Latest / The Newsletter Podcast with Fexingo: Substack, Beehiiv, and Email Newsletter Business / How Newsletters Use AI Writing Assistants to Save 10 Hours a Week
Transcript
- Lucas: So there's this stat I keep coming back to — a survey from last fall of about five hundred newsletter operators, and the ones using AI writing assistants reported saving an average of ten hours a week. Not five, not eight — ten. Luna: Ten hours is basically a full workday. That's huge. But I feel like there's this unspoken anxiety about it — like using AI makes you less of a 'real writer'. Lucas: Exactly. And that tension is real. I think a lot of newsletter creators worry that if they use ChatGPT or Claude to draft a post, readers will somehow smell it. The prose will feel generic, the voice will flatten. Luna: Right — like the uncanny valley of email. It's technically correct but it doesn't feel like a person wrote it. Lucas: Yeah. And some readers actually can tell. There was a Twitter thread from a Substack author who ran an experiment — she published two posts, one fully human-written, one ai assisted. Her readers flagged the AI one in the comments within an hour. But here's the nuance — she didn't have a human-in-the-loop workflow. She basically hit 'generate' and published. Luna: Oh, that's the mistake. The successful newsletter operators I've talked to use AI as a co-pilot, not a pilot. Outlines, subject line variations, maybe a first draft they heavily rewrite. Lucas: Right. So the question is — what does a good human-in-the-loop workflow actually look like? Let's walk through a real example. Say you're writing a weekly newsletter about the semiconductor industry. You've got your research. Maybe a few links, some notes. Instead of staring at a blank screen, you feed your raw material into Claude and ask for an outline in your voice. Luna: And you'd have to train it first, right? Give it samples of your past newsletters so it understands tone, sentence length, how you open. Lucas: Exactly. That's step one. Then you get an outline. Maybe three sections, some bullet points. You rearrange, cut, add your own angle. Then you ask it to write each section, but you edit heavily. The AI might give you a decent paragraph about ASML's latest lithography machine, but you know the context — you worked in that industry for five years. So you rewrite half of it. Luna: So the AI saves you the mechanical part — the first draft — but you still do the thinking, the voice, the nuance. That's where the ten hours goes. Lucas: Precisely. And the best part is, you can also use AI for the stuff you hate doing. Like subject lines. A lot of writers dread writing subject lines. You can ask for twenty variations, pick two or three that feel right, maybe tweak one, and you're done. That alone can save thirty minutes an issue. Luna: And subject lines are so data-driven now. Some of the Beehiiv power users I follow A-B test subject lines and the ai generated ones often win — because they're less precious, more direct. Lucas: Right. The AI isn't attached to a clever pun. It just writes 'Why Intel's new chip matters for your portfolio' — which might outperform the writer's more creative version. Luna: But there's a flip side. I worry about homogenization. If everyone uses the same AI models, won't all newsletters start to sound the same? Lucas: That's a real risk — especially if people don't do the editing step. The AI defaults to a kind of corporate-friendly, mid-Atlantic English. No edge, no personality. But the writers who succeed are the ones who use AI to amplify their voice, not replace it. Luna: So it's like having a really efficient research assistant who can also draft — but you're still the editor-in-chief. Lucas: Exactly. And I think that's the right mental model. One of the creators I follow, a guy who writes a popular marketing newsletter on Substack, told me he uses AI for about forty percent of his writing process now. He drafts the intro and conclusion himself, uses AI for the middle section — the data-heavy part — then rewrites the whole thing. He says his output went from one issue a week to three, and his open rates actually went up. Luna: Because the content got more valuable, more frequent. Readers felt like they were getting more from him. Lucas: And his voice was still there because he was curating and editing. The AI just made him faster. On the other end of the spectrum, I talked to a finance newsletter writer who tried AI and hated it. Said every draft sounded like a McKinsey report. He went back to writing everything from scratch. Luna: Which is totally valid. AI isn't for everyone. But I think the most interesting use case is personalization at scale. Imagine you have ten thousand subscribers. You can use AI to write three different versions of the same newsletter — one for beginners, one for intermediate, one for advanced — based on what they've clicked before. Lucas: That's where Beehiiv's new AI features are heading. They rolled out a tool that lets you create dynamic content blocks. So the same email can show different paragraphs to different segments. The AI writes the variations, but you set the rules. Luna: That's powerful. But it also raises the stakes on the editing. One mistake and a beginner gets a paragraph about options strategies and unsubscribes. Lucas: Right. So you need to test. Most creators start with a small test segment, like five percent of their list, before rolling out to everyone. And they monitor the open and click rates closely. If something feels off, they tweak the prompt. Luna: And that's the thing — the prompt is the skill now. Writing a good prompt is almost as important as writing the newsletter itself. Lucas: Absolutely. I've seen people share prompts that are longer than some of their newsletters. They include tone guidelines, formatting rules, examples of what they like and don't like. It's essentially a style guide for the AI. Luna: And you can reuse that prompt for every issue. That's where the time savings compound. Lucas: Exactly. You invest an hour upfront to build a great prompt, and then every week you save two hours because you don't have to reinvent the workflow. And if these marketing conversations have sparked something you've actually used in your own newsletter, I'll just say — a couple of dollars a month is genuinely what keeps these episodes coming. You can find us at buy me a coffee dot com slash fexingo. No pressure, but it makes a real difference. Luna: Yeah, and we really appreciate it. It keeps the podcast ad-free and lets us spend time on deep dives like this. Lucas: Alright, back to the AI workflow. Another tool that's gaining traction is using AI for research summaries. Instead of reading ten articles, you can paste them into Claude and ask for a one-paragraph synthesis with key stats. That alone can save an hour per issue. Luna: But you have to fact-check. AI still hallucinates, especially with numbers. I've seen it make up a statistic that sounded totally plausible but was completely wrong. Lucas: Critical point. Never trust AI with a number you haven't verified. Use it for structure, for phrasing, for ideas — but if it cites a study, go find the original source. That's part of the human-in-the-loop model. Luna: So the bottom line is — AI can save you ten hours a week, but only if you're willing to do the editing and fact-checking. It's not a set-it-and-forget-it solution. Lucas: Exactly. And I think that's the right framing. The writers who thrive with AI are the ones who treat it as a tool to amplify their own skills, not a crutch. They still bring the voice, the insight, the judgment. The AI just handles the grunt work. Luna: And the results speak for themselves. Higher output, better subject lines, more personalization — all without burning out. That's a pretty compelling case. Lucas: Yeah. I think in a year or two, using AI will be as normal as using spellcheck. It'll just be part of the writer's toolkit. The key is learning how to use it well now, before everyone catches up. Luna: And keeping your voice. That's the thing readers come back for. Lucas: Exactly. So if you're thinking about trying an AI assistant, start small. Pick one part of your workflow — subject lines, or outlines — and experiment. See how it feels. You might find those ten hours.