I believe people are the whole point of everything. That's why.
I'm not pro-AI because people can't fucking use it properly. The output of careless prompting is the specific reason I won't put my name on the pro-AI side of this conversation. The pro-AI camp wants the headline to be capability. The actual headline is what gets shipped, and what gets shipped is mostly garbage.
I'm also not anti-AI. I run my marketing in Claude. I built a paid prompt library aimed at marketers using Claude. I've spent two years figuring out what good looks like. The output of careful prompting is the specific reason I won't put my name on the anti-AI side of this conversation either.
The AI conversation has settled into two camps that hate each other and agree on the wrong thing. Both treat the tool as the variable and people as the variable too. Both miss the structural fact: people are the constant. The tool is the variable. You pick the tool to serve them. Not the other way around.
The position the conversation can't make space for is the only one I find defensible. I'm not pro-AI. I'm not anti-AI. I'm pro-people, and AI is one of many tools that either serves them or doesn't, depending on how it's wired up.
That's not a centrist take. It's the only honest one. The reason it's not louder is that both extremes get to be loud. The middle requires actually doing the work and showing it.
The pro-AI camp treats capability as the headline. "Look what the model can do." Every AI marketing pitch is a demo of the tool's ceiling, run on a perfect input, with the cherry-picked output framed as the new normal.
The pitch assumes adoption is the goal. If only marketers would use it more. If only the laggards would catch up. If only the holdouts would see what's possible. Adoption framed as the metric, as if the question of WHAT we're adopting and HOW were already settled.
It's not. Adoption rate is the wrong metric because the value of the tool depends entirely on the user's judgment. A team using Claude carelessly produces more slop, faster. Adoption-rate metrics treat that as success. The audience doesn't.
The thing pro-AI keeps missing is that the bottleneck isn't the tool. The bottleneck is taste. Faster shipping doesn't fix taste. It exposes the absence of it.
The anti-AI camp is right about the OUTPUT. Most AI-generated marketing content is slop. The proof is in your inbox. The proof is on your LinkedIn feed. Sameness, hollow specificity, structurally-perfect emails that nobody finishes. The diagnosis is correct.
The anti-AI camp is right that careless adoption causes real damage. Trust erosion. Voice sameness across competitors. Real human writers laid off because their managers thought a junior could be replaced by a prompt. Real audiences trained to filter out anything that smells generated.
The anti-AI camp is wrong about the source. The slop isn't coming from the tool. The slop is coming from the careless user. The same model that produces "Direct, confident, approachable. Clear, concise language" for a lazy prompt produces a measurably specific, defensible voice guide for a structured one. The model isn't the variable. The user is.
The anti-AI camp is also wrong about the solution. Refusing the tool doesn't bring back the people who got laid off. It doesn't reverse the trust erosion. It doesn't recalibrate the audience's filters. Walking away from the tool just means the people doing it carelessly own the playing field.
The thing both camps refuse to look at is that the question isn't WHETHER. It's HOW.
Whether to use AI is a settled question for anyone whose job is shipping marketing in 2026. The competitive cost of opting out is too high. The marketers who refuse the tool entirely are quietly losing to the marketers using it deliberately, regardless of what either side says publicly.
The interesting question is HOW. Specifically:
These questions don't have viral takes attached. They require actually doing the work. They produce libraries instead of opinions. They build capability that compounds.
This is the work I do. The Marketing Prompt Hub is what comes out of it. Not a position on AI. A position on how to use it without being part of the problem.
Five concrete things I do, all of which the prompts in the Hub are built around:
Refuse-conditions in every prompt. "Refuse to declare a trend from fewer than two data points." "Refuse to recommend a positioning the brand can't credibly defend." Without these, Claude defaults to maximally helpful, which on a question the data can't support looks like a confident answer to a question that doesn't have one. The refuse-condition is the lever that prevents the most common failure mode.
Real source material pasted into every prompt. A brand voice prompt without samples produces generic adjectives. A competitive teardown without competitor content produces inferences dressed up as findings. The prompts in the Hub require you to do the work of gathering inputs. They won't run otherwise. The constraint is the feature.
Output specs that force structure. Not "tell me about X." Instead: "Output exactly five attributes, each paired with its opposite. Then a 150-word executive summary. Then three positioning hypotheses with kill criteria for each." Specs prevent the 4,000-word generic essay nobody asked for.
Conversational register, not function-call register. I talk to Claude like a smart junior on the team, not a search engine. The output mirrors the input. Search-query inputs produce search-query outputs. The whole post on this is here.
Refusing to use entire categories of prompt. No engagement-bait LinkedIn comment generators. No "rewrite this in [famous founder]'s voice" novelty prompts. No mega-prompts that promise to do everything badly. The library you should keep is the one you can defend prompt by prompt.
None of this is hard. It's just work. The pro-AI pitch promised you wouldn't have to do it. The anti-AI pitch tells you not to bother. Both are wrong.
The marketers using AI carelessly are paying a tax already. They just haven't noticed because the tax is paid by the audience, not by them.
Voice sameness. Trust erosion. The brand-shaped hole where a person used to be. The cleanup time when the AI-drafted email that "sounded fine" turns out to misrepresent the product. The slow drip of the audience deciding that everything from your category sounds the same and is therefore worth less attention.
The marketers using AI deliberately are compounding. Better output over time, because the prompts improve. Better instincts about when not to use it, because the failures get audited. Better voice consistency, because the voice is structurally extracted instead of sloppily described. Better arguments at the team meeting about what to ship, because the time saved on drafting is reinvested in deciding.
The two sets of marketers do not currently look very different. They will, by the end of 2027. The careful ones will be the ones still trusted by their audience. The careless ones will be the case studies in why "AI hype was overblown" thinkpieces, conveniently blaming the tool for what was always a judgment failure.
I'm not pro-AI. I believe people are the whole point of everything. The AI conversation keeps missing this because both camps argue about the tool. The tool isn't the question. The tool is a variable.
The constant is people. The marketers who use the tool to serve them are doing the work. The marketers who use the tool to replace them are paying a tax they can't see yet. The marketers who refuse the tool entirely are losing to both.
There's a third position. It's the one I take. It's not loud. It's just defensible.
People are the whole point. That hasn't changed.
Zara
This essay sits inside the Marketing Prompt Hub: 30 tested Claude prompts across 8 marketing disciplines, 6 playbooks, 5 agent blueprints, full reporting kit. Pay once, own the files. No email gate, no per-seat fees, no subscription.
See the hub →