When we launched the AI Workshop Kit licensing program, I thought we might get 50, maybe 100 applications. You know, a manageable number I could handle manually.

We got over 600.

Suddenly, preparing individual license agreements wasn't just time-consuming - it was practically impossible.

Harminder, my business partner was suddenly a full-time license agreement preparer!

Each application needed to be reviewed, details extracted, and converted into proper legal language to be written up into agreements. Doing this manually would have killed the momentum of the launch completely. And maybe killed Harms too!

That's when we realised - when you're dealing with volume, automation isn't just nice to have. It's imperative.

Let’s get started:

When Automation Makes Sense

This part is more advanced and only for if you have a recurring task. If you're creating a one-off contract or occasional documents, the processes we covered in the previous parts are plenty.

Don’t automate one-offs or even small batch tasks. It’s tempting but a waste of time. Just do the work. Boooo!

But if you're regularly creating similar legal documents - like service agreements for clients, NDAs for contractors, or licensing agreements for products - look into automating the process.

Specifically here’s what we're trying to solve: Your clients rarely give you information in a format that plugs straight into a legal document.

They might tell you what they want in everyday language on a call or via email. And converting that into proper legal terms takes time.

Do this once? No problem. Do it hundreds of times? That's a different story.

The Basic Flow

Think of it like this:

• Input = Client details and requirements

• Process = Your legal document template + AI processing

• Output = Clean, properly formatted legal document

First, we need to know exactly what information we need to collect. Here's the prompt to help with that:

Review this legal document template and identify all information that needs to be collected from the second party. List:

Required personal/business details
Specific terms that need to be agreed upon
Options or variations that need to be selected
Any conditional elements that might apply

Format the response as a structured data collection plan.

Once you know what you need to collect, it's time to set up your form.

I use HeyForm but JotForm or TypeForm work too. The key is making sure it can connect to Zapier or Make.com so we can set up a workflow!

Create your form using the fields identified by the prompt above. Make it easy for clients to fill out, but thorough enough to get all the information you need. Then publish it with a public link. That’s input sorted.

The Automation Magic

This is where it gets interesting. In Zapier or Make.com, you'll create a flow that:

1. Captures form submissions

2. Sends the data to your AI of choice (Claude or ChatGPT)

3. The AI processes the response

4. Creates your document

For the AI processing, here's the crucial prompt:

Using the provided legal document template and the following form responses, create a properly formatted legal document. Maintain professional legal language while incorporating all specific details provided. Format all dates, numbers, and proper nouns according to legal document standards.

Template: [Your template]
Form Responses: [Form data]

This is a starting point. You’ll likely have to make adjustments based on what sort of document and inputs you are working with. Crucially we make adjustments based on doing dry run.

Test this ruthlessly before going live. Start with dummy data, then real test cases. Look for edge cases and weird inputs that might break your flow. I covered all of this in the Prompt Playbook on Blueprint Prompts. Hop into that course for more techniques on refining reusable prompts.

For output format, think about your needs:

• Need to edit? Output as .doc

• Ready to sign? Go straight to PDF

• Want to streamline further? Connect directly to DocuSign or PandaDoc

This will depend on your human workflow once the basic draft is ready.