Prompt Playbook: AI Legal Assistant PART 2

One of my first “proper” businesses was a TV station I co-founded in Vietnam.

Out the gate we hired fancy legal experts to help with everything - business setup, licensing, contracts, the works.

One day, I asked for a specific contract, and within minutes (literally minutes), I got an email back. It had a template attached and a list of basic questions to fill in.

In that moment, they'd accidentally shown their hand.

Their whole process was this: take a template, get some basic info from the client, plug it in, make it sound fancy with some legal language. Then charge hundreds per hour for the privilege.

Huh.

Here's the real kicker - I went online and found the exact same template. The one they were using? You could buy it for $50. It wasn’t even hard to find - it was sitting right at the top of Google. 😅 

I'm not sharing this to knock lawyers. But here's the truth: the majority of business legal work follows this exact pattern. Lawyers have their boilerplate templates, they collect specific information from you, and make a handful of changes.

That used to be enough to justify the fees. Not anymore.

AI can do this faster and at a fraction of the cost.

Let’s get started

Boilerplate + Interview method

Getting Your Base Template

First we need our boilerplate template.

There are two basic options.

1. Buy/find an existing template

2. Use AI to generate the template

Honestly the second option works well. AIs have been trained on thousands of legal documents already and can give a good base level boilerplate. It also means we can be sure it's not copied from anywhere. Bonus.

Let's use an NDA as an example, since it's something most businesses need. But this process works for any legal document - employment contracts, terms of service, you name it.

Here's our base prompt

Create a standard template for [type of document]. This should be comprehensive and suitable for business use. Include all standard clauses typically found in this type of document.

Simple as that. The AI will generate a solid base template. For our NDA example, we'd simply specify "Non-Disclosure Agreement" and get a template with all the usual confidentiality, term, and breach clauses.

But here's where most people go wrong - they just take this template and run with it. That's not what we're doing. This is just our starting point.

The Interview Method

We’ll now use our AI as a lawyer to walk us through the customisation process. The basic method is we’ll ask the AI to devise a list of questions (based on the document type), we’ll provide answers and the AI will use them to customise.

Do this below your template. Here's the prompt

You are a lawyer preparing a draft of [type of document]. Interview me to gather all necessary information. Ask one question at a time. Focus on understanding:

All parties involved
Key terms and conditions
Specific requirements or restrictions
Special circumstances or exceptions

Wait for my response before asking the next question.

This interview process is crucial. It's what separates amateur hour from professional-grade documents. The AI will systematically work through everything needed to customise the document properly.

From Answers to Document

Once you've gone through the interview process, you've got all the pieces needed. Now we combine everything:

Using the template provided and my answers to your questions, create a first draft document that reflects our specific requirements. Maintain professional legal language while incorporating all the specific details and requirements we discussed.

The key here is that we're not just filling in blanks. We're creating a document that's properly tailored to the specific situation. As if a lawyer was compiling the document for us…just, a lot cheaper!

But we are not done. We’re going to refine and bulletproof our document.

In Part 3, we'll look at how to review and refine your document. I'll show you how to use AI to punch holes in your document, find weaknesses, and make it stronger through multiple iterations. We’ll use combative prompting to have it find flaws (that we’ll then fix).

Keep Prompting,

Kyle