I watched a cybersecurity SaaS company gut their content strategy in real time last year. Smart people, solid product, Series B funded. They'd spent eighteen months building an audience through sharp analysis of threat landscapes and incident breakdowns. Then someone in finance ran the numbers and decided three staff writers cost too much. By late 2023, they'd replaced their editorial team with ChatGPT and a junior coordinator making $50K. Within two weeks, engagement dropped 62%. Bounce rate climbed. Returning visitors—the ones who'd actually bought into their expertise—vanished. Three months later, they quietly started rehiring. I keep seeing versions of this story, and I'm tired of watching companies learn expensive lessons they could have avoided by asking one basic question: What are we actually paying for when we pay for writing? The Commodity Trap The logic seemed airtight when generative AI hit scale in late 2022. Writing is words on a page. AI produces words faster and cheaper than humans. Therefore, AI should replace humans for writing. Silicon Valley's favorite syllogism: clean, simple, completely wrong. What died in that equation was the distinction between production and synthesis. Yes, AI produces text at speeds no human can match. But companies didn't need more text—they were already drowning in it. What they needed was judgment about which arguments mattered, which framing would land with specific audiences, and which claims could survive scrutiny. production synthesis The AI couldn't provide that because it fundamentally can't. Not in the "it's not sophisticated enough yet" sense, but in the architectural sense. Large language models predict plausible next tokens based on statistical patterns. They don't evaluate whether an argument is sound, whether a metaphor clarifies or obscures, or whether a claim will hold up when a customer's CISO reads it at 2 AM while evaluating vendors. That distinction didn't matter when the metric was "publish three blog posts per week." It mattered enormously when the metric was "generate qualified leads" or "reduce support tickets through better documentation." What the Market Actually Values I've been covering tech for long enough to recognize when received wisdom is collapsing under its own weight. Right now, the "AI will replace writers" narrative is hitting reality and losing badly. Here's what I'm seeing in my own network: editors at major tech publications are rejecting more submissions than ever, despite (or because of) AI making it easier to produce them. One editor told me bluntly: "This looks clean, but it doesn't say anything. We're looking for reporting, not summaries." That's the unbundling in action. AI made the production of clean-looking text essentially free. In doing so, it made that skill worthless and made everything else—research, synthesis, judgment, voice—dramatically more valuable. The job postings reflect this. I checked listings for senior content roles last month: $120K to $180K for positions that specify "strategic thinking," "subject matter expertise," and "distinctive voice." These aren't legacy media companies clinging to tradition. They're venture-backed startups, enterprise SaaS vendors, and cybersecurity firms that just watched their competitors publish themselves into irrelevance with AI slop. Meanwhile, the companies that went all-in on AI content are quietly walking it back. A fintech company I follow pulled down an entire batch of AI-generated articles after readers flagged factual inconsistencies that should have been caught in editorial review. The reputational cost dwarfed whatever they saved on writing. The Clarity Crisis What surprises me most is how few companies saw this coming. The warning signs were obvious if you were paying attention. Take technical documentation—arguably the clearest test case for "can AI handle this?" The answer turned out to be "sort of, but not in ways that matter." AI can generate grammatically correct documentation. It cannot generate documentation that anticipates where users will get confused, or that structures information in the sequence people actually need it, or that knows which edge cases are common enough to warrant explanation. I've talked to enough product managers to know: the cost of bad documentation isn't the documentation budget. It's the support tickets, the extended sales cycles, the features users can't find, and the churn from customers who never understood what they bought. One enterprise security vendor I covered spent six months trying to AI-generate their knowledge base. Everything parsed correctly. Usage metrics were disastrous because the articles answered the questions users asked, not the questions they actually had. They ended up hiring two senior technical writers who understood the product deeply enough to bridge that gap. The payback period was six weeks. What AI Actually Revealed Here's the uncomfortable truth that's emerging: AI didn't reveal that writing was easy to automate. It revealed that a lot of what we called "writing" was already automated thinking—just slower and more expensive. The blog posts that disappeared when companies switched to AI? They were never providing unique value. They were executing a content marketing playbook: identify keyword, hit word count, include relevant terms, publish. An AI can do that perfectly because there's no actual thinking required. What AI can't do—what's now commanding serious premiums—is the harder stuff that we somehow convinced ourselves was optional: Taking a technical product and explaining it in terms a non-technical buyer can evaluate
Synthesizing six conflicting sources into a coherent position with explicit tradeoffs
Building an argument that withstands informed skepticism
Developing a voice that readers recognize and trust Taking a technical product and explaining it in terms a non-technical buyer can evaluate Synthesizing six conflicting sources into a coherent position with explicit tradeoffs Building an argument that withstands informed skepticism Developing a voice that readers recognize and trust These aren't decorative skills. They're the entire job, once you strip away the text production that AI now handles. The Economics of Synthesis The market is already sorting this out, though not everyone's figured it out yet. Companies are discovering that content marketing ROI depends entirely on whether the content does anything besides exist. The math isn't subtle. Content that generates leads costs about 62% less per lead than traditional advertising while delivering three times the volume. But that's only true if the content actually works—if it builds credibility, clarifies value, and moves prospects toward decisions. Generic AI content doesn't do any of that. It ranks poorly (Google's gotten better at identifying it), converts worse (readers can smell it), and actively damages brand perception when it's obviously synthetic. So companies face a choice: pay for speed and get content that might as well not exist, or pay for judgment and get content that actually performs. When you frame it that way, the economics flip entirely. The question isn't "can we afford writers?" It's "can we afford not to have them?" What Happens Next I don't think we're heading toward AI replacing writers any more than Excel replaced accountants. What's happening is more precise: AI is handling the mechanical parts of writing, which makes the non-mechanical parts—the thinking, the synthesis, the judgment—vastly more important. The companies that understand this are already pulling ahead. They're using AI as a drafting tool while keeping humans in the loop for everything that matters: argument structure, factual verification, audience calibration, strategic positioning. The ones that don't understand it are still trying to optimize their way to relevance, publishing more content that performs worse while wondering why their competitors are eating their lunch. This is where I get opinionated: the market will force clarity here whether companies like it or not. Audiences are developing allergic reactions to AI-generated mediocrity. The brands that win will be the ones that deliver actual insight, actual expertise, actual value. That requires writers who can think, not just type. And since AI made typing essentially free, that's all that's left worth paying for. The great unbundling isn't coming. It's here. The question is whether you're on the right side of it.

Google

The Great Unbundling: How AI Separated Writing Speed from Writing Skill

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

AI in the Classroom Promises Efficiency—But at What Cost to Learning?

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

Darwin's Hybrid Intelligence to Align AI & Human Goals for Startups & VCs

The Noonification: White Man (11/26/2022)

The Noonification: The Metaverse is a Sh*tshow (11/2/2022)

100 Days of AI Day 1: From Newsletter to Podcast, Leveraging AI for Audio Transformation

AI in the Classroom Promises Efficiency—But at What Cost to Learning?

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

Darwin's Hybrid Intelligence to Align AI & Human Goals for Startups & VCs

The Noonification: White Man (11/26/2022)

The Noonification: The Metaverse is a Sh*tshow (11/2/2022)

100 Days of AI Day 1: From Newsletter to Podcast, Leveraging AI for Audio Transformation

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps