Taste Killed More Good Content Than AI Ever Will

I recorded an episode with Kayte Grady this week, Content Director at BlueConic, and I went in expecting her to agree with me. She didn’t. That turned out to be the most useful thirty minutes I’ve had in a while.

Here’s the question I asked her, and I’ll admit it was a loaded one.

Before any of this AI nonsense showed up, I watched CEOs take good copy in final review and kill it. Not because the research was wrong. Because they personally didn’t care for it. “My taste says put this on the homepage instead.” No data. No test. A copywriter’s week in the trash, and a homepage that reads beautifully to exactly one person.

Now I’m watching the same leaders do the same move with a new tool. ChatGPT hands them something that sounds like them, they like how it sounds, and it goes live. Same instinct. Faster delivery.

So I asked Kayte why leaders find it so hard to trust the people they hired.

She defended them

Her answer was better than my question.

“Think about a book you like. Think about a style of writing you like, a style of music you like. Taste is personal, taste is individual, taste is unique.”

Then the part that stopped me:

“I think I have the best taste in writing, but there are things that I may not take into consideration. If you have a CEO or a CMO or some C-suite who comes in and they’re like, ‘Oh, I don’t like that,’ that it’s based in taste, not necessarily in wanting to undermine or, you know, dog on the level of work and thought that went into it.”

Her read is that the executive killing your draft usually isn’t dismissing your expertise. They’re having a reaction, the same reaction any of us has to a song or a book, and they happen to have the final say. And her advice to the writer on the other side of it is to stop treating that as an injustice and start treating it as an argument you can win:

“If you did the research, you know what’s right and what’s correct, you fight for it. You give them a reason.”

She compared it to how product teams operate. Nobody expects a PM to get a feature approved by being offended that it was questioned. They come with a reason. Writers, she thinks, should hold themselves to the same standard.

And then she conceded the thing most people won’t:

“At the end of the day, you know, your boss or your boss’s boss is gonna have the final say, and if their taste is what wins, their taste is what wins. Like that’s, that’s the way it’s gonna be.”

I told her on the recording that I thought this was too forgiving to leaders (I still do).

But she also named something I’ve been circling for a year without landing on it, which is that most of these fights aren’t about competence at all. They’re about two people using the same word for different things.

Quality is measurable. Taste is personal.

These get treated as one axis, but they aren’t.

Quality is checkable.

• Does this answer the question the reader actually asked?
• Is it accurate?
• Does the structure hold up?
• Would a customer who uses the product recognize it as true?

Every one of those has an answer that exists outside your head.

Taste is whether you like the tone. There’s no test for that, because the only instrument is your own preference, and your preference is not the market.

The confusion between them is expensive in a specific and under-discussed way: killed content generates no data. If you spike a draft before it ships, nothing ever contradicts you. There’s no bounce rate, no conversion, no comment section. You get to be right forever about work that never existed.

That’s an unfalsifiable process, and unfalsifiable processes drift toward whoever has the most authority in the room.

Nobody would accept this anywhere else. Nobody kills a feature in staging and then claims the roadmap is validated.

Guardrails are how you take taste out of the loop

Kayte’s answer to keeping one brand voice intact across a team was the least abstract answer I’ve heard anyone give:

“I think protecting the brand voice just gets this abstract, like POV, right? We talk about it in a really abstract way, but it’s really, it comes down to fundamentals. And so the first thing is guardrails.”

And what those guardrails contain:

“Words we naturally use, things that we avoid, how we talk about the tech, how much depth we go into for the tech, you know, is this our tone? Is this the structure we typically go by?”

The part that matters most is who they apply to:

“Whether it’s myself, whether it’s AI, whether it’s another person, whether it’s, you know, demand gen creating ads, it, it doesn’t matter. We have guardrails in place.”

Herself included. That’s the whole thing. Guardrails that exempt the person who wrote them are a style guide for junior staff, not a quality system.

And this is what actually resolves the tension between her position and mine.

She’s right that you’ll never remove taste from leadership, because someone with final say will always have a gut reaction to the work. But you can make the criteria specific enough that “I don’t like it” stops being a sufficient reason to kill something clearing every measurable bar. Make the leader argue against the standard rather than their mood that day. And if the standard turns out to be wrong, change the standard, in writing, so the next draft benefits.

That’s the difference between a decision and a preference.

Taste is also wrong sometimes, and she proved it on herself

The best evidence against gut reaction in this conversation came from Kayte’s own story.

When ChatGPT first launched, her CMO called her and her director of content in to show it off, and typed a prompt asking it to write a story about the two of them.

“Robots are running the world. They’re gonna destroy my career. Like, this is the worst thing I’ve ever seen.”

She shut it down. Wouldn’t use it. In her words, it gave her “the ultimate ick.”

Today she runs an agent stack she describes as far too complicated, built out of Claude and n8n, and she’s blunt about the fact that she built the engine and still puts the keys in herself every time instead of letting it run.

She was completely certain, and completely wrong, about a technology she now works in daily. That’s not a knock on her. It’s the honest case for why no single person’s reaction, however senior and however experienced, should be the final quality gate. Taste is a real signal. It just isn’t a reliable one, and the people with the most of it are the ones who most need something to check it against.

The 2027 version of this problem

She also said something that I think is the actual stake here. On the SEO-templated content most of us produced for a decade:

“A lot of it was performative. I mean, we, we did it, right? And we did it for, as a means to an end, but I, like, welcome the replacement of that.”

Then the question underneath it:

“If the machine can do it, right? And the machine, we know the machine can do a lot. It might not be able to do it all well, but it can do it a lot. What is the human there for?”

Her answer is judgment, and she’s specific about where hers came from. She used to print what she wrote, take a red pen, read the words out loud, and circle the things that didn’t make sense.

“So would somebody actually say this? Like, does it sound good, or does it make sense? Because those are two different things.”

Ask whether she’d rather have skipped those years and let the tools handle it and the answer is no:

“Without that skill, I wouldn’t be able to use AI as well as I do now.”

That’s the part people racing to automate keep missing. The judgment isn’t left over after the machine does its part. It’s the thing that makes the machine’s part usable, and it was built by doing the boring work first.

An honest admission

I don’t think you can get taste out of leadership, and I’m not sure you should want to. Kayte is right that a leader with a strong reaction is often reacting to something real that the writer didn’t consider.

She’s also right that most writers lose those arguments because they show up offended rather than prepared.

Where I still (respectfully) disagree with her: “their taste is what wins” is a description of how it works, not necessarily a defense of it. Plenty of things that are true about org charts are bad for the business, and this is one of them. The fact that a CEO gets the final call doesn’t mean the final call should be made on a preference nobody wrote down.

What I’d add to her framework is a burden of proof. If it clears the guardrails and someone kills it anyway, that’s fine, but it should cost something: a written reason, and an update to the standard so the next person isn’t guessing. Preferences that survive being written down are usually judgment. The ones that don’t survive it were never judgment to begin with.

Three people doing the work of fifteen only works if quality is defined somewhere other than one person’s gut. Otherwise you’ve automated production and left the bottleneck exactly where it always was, except now it processes drafts faster.

In a factory like Willy Wonka’s, pipes come before the chocolate river can exist. Build the system that decides what’s good before you build the system that makes things fast.

---

Kayte Yerga Grady is Content Director at BlueConic. The full conversation is on Barely Shipping. Quotes are from the episode transcript, lightly trimmed for filler.