What a gem

My favorite LLM right now, and how I'm using it

Hey everyone,

Welcome back to the Markfive newsletter, where we talk about AI strategies that I’m seeing from the frontlines.

Last week I mentioned I was building an automated cold email system to help a founder get more leads and keep his sales pipeline full.

Today I want to share part of the process I’ve been using that you can implement today using my favorite language model app — Google’s Gemini 2.5 Flash.

Here’s how I’m using it to help run cold email at scale:

Good cold email has five primary elements. Short-change any one of these, and your campaign won’t work as well as it should.

  • How compelling your offer actually is

  • The quality of your list of targets

  • Personalization of the cold email script (this is the one we’ll be diving into)

  • Setting up your infrastructure properly

  • Responding to interested leads right away and nurturing them

When you’re signed into Gemini, there’s a feature called “Gems” which essentially lets you build a custom ChatGPT that’s expertly trained on a specific amount of content that you give it (at absolutely zero cost).

Here’s an official guide on how to do it by Google if you want to follow along.

In my Gem, I uploaded an advanced report on my client’s ideal customer profile, an in-depth description of my client’s offer (given to me by my client), as well as a bunch of internal SOP’s I use to position offers in an ideal way for cold prospects (if you’d like a version of this for yourself, just reply to this email and I’ll see if I can make one for you).

Then I gave it a system prompt/underlying instruction to act like an expertly trained marketer on the topic of demand generation offer positioning for cold prospects.

Finally, I asked my Gem to develop a few scripts to use in our cold email campaign, based exclusively on the prior knowledge I uploaded to it earlier.

Tip: If at any point, you’re struggling to prompt AI, just ask AI what you should ask and what your prompt should be!

The results were pretty damn good. Like any AI output, I had to whittle it down to the refined production-ready level, but not having to come up with these personalized scripts at scale from zero is a huge timesaver. Here’s a semi- redacted version of the scripts we’re moving forward with.

Subject line: {their dream outcome}

Hey {their name},

Seeing many {their_role} like yourself at {company name} are overwhelmed with {their biggest problem we can solve}.

Would you be open to a 15-minute call where we can show you how to {the fix to their industry-specific problem} with {my client's product and their unique mechanism}. 

No commitment, just a practical demo of {my client's value prop}.

Cheers,
{Name}

PS - {personalization blurb}

This is what we’re using to validate the offer for cold prospects on our lead list, if it goes well we’ll be doubling down and using Clay to do this at a higher volume - but that’s a topic for a future issue.

That’s all I got for today.

Talk soon,

Momin

PS - If you want help building an automated cold email system that actually drives qualified leads to your business, book a call with me here. We'll craft a highly compelling version of your offer for cold traffic, identify your best target segments, and map out the tech stack you need to keep your sales pipeline full.