The person
who writes
your story
actually writes.

Every founder, executive, and creator I work with has access to the same tools I do. They have ChatGPT, Claude, Gemini, and three more in beta they signed up for last week. And yet — they are still hiring writers. Specifically, they are still hiring writers who cost more than they did two years ago.

The reason is not what the AI companies would like you to think it is. It is not that humans “edit better” or “add the human touch” or any of the other vague hedges thrown around by the consultancies trying to sell you the prompt-engineering certification. The reason is simpler and more structural: AI is a compression engine. It does not invent. It averages. And averages, when applied to brand voice, produce exactly the thing you are trying to avoid.

Run the same prompt through the same model twice with two different companies' positioning briefs. You will get two outputs that are remarkable for how similar they are. Both will be confident. Both will be readable. Both will include the words “empower,” “seamless,” “reimagine,” and the construction “we don't just X — we Y.” Both will sound exactly like the rest of the internet. That is not a bug. That is what the model was built to do. It surveys the corpus of existing writing and produces a confident average of it, which is by definition the most generic thing it is capable of producing.

There is one thing AI cannot do, and it is the thing that actually makes copy convert: it cannot have a specific, defensible point of view. Voice is not adjective selection. Voice is the residue of someone with an opinion making decisions on the page. What they include. What they refuse to say. The specific examples they reach for. The metaphors they use that no one else would. The arguments they are willing to make that other people in their category quietly avoid because the avoidance is more comfortable. None of this can be generated. It can only be observed in a writer and captured.

This is why your “AI-assisted” thought leadership reads like everyone else's. It is not because the model is bad. It is because the model has no skin in any of the arguments it is making. It will not pick a side because it does not have a side. It will hedge every claim because hedging is the safest path through the data it was trained on. What you are getting back is the literary equivalent of an executive who agrees with whichever board member spoke last.

What AI Is Actually Good At

I want to be precise here, because the framing of “AI versus humans” is mostly false. AI is genuinely useful for a specific set of tasks in the writing process. It is excellent at first-draft scaffolding when you already know what you want to say and need to get something on the page to react to. It is excellent at light editing — finding repetition, smoothing transitions, catching the third instance of a word you have already used twice. It is excellent at format conversion: turning a long essay into a LinkedIn post, an email sequence, a tweet thread.

What AI is not good at — and what no amount of prompt engineering will make it good at — is generating the specific point of view that makes a piece of copy worth reading in the first place. The argument. The opinion. The thing your reader has not heard before, or has heard before but never heard said this clearly. That is upstream of the writing. That is the strategic work that has to happen first, and that work has nothing to do with the model.

The market is correcting toward this faster than most founders realize. The companies that built their content engines around AI in 2023 are already seeing diminishing returns. Their LinkedIn posts get the same engagement as a quiet day. Their newsletters get unsubscribes at the rate of the very generic content they are now indistinguishable from. The advantage they thought they were buying — “produce 10x more content for 1/10th the cost” — turned out to be an advantage in producing the kind of content that has no advantage.

The brands winning attention right now are the ones with a writer or an editor at the top of the funnel making specific, opinionated choices. Sometimes that writer is the founder. Sometimes that writer is a ghostwriter who has spent enough time inside the founder's head to think the way they think. Either way, the bottleneck is the same: someone with a position has to decide what to say. AI can help with everything after that. It cannot do that part.

If your content strategy assumes that AI will eventually be able to do the upstream work — that with the right prompts, the right context window, the right fine-tuning, you will get something that has a real point of view — you are misreading what these models are. They are very fast averagers. The future where averagers replace specific, opinionated humans is not the future. It is just a slightly faster version of the present, with more content and less to read.

You still need someone to decide what is worth saying. That has always been the actual job.

When a founder hesitates to spend $5,000 on a brand rewrite, the hesitation is almost always framed as a budget question. It is not a budget question. It is a math question — and most founders avoid the math because the math is uncomfortable.

Here is the math. Take your average client value. Multiply it by your typical sales cycle conversion rate from website visitor to closed deal. Then ask: what happens if the copy on your homepage moves that conversion rate from 1.2% to 1.5%? It is a small absolute change. It is also, if you are running any meaningful traffic, the difference between a single mid-size client per quarter and two of them. Annualize it and the number is rarely under six figures.

Now apply the same math to your sales page. Your founder bio that's on every pitch deck. Your LinkedIn profile that every prospect opens before they take the meeting. Your cold email subject lines. The copy that runs your business is not a creative line item. It is the friction coefficient on every revenue motion you have. And friction compounds.

The founders who avoid this math are usually the same founders who will spend $40,000 on a quarterly ad budget without flinching. They will pay an agency $15,000 a month to manage that budget. They will hire a sales rep at $120,000 base. And then they will send all of that traffic to a homepage they wrote on a Tuesday in the back of a Lyft, three years ago, when they were a different company.

The asymmetry is staggering. You are paying for the audience. You are paying for the funnel. You are paying for the people. The single piece that sits between every prospect and every conversion — the words on the page — is the cheapest and most-neglected component of the entire stack. And it is the only one with a multiplier effect on everything else.

The Hidden Costs No One Tracks

The visible cost of bad copy is the conversion rate you can see in your analytics. The hidden cost — which is significantly larger — shows up in places you are not measuring.

Bad copy raises your sales cycle length. Prospects who arrive uncertain about what you actually do take longer to qualify, ask more clarifying questions, and require more sales touches before they close. Each additional touch costs you about 20-40 minutes of senior time and reduces your sales team's capacity for high-value work. Run the per-deal cost out and the difference between a 5-touch cycle and a 7-touch cycle, across your annual deal volume, is rarely under a quarter of someone's salary.

Bad copy raises your customer acquisition cost across every channel. When your landing pages do not pre-qualify visitors, your sales pipeline fills with unqualified leads who waste demo slots and inflate your CAC numbers. Your CFO sees the CAC and tells the CMO to cut spend. The CMO cuts spend and pipeline shrinks. The shrinkage gets attributed to “the channel” — but the channel was fine. The copy was sending the wrong people through.

Bad copy raises your pricing ceiling. The most underdiscussed effect of weak positioning is that it caps what you can charge. When your homepage describes you in the same generic terms as the rest of your category, you are commoditized by default — and commodities compete on price. Sharp positioning creates premium permission. The same exact service, with copy that frames it as distinct and considered, sells for 30-60% more without changing the deliverables. I have seen this play out across enough engagements to stop being surprised by it.

A handful of founders run this math and decide it is still not worth investing in. They are usually running businesses where the copy is in fact already strong, or where the bottleneck is somewhere else entirely. Most of the time, the math produces a number that makes the typical fee for a serious rewrite look comically inexpensive. Spending $5,000-$15,000 to recover an annual $60,000-$400,000 leak is not a budget decision. It is the most obvious ROI move on the list.

The reason founders avoid this math is not financial. It is psychological. Copy feels subjective. It feels like the kind of thing a smart founder should already be able to do. Spending money on a writer feels like admitting that the version of the company you have personally written into existence is not landing. That is uncomfortable. The math, on the other hand, is not uncomfortable. The math is just the math.

If you have not run it on your business yet, run it. If you have run it and the answer is bigger than you thought, that is the answer.