How To Send Cold Emails With AI

AI can make cold outreach better. This isn’t a debate anymore. And if you’re using Hunter, then you’re already leveraging AI which we’ve implemented across the platform.

Through our research, we found that 67% of decision makers don’t care if you use AI to write your cold emails. 

At the same time, there’s a problem with how AI is used in cold email.

The emails that I tend to get nowadays:

• Are packed with AI buzzwords, 
• Use irrelevant AI-powered personalization.

AI has its place in your cold email toolset, but we shouldn’t use it for automation only. It can and should serve to make your messaging more relevant in a scalable way.

In this article, I’ll share specific use cases AI has in the context of cold email, and how to prompt it to give you what you need, from copywriting to researching your ICP and evaluating your sequence.

Write cold email copy with AI

We’ve all seen AI slop, but it doesn’t change the fact that AI is great at writing. It’s fast, the quality of its output is directly proportional to the effort you put into your prompts, and with sufficient prompting, its writing can truly impress.

There are two problems, though, when you try writing cold emails with AI:

1. AI doesn’t know what you know.

Whether you realize it or not, you possess a trove of tacit knowledge about your product/service and your prospects, and you tap into it whenever writing your messaging. 

AI doesn’t even know the trove exists. It “knows” a lot thanks to its training, but it doesn’t have an understanding of your business. And unless you provide it with that crucial context, its output will be mediocre.

2. AI isn’t good at doing two things at the same time. 

There’s a reason why many cold email tools offer a similar workflow: you add recipients, then create copy and personalize.

It’s because these steps naturally flow from each other. You first need to know who you’re contacting, and only then can you create relevant messaging.

When you ask AI to do everything at once, you disrupt that flow, and email relevance gets lost in translation.

You can address both problems by providing contextual information and by breaking down the copywriting process into several steps. Here’s how.

Document key context

Just like you, AI needs to understand who your recipients are, what they care about, and how you can help them. 

Document these in great detail:

• Who your ideal customer is
• What problems you’re addressing
• What the solution you’re offering is
• Your unique value proposition

Use a step-by-step prompting workflow 

We have tried dozens of approaches to prompting for cold-email copywriting, and we’ve tested many existing solutions. We’ve realized that bad results come from the fact that you don’t provide enough information, and that you’re trying to do everything at once. An ideal approach that we implemented in our AI writing assistant is to break down the process into two main steps:

1. You give AI your broad ICP information, and as much context about the problem/solution as possible, in order to create an email template that is not personalized but highly relevant for everyone in the audience.
2. You use AI to personalize every email, using the available data points about every recipient to reshape the template created at the previous step.

Here’s a prompt that resembles what Hunter’s AI Writing Assistant does:

Here's the output:

In the second step, personalize the template for the individual recipient with this prompt:

You are a world-class BDR, and you're doing your best work. Create a personalized cold email by contextualizing the template I provide to make it maximally relevant to the specific recipient.

Here’s the contextualized output:

When it comes to choosing a specific model to write your cold-email copy, I recommend choosing advanced models with reasoning modules like OpenAI’s o3. Especially if you just want a template to later personalize for every recipient, you want the best model available, because you won’t be spending tons of credits on this anyway.

Otherwise, you can give Hunter’s AI Writing Assistant a try. We built it to get your email sequence 80 % there, as long as you provide it with solid input.

Figure out what your prospects want

To be relevant, you need to know what your recipients care about and how your product or service helps. 

AI can help you:

• Challenge your assumptions about your ideal customer’s pains and gains.
• Narrow down your value props for subsegments of your market.

Here’s a prompt that I absolutely love: 

Here's the output:

When you segment your audience to create multiple smaller campaigns, you can do this for a list of personas to understand how your messaging can differ.

This uses AI's built-in knowledge about the given job titles, and the more details you can provide, the better the output will be. 

Taking it to the next level, you can supplement what AI already knows with your own data: blog articles, social posts, or even sales call transcripts.

Figure out the best person to contact

If you have multiple contacts at a single company, it’s often a good idea to contact several of them… But not at the same time. Start by selecting the single best person to contact.

Sometimes, it’s a no-brainer. However, there are also times when choosing the recipient can be tricky, either because job titles are ambiguous or simply because of scale.

Furthermore, you don’t always know the job titles, and sometimes you’re dealing with generic email addresses like office@example.com. 

AI can help. If it’s a single company that you’re struggling to figure out, use this prompt: 

Here's the output:

At scale—for example, if you used the Bulk Domain Search and you have a ton of companies and profiles to go through—you can paste the CSV into the AI tool you’re using and ask it to produce a table with a single recipient chosen per company. This can be imported directly into Hunter to easily contact these recipients with a campaign.

Create synthetic recipients and ask AI for their reactions to the emails you send.

This is my favorite use case. 

A major pain point with cold email is that you never know how your recipients will react to what you send them, especially when you start sending a new campaign.

AI can help you predict the various reactions your messaging may generate and adapt your sequence before you burn too many leads. 

Here’s a prompt:

And here's the output from GPT o3.

Yes, you could do this on your own. But let’s be honest: AI has near-limitless creativity, and it doesn’t have your biases, which means the personas it creates are bound to uncover something you didn’t consider about your campaign.

It’s astonishing how much this approach can teach you about your email draft. By giving different perspectives a name and backstory, AI helps you understand how to preemptively address concerns and improve your messaging.

Create follow-ups that add value

For most cold email campaigns, following up is a surefire way to get more replies.

But writing efficient follow-ups isn’t always easy. Most people aim to make the first email in the sequence packed with value, leaving seemingly nothing to add later down the line. That’s one of the reasons why many follow-ups are nothing more than a boring “Hi, have you seen my previous email?”

AI can draft entire email sequences to give you ideas for how to reengage your recipients with consecutive messages.

The trick to getting good output is to force it to write follow-ups that offer alternative CTAs and value props.

Here’s an example prompt:

This is my cold email: {paste your initial email}

I want to send this to {paste your recipient information here}

The output is really good, especially for a first step:

Figure out a goal for your campaign

Before you let AI loose on copywriting, decide exactly what the sequence should accomplish. Do you want to book demos, validate interest, or route prospects to a piece of content? Defining a single, measurable objective shapes your call-to-action so it’s clear and friction-free.

With some campaigns, there are multiple CTAs you can try, and figuring out the best one takes time. 

Using AI, you can create quick previews of the same email with varying CTAs, and anticipate how your recipients will react.

Here’s a prompt:

And here's the output with various CTAs to test.

Change the tone of voice

Tone of voice represents subtle, yet significant, changes in your messaging. Changing the tone of voice doesn’t change anything about your offer, but it impacts how you come across.

Even a perfect message can flop if it “sounds” wrong. AI makes it painless to create voice variants that are wildly different from your typical voice.

And here's the output:

Turn an email into a LinkedIn DM/voice message script

Cold outreach doesn’t live in the inbox alone. Repurposing your email for social or voice keeps messaging consistent while meeting prospects where they already spend time.

1. LinkedIn DM – tighter character limit, faster payoff.
2. Voice note – conversational, human, perfect for objections and micro-stories.

You’ll get two modality-ready assets in seconds—no need to reinvent the wheel.

Ask where to find data

Personalization lives or dies on data. While Hunter has tons of datapoints on companies and professionals, we won’t ever satisfy every single data need you might have. 

When you’re not sure what to enrich or where to source it, let AI brainstorm for you.

With AI's knowledge of the various data sourcing methods, you're bound to get inspired by the output.

From personal experience, I have to say that AI can be too optimistic about easily acquiring data points like G2 or Capterra reviews: in practice, they are really hard to scrape. Still, this can give you a starting point, and with some research, you can find methods of getting data from just about any location online.

Wrapping up

I used to be sceptical about AI when it first saw it impact my inbox, and I was wrong. The question isn’t if, but how we should use LLMs to make the job of connecting with our prospects easier.

When writing this article, my main goal was to inspire you to use AI for your outreach campaigns beyond simple personalization tricks. 

If you found any of this interesting, make sure you give Hunter’s AI Writing Assistant a try—we’re building it precisely to help you be more relevant.