AI Won’t Save Your Business

Hey honeys and hustlers,

beehiiv recently introduced a new feature called Copilot, an internal AI tool that can read your newsletter data and automate tasks within your publication. I wasn’t expecting this to be announced during their summer release, but I was very eager to try it out once I heard about it. I didn’t expect a newsletter platform to start sounding like my second brain, also known as Notion AI, where I spend a great deal of time. And yet, here I am, depleting my weekly AI token allowance in one day asking beehiiv Copilot things like, “Can you please explain why my open rates are dropping while my click-through rate is increasing?”

Because that’s the beauty of AI inside a closed system: it’s not trying to be everything to everyone. It’s trying to be useful to you, specifically, in the tiny little universe where all your work already lives. Copilot isn’t a third-party chatbot you have to manually feed context. It already knows what you’ve published, how it performed, what your audience actually clicks, and which subject lines you swear were funny (but the data says otherwise). So I started using Copilot the way I use Notion AI: as a high-speed thought partner with access to receipts.

First up: my welcome sequence and welcome survey. I’ve only rewritten it once—which is saying something considering I’ve been writing this newsletter almost 3 years. It’s long overdue for a change. Copilot helped me tighten the copy, clarify what I actually want to learn about new subscribers, and (most importantly) build better segmentation. Not the vague “How can I best help you?” segmentation. I mean the specific, helpful kind: what people want more of, what they’re struggling with, and what they’re secretly hoping I’ll send them at 7:02AM on a Tuesday (even though I publish on Mondays, Thursdays, and Saturdays).

Then I got greedy. I asked it to auto-tag previous articles, which felt like a huge organizational task being lifted from my shoulders (and is the reason I blew through my weekly token allowance twice). It’s one of those tasks that’s easy in theory and maddening in reality—especially when you’ve been publishing long enough to have a back catalog that could qualify as “historical archives” (it tagged 277 articles, to be exact). Watching Copilot do in minutes what would take me a weekend and a questionable amount of caffeine? Pure. Joy.

I asked it to give me a monthly analytics recap with context. Not just “here are the numbers,” but “here are the numbers and why they look this way.” I asked it why my open rates were dipping, but my CTR was climbing (I mean technically I could’ve simply asked you guys, but that defeats the purpose of using AI). Copilot helped me interpret it: my growth is outpacing my engagement, but the people who do open are more aligned. Maybe I need to do more to ensure that new subscribers feel welcomed, are directed to content that’s most interesting to them, and can meet others in the community.

And because it already has access to lifetime performance data across time, it can help me make sense of the patterns I’ve been seeing but unsure how to interpret or improve. It can connect dots between topics, formats, and subscriber behavior without me doing the spreadsheet equivalent and wanting to pull my hair out.

Not everything has been a full-blown “wow.” I asked Copilot to develop an organic growth plan, and it was… fine. Helpful-ish. Like getting generic advice from an online content creator who’s tried once but hasn’t truly done it long enough to teach: solid fundamentals, fewer specific insights. That said, even the “somewhat helpful” outputs become more valuable when you treat them like a draft you’re allowed to mold into shape.

Where Copilot really shines for me is the stuff that benefits from repetition, pattern recognition, and access to history: rewriting automation sequences, building segments, auditing lead magnet flows, and eventually generating campaigns for new products—like my article image generator—without me starting from scratch every time.

There’s just one small problem: tokens. Weekly tokens, specifically. There aren’t nearly enough. Not if you’re running even one publication with any consistency, and definitely not if you’re managing two. Copilot makes me want to use it more, which is rude when the meter runs out mid-task like, “Okay, you’ve had enough improvement for today.” I’m hoping beehiiv takes the hint and bumps the weekly allowance the way they did with daily tokens for website vibe coding. I’m not trying to become an AI maximalist—I’m just trying to move faster without burning out.

If you’re new here, you may not know this, but I’m a big fan of Notion. To me, Copilot and Notion AI feel like cousins. Not because they’re identical, but because they both become useful the same way—through proximity to your real work. Notion AI works for me because my Notion workspace has everything: analytics, progress notes, goals, quarterly focuses, production notes, content calendars, the tiny thoughts that become big decisions. Beehiiv Copilot works because beehiiv has everything related to my newsletters: every post, every performance metric, every tag, every ad, every click, every subscriber source.

Neither tool will take you from zero to hero. If you’re just getting started, AI can’t invent momentum you don’t have. But if you do have motion—if you’ve been publishing, testing, iterating, collecting data—these tools can absolutely increase your trajectory. They can help you spot what’s working, improve what’s not, and automate the kind of maintenance work that quietly drains creative energy. Basically: the magic isn’t that AI writes for you. The magic is that it helps you think with your own evidence.

So get to creating. I’m rooting for you.