the analytical skeleton your judgment runs on. The shift is from "produce output" to "structure the problem."

A client is entering Turkey with a crypto product. You need a regulatory map. The traditional workflow, scan the regulations, cross-reference the relevant instruments, build a structure, write it up, takes most of a working day. The thinking that matters takes maybe an hour of that.

With the right prompt, Claude produces a usable first-pass skeleton in ten minutes. You spend your time on what actually requires a lawyer: risk interpretation, product-specific nuance, judgment calls on edge cases, and the executive framing that makes a memo actionable.

The "REQUIRES CLARIFICATION" tag is not a workaround, it''s the point. It forces the output to distinguish between settled analysis and interpretive gaps, which is exactly the distinction that matters when you''re billing for judgment.

Marketing compliance is where I spend a disproportionate amount of my time with tech clients. Web3 and AI companies generate content continuously, landing pages, app store listings, influencer briefs, campaign copy, token explainers. Each one carries regulatory exposure. Reviewing everything manually is neither fast nor scalable.

With the right framework, Claude can run a first-pass compliance review that flags problematic claims, explains the regulatory basis for each flag, and suggests alternative formulations. You review the output, make the calls on borderline items, and sign off. The growth team moves. Legal is no longer the bottleneck.

The framing matters. "Rewrite this" produces mush. "Flag, explain, and suggest" produces something a growth team can actually act on.

"Should we launch in Turkey, DIFC, or under MiCA?" This is usually the first legal question a Web3 founder asks, and it''s genuinely difficult to answer well, not because the law is impenetrable, but because the answer depends on product architecture, timeline, budget, investor geography, and risk appetite simultaneously.

Claude can''t make that call. But it can produce a structured comparison across the key variables fast enough that you''re having the real conversation with the founder in the first meeting, not the third.

The table is the starting point, not the deliverable. Your value is in interpreting it for that specific founder, with that specific product, at that specific moment in the regulatory cycle.

You''ve written a thorough, well-structured legal memo. It''s accurate. It''s defensible. The founder opens it, reads the first three paragraphs, and asks you to summarize it on a call.

This is not a failure of the lawyer. It''s a format problem. A 20-page legal analysis and a founder decision brief are two different documents. Most lawyers produce only the first.

The memo doesn''t get shorter, it gets reoriented. That''s the difference between information and advice.

The next issue in this series covers the other side: which prompts consistently fail, why the framing breaks down in regulated tech specifically, and what prompting well actually requires as a legal practitioner. If you found this useful, the best thing you can do is share it with one lawyer who would benefit from it.

Alp Mete Şirin advises AI, Web3, and fintech companies on market entry, marketing compliance, product structuring, and regulatory strategy. He works independently under his own practice at alpmetesirin.com.