Dear B2B CEO,
Your Head of Marketing sends the new homepage copy for approval. Before you reply, you paste it into ChatGPT and type “make this better.”
Ten seconds later you have a cleaner version. Tighter sentences, a punchier headline. You forward it with two words: “use this.”
It feels efficient. It’s the third time this month.
I want to tell you what that email looks like from the other side, because nobody on your team is going to.
What Your Team Hears
This spring, Rachel Karten, who writes a widely read newsletter for social media marketers, published a piece called “Does your boss have AI brain?” She filled it with what her readers told her about their leaders.
It reads like the minutes of a support group.
“It didn’t matter that I was writing excellent content,” one marketer wrote. The CEO “fully believed AI was better at everything.”
Another: “CEO was like ‘don’t share anything (i.e. drafts for review) if you haven’t run it through AI.'”
Another: “Literally every action item from a meeting is met with ‘just ask Claude it’.”
And one that should worry you, from a copywriter: “I’m a copywriter contemplating leaving the industry bc the art of feedback has been completely outsourced.”
It hasn’t let up since. In August, a marketer posted on Reddit about a manager who rejects almost every piece of writing outright, without asking for changes. “They’ll then take the general idea, put it through AI and go with that output.” The post ends: “I’m feeling a little depleted and untalented.”
Karten names the gap in two sentences. “To a boss, that’s just them being ‘efficient’.” To the employee, it sounds like the boss saying, “I trust AI more than I trust you or myself.”
Read the last two words again. Or myself.
Your team concludes that you’ve stopped trusting them. Worse, they conclude that you’ve stopped trusting your own judgment as well, and handed both to a machine.
What the Machine Does With Your Question
Think about what actually happens when you type “make this better.”
A language model writes the most likely next words, based on everything it has read. That’s what makes it fluent. It also means its natural home is the middle. Karten again: “we’re watching decision makers run big ideas through a machine that finds the median.”
The research backs her up. Wharton researchers, publishing in Nature Human Behaviour last year, found that people brainstorming with ChatGPT came up with narrower sets of ideas than people working without it. In one task, participants had to invent a toy from a brick and a fan. Nine of them, working separately, gave their toy the same name: “Build-a-Breeze Castle.”
It gets closer to your desk. Researchers at NYU Stern asked leading models for strategic advice across thousands of simulated companies. The models “almost uniformly select the same trendy strategies,” whatever the company’s situation. The researchers have a name for it: trendslop.
Then there’s the second problem. These models are trained to please the person asking.
Stanford researchers tested 11 leading models and published the results in Science this March. The models affirmed users’ actions 49% more often than other people did. After a single conversation with an agreeable model, participants were more convinced they were right, and they trusted that model more. The study looked at personal conflicts rather than business plans, and other research suggests the pull is milder when people make business decisions. The direction is the same, though. The machine leans toward you.
Even OpenAI admitted it last year, when it rolled back a ChatGPT update that it said had become “overly supportive but disingenuous.”
Ask “can this be better?” and the answer is always yes. Ask whether your team’s idea is too risky and you’ll get a tidy list of risks. The machine hands your doubt back to you, in confident, polished prose.
So the loop runs like this. You already had doubts about marketing. You asked a machine that leans toward the average and toward agreeing with you. It agreed, and it sounded smarter than your team.
Why It Feels Right
I get why you do it.
You’ve probably been burned. An agency that delivered decks instead of pipeline. A marketing hire who talked well and shipped little. When you’ve paid for confident opinions that never paid off, a tool that gives you a second opinion in ten seconds, for twenty dollars a month, without getting defensive, feels like getting some control back.
And the tool is genuinely good at parts of this. I build AI-enhanced marketing operations for a living and use these tools every day. They catch clumsy sentences, find holes in an argument, and produce a first draft faster than any person can.
But look at the instruction. “Make this better.” Better for whom? Better at what? Compared to what?
The machine fills those blanks with the average answer, because you gave it nothing else to work with.
What It Costs You
The bill arrives about six months after that email.
First, your experts stop arguing. Why fight for the strange, specific headline that came out of three customer interviews if it will be “improved” back into the generic version anyway? Karten’s readers describe that resignation in plain words: “A huge waste of time and effort on everyone’s part, not to mention demoralizing as hell.”
Second, they start pre-chewing. If every draft gets run through AI at your desk, the sensible move is to run it through AI before it reaches you. Now nobody in the chain applies judgment. Everyone approves the machine’s version of the machine’s version.
Third, your best people check out. The copywriter thinking about leaving the industry is one person in one newsletter. In your company, it’s the person who knows why your best clients actually bought, sitting quietly while the homepage turns into everyone else’s homepage.
Fourth, your outside experts stop telling you the truth. A consultant who works with small businesses put it bluntly in a Reddit thread this August: “I’m currently watching one particular CEO destroy his business as I work on ending my contract so I don’t get blamed for this mess. He took on product design and marketing by himself with Claude’s aid, and has destroyed his 2 lead products and ruined the marketing campaigns.”
He saw it coming, ran out of polite ways to say so, and is leaving before the blame arrives. The CEO will probably never hear why.
If you run a services firm, you may already know this habit from the receiving end. In an October thread, another consultant described clients who now “argue with findings or outputs over email which is clearly chatGPT reviews of the work, without any context.” He was about to fly to a client in Holland “to sit them in a meeting room for 3 days to take them away from ChatGPT.” You know how that feels when a client does it to your team. Your own team feels the same way when you do it to them.
The end state is marketing that sounds like the median of your category. In B2B with long sales cycles, that is the one position you can’t win from. Buyers already struggle to tell vendors apart. If your copy comes from the same machine as your competitor’s, polished by the same three words, you’ve paid a team to help you become harder to remember.
For a services firm, the damage lands on the proposals and capability decks where your expertise is supposed to show. For a software company, it lands on the launch copy and positioning that should make a buyer ask for a demo or discovery call instead of a comparison sheet.
Where the Machine Belongs
The fix is to change the job you give the machine.
AI is a strong checker and a poor judge. It can tell you whether a piece of copy matches a standard. It can’t tell you what the standard should be. That has to come from you and your team: who you serve, which problem you end, what you believe that your competitors don’t, and what you would never say.
From my experience, most companies have never written that down. That is the real gap. When a marketing function has no written point of view, there is nothing to check the machine’s output against, so the machine’s output becomes the point of view. Structure closes that gap: a documented position, a defined review step, and a clear answer to who decides.
As a managing editor, I learned that good feedback rests on two things: someone with the expertise to give it, and a standard everyone has agreed to. Take away both and you get opinions. Take away both and add a chatbot, and you get the average opinion, delivered with confidence.
With a written standard in place, AI becomes genuinely useful at your desk. Paste in the copy and your positioning document, and ask: “Where does this drift from what we said we stand for?” That question has an answer. “Make this better” doesn’t.
Three changes, all small:
- Write the standard down, once, with your team. Then point the machine at it.
- Keep feedback human. If you think the copy is weak, say why in your own words. If you can’t say why, that tells you something too.
- Decide where AI sits in the process, and say it out loud. Your team should never have to guess what you’ll do with their work.
Two Words Instead of “Use This”
Next time the homepage copy lands in your inbox, try a different reply. It’s also two words.
“Why this?”
Let your Head of Marketing walk you through the choices. Within five minutes you’ll know whether there is judgment behind the work. If there is, you just got something no chatbot can give you. If there isn’t, you’ve found the real problem, and no amount of “make this better” was ever going to fix it.
Your team has been waiting to be asked.
Yours sincerely,
