From pre-seed to Series G, Peter Purcell spent 12 years in B2B SaaS marketing before noticing something hiding in plain sight. Founder-led content on LinkedIn was quietly outperforming brand content by ~10x. Most founders knew it, yet almost none of them had time to do anything about it.
So Peter started a ghostwriting agency to bridge the gap, realized the agency model itself was the bottleneck, and built Poplar — a platform that creates a living AI voice profile for executives and turns it into a content engine that actually sounds like them.
Here are three things I took away from our conversation.
The Knowledge Base Is the Product (Not the AI Writing)
AI writing isn't that interesting. By now, we know Claude or ChatGPT can write a LinkedIn post. That's not the point.
Poplar is building something closer to an AI digital twin — a voice profile that crawls your LinkedIn history, your website, your meeting notetakers, your Notion docs, your knowledge bases, and keeps learning from all of it over time. The more data it has, the more it sounds like you and only you. Because it's also monitoring post performance and learning from what actually lands with your specific audience, it compounds. The writing gets better and the timing gets smarter. You get an AI social media strategist meets ghost writer.
We Can All Spot AI Slop By Now
The (beloved) em dash everywhere, the "it's not X, it's Y" framing, the aggressively chopped-up paragraphs with never more than two sentences in a row. You know exactly what Peter's talking about, because we've all seen it.
The irony is that those patterns exist for a reason: Shorter sentences, broken-up text, and punchy formatting all genuinely perform better in the LinkedIn algorithm. (Trust me, I'm in PR). Every AI model learned to write that way, so now the thing that "performs well" is also the same thing that screams "a human did not write this."
Poplar's response is to make subject matter primary and formatting secondary. If the content is pulling from genuinely specific, individual knowledge (your take, your experience, your domain) it's harder to replicate than anything a generic prompt can produce. Proprietary perspective becomes the slop antidote.
AI Overwhelming Us Might Be Underhyped
I asked Peter about the "AI will replace entire job functions" narrative, and he thinks it's overhyped, at least in the way and timeline most people are imagining. Human strategy, context, and experience are still driving business, and that won't change as fast as the headlines suggest or some of us worry about.
What he thinks isn't talked about enough, however, is AI going from highly convenient to quietly oppressive before we notice. For example, cybersecurity developers can push 10x the code they could two years ago, but application security teams can't scale at the same rate. So now you need AI to secure the AI-generated code. The same dynamic is showing up everywhere else: More capacity in one place creates a new bottleneck somewhere else.
Peter's take isn't exactly pessimistic, but more of a heads-up. The tools that "win" long-term will be the ones that scale intelligently, with full context, across an entire organization all at once. That's the version of augmentation he finds interesting, and not coincidentally, what he's building.
Listen to the full episode of Actually Intelligent to hear more from Peter Purcell on how he went from agency to startup founder, why he thinks the best AI tools are the ones solving for the human bottleneck, and what happens when Poplar customers see their first generated post.
LEAVE A COMMENT