Using AI for Cold Email Feedback

A little show and tell today.

You’re likely sending cold emails / cold LinkedIn messages to get customers for interviews, CITs, and MVPs. AI tools can be helpful for this if you use them right. I particularly like Claude, which I think writes more like a human than ChatGPT.

All AI tools, at this point, are best used for inspiration and perspiration — things like generating example cold emails you can riff on or creating summaries of 10,000 word research papers or a 90-minute call.

Today, we’ll talk cold messages.

Let’s say I want to send cold emails to plumbers because I’m running a CIT for software that’ll help them better schedule appointments.

Maybe I did a bunch of customer interviews and learned that plumbers don’t have a thoughtful system around inbound business. They’re extremely busy, field a lot of calls, and basically respond to them in the order they received them. Maybe I think there’s massive opportunity here. A tool could help figure out the highest value potential customers, take into account urgency, prioritize previous customers, optimize routes, etc.

But now I need to test the big assumption — that plumbers care enough about this problem to change. Will they actually adopt new software to solve this problem?

So, I need cold emails.

And that’s where Claude (and Tacklebox) comes in.

Our best cold emails for this situation have four characteristics. They…

1. Lean into Problem language

2. Highlight the risk the customer feels immediately

3. Describe “Wild Success”

4. Pitch a wedge problem as the CTA

So, I want Claude to help me write with these in mind.

I open it up and begin to describe what I’ve got and what I want.

I write:

“I'm building a tool to help plumbers manage their inbound customer requests. I’d like to get some first customers, so I’m writing a cold email to try to attract them.

I’d like for this email to have a few key characteristics:

1. I’d like to use “problem language” for the email subject and the first line. Meaning, I want to describe a problem they have to get their attention. For example — I’m a 6’5 guy with long arms so most shirts don’t fit. A clothing company using “problem language” might write — “No shirts fit because your arms are too long” to get my attention. That’s problem language.

2. I’d like to then hit on a risk — the first risk that pops into their mind about using the service should be immediately assuaged.

3. I’d like to show what wild success would look like.

4. I’d like to end with a call to action that seems manageable and easy to respond to and solves an immediate problem — a wedge, a foot in the door.

Here’s some information on the business and the prompts above:

I did a bunch of customer interviews with plumbers and realized that they struggle to manage inbound business. They get a lot of calls and it’s tough ranking them by importance — how much potential revenue is there, is this a repeat customer, is this a new customer — and then scheduling / responding accordingly.

I’m planning to build a tool to help with this — I’ll use AI to manage intake, assess opportunity, then create a schedule based on NPV and happy repeat business. The AI will text with customers to get the information needed, then schedule.

1. The problem - it’s hard for plumbers to manage inbound in a way that maximizes revenue

2. The immediate risk - A new system is hard to learn and messes up their current system.

3. Wild success - Prioritizing the most profitable customers and de-prioritizing the least profitable

4. Call to action / Wedge - Give them a customized link they can send to one inbound customer to test — it’ll give them an estimate of lifetime value and a suggested “ranking” of the customer

Please create a cold email template that’s no longer than six sentences in the form from above:”

I pop that in, and here’s what I get:

This is…fine. As I said, it’s inspiration.

Now comes the editing - you treat this like a conversation and keep prompting. Things like “speak more casually” or “do it in only four sentences” or “do it as if you’re a business made by plumbers” work pretty well.

What I’ve found works really well is to paste in a whole example of a cold email — or 10 — that you really like.

Eventually, with some human tweaks, you end up with something like:

Subject:

You’re an overworked plumber losing revenue because inbound is out of control

Body:

You're missing out on serious revenue by prioritizing less profitable inbound customers over higher-value opportunities. This is a tricky problem. You’ve got to prioritize existing customers, assess new customers with incomplete information, protect your Google ranking, and figure out your schedule accordingly.

We’ve built a tool that’ll integrate with your current intake system to help you understand the quality of inbound customers. There’s zero extra effort on your end. We then build you a smart schedule based on profitability, customer quality and route (minimize your driving time).

You don’t have time to test this new system on your own. Text this number (number) and we’ll show you how it looks for a customer, then give you a link you can test with a new customer on your end.

Shoot a text to (number) if interested or if you’ve got any other questions.

And there you go. A pretty good customer cold email that’ll hopefully convert. With the help of Claude.