Another round of advice from friends to our community, served fresh!

This time around, we’re navigating somewhat “murky” waters in the email space…

AI in your email process or flows.

This is a tricky topic because there’s so much noise out there, so we thought, who better to ask to share their experience than two people who incorporate AI into their strategy in a massive way (and with a heap of success).

Cagri Sarigoz is the Co-founder and CEO of HeyNews, an AI newsletter software company helping publishers produce more, earn more, and stress less about the weekly grind.

Alongside him is Eren Daşkesen, Co-founder and Chief Creative Officer, and the person responsible for making sure everything HeyNews produces actually sounds like a human wrote it.

Together, over the past 12 months alone, their team has produced more than 550 newsletter issues across 10 different formats for clients around the world.

(you can see why we’ve asked for their thoughts)

They’re here today to help our community cut through some of that noise, share how they've made it work without sacrificing quality and leave you with something that’s genuinely useful.

Let’s get stuck in!

Time to save you the debate.

Every week, we see the same argument play out in newsletter communities. Half the room is convinced AI will destroy the authenticity that makes newsletters worth reading.

The other half is using ChatGPT to pump out issues in 5 minutes and calling it progress.

Sigh…Unfortunately, they're both wrong.

The question isn't whether to use AI. It's whether you're using it in a way that actually serves your readers, or just makes your life feel easier for a few weeks before your open rates start plummeting.

We have been in the newsletter business long enough to have watched both play out.

And, we also agree that there’s a problem with most AI-for-email content…

And it’s something most of us are seeing in a fair number of emails we’re consuming lately.

Much of the AI-generated newsletter content out there is pretty poor.

Not "slightly generic" kinda bad.
We’re talking about that “LOL, ChatGPT totally wrote this" kinda bad.

And, we can only blame the generic nonsense that LLMs spew to a point; the real issue here stems from the poor processes these senders use.

When people use AI as their “saviour-of-a-ghostwriter”, they often wind up with a newsletter that sounds like, well…a generic newsletter.

It misses the nuance of what makes the writer exactly that: the writer.

Whatever generative AI they’re using has no idea why they started their newsletter in the first place, what their readers actually care about, or what makes their perspective worth anything.

And the result (as you’ll most often see) is that content may technically be correct, but it’s only vaguely interesting and, honestly, completely forgettable 🙃

And forgettable newsletters lose subscribers (or a place in the inbox for ongoing attention). This is the danger of slop.

How do we (honestly) use AI at HeyNews? (literally, what we’ve refined over 550+ emails generated in the last year):

To begin, we’re not replacing our writing, but replacing the work writers don’t need to be doing manually!

If we’re specific about it, we’re:

These are the parts of newsletter production that drain our energy without adding to our creativity.

And it’s these repetitive, time-consuming tasks that are exactly what AI is built for.

Creative judgment, the angle selection, the editorial voice that makes a newsletter feel like a real person? That’s where that grey matter between our two ears is being applied.

And that makes all the difference.

Skynet won’t win for a while 🤖

When we tell folks that we've produced 550+ issues with a small team across 10 different formats, the first question is always "how?"

And our answer always goes back to what we were just discussing: We’re using AI for the grunt work, and our editors for everything that matters.

Every issue goes through human editing before it ships.

Not a quick skim (though we’re structuring for all you readers skimming things, like this email).

We’re talking real editing. The honest response to the “does this TRULY sound like us” question is where the focus lies.

This question is what everyone using AI or not needs to answer. Every. Damn. Time.
This separates trust from mere tolerance in our engagement.

And we have seen what happens when you skip this step.

On the surface, everything looks okay, doesn’t it? You don’t see shifts in your email metrics initially either…but things feel “off” and your readers sense this.

That’s when replies drop off a cliff. Unsubscribes start ticking up. Sales start diving…
And the culprit is care. If you don’t care anymore, then frankly, why should your readers?

That human layer is intangible yet fragile. And humans are pattern-seeking creatures, so they may not realise immediately, but it becomes palpable.

Okay, the “steal this framework” heading is cliched, because we’re giving it to you anyway, but this is for all you kleptomaniacs seeking a thrill 😉

(And, seriously, it’s not rocket science, but it works for us and we’re pretty certain it’s going to help you out, too).

If you want to start using AI well, without turning your newsletter into unseasoned robot chowder, here's where to begin:

1. Don’t use AI for storytelling; use it for sourcing. Give it the angle and let it outline the story. Let it help you surface the context within the confines of who your audience is and what they care about.

You then decide what matters and why. The “why” people should care can’t be told to you, at the end of the day.

2. Build from your own archive. The best AI outputs come when the model has read and understood your past work. Feed your previous emails, email replies, post content, heck, even your website or call transcripts if you have to and build out a markdown file that’s practically a clone of YOUR way of writing.

This cuts down prompting time or telling it to “sound like an [x] professional with 10 years of experience (which never works anyway).

3. Never skip the edit. Like ever! Even a five-minute read-through where you ask yourself questions like “Would I actually say this?" is enough to catch the moments where your chosen LLM has drifted or imagined what you would sound like.

This is, honestly, where the majority of our time is spent in creation, because it’s the stuff that matters.

4. Measure what your readers actually do. If your AI-assisted issues are getting fewer replies, fewer clicks and less engagement, those efficiency gains simply aren’t worth it. This means you need to set a schedule to review what your metrics are doing.

Quality is quite literally a business metric (besides being an ethical one, and you owe it to your readers). You just can’t let it slide, guys. This is where you lose trust, engagement, readers and money.

It’s easier than you think, but we’re not (yet) at the point that we can fully trust the robots to actually BE us, so make sure you put in the work.

We’ve been building our system internally over the past year already, and we have been using it exclusively with our clients until now.

Essentially, we’ve put together a system that learns your tone and style from everything you’ve already written, generates drafts that sound like you and handles sourcing by pulling relevant content from your archive (all while suggesting new stories from our curated pool).

So, if you’re strapped for time in your email process and need some help, feel free to reach out, because we’re keen to give back to this community we’re part of.

We’re letting 10 readers who get in touch use this for a month (on the house), simply for giving this a read.

If you’re keen, hit the button below, say hi, and we’ll be in touch to share that. (And, this also won’t be a pitch. We’ll share the tool, and that’s it.)

P.S. if you’re keen to find more about how we think about newsletters, AI, and building publications worth reading, head on over to check out our newsletter here!

Your friends, Cagri & Eren.

AI agents already outnumber human visitors to your docs — now you can track them.

See your AI traffic!

Ready to use AI in your emails the right way, {{name|Friend}}?

If you have any feedback or knowledge to share, click here! Oh, and please share this email with your friends and colleagues if you think they’ll find value over here.

Your feedback only makes us better.

Your friend in email,

Des