AI can generate infinite music, but who stops the Salt & Vinegar donuts? Why agentic AI music labels still need humans with taste, judgment, and veto power.

I caught a story on CBS Sunday Morning about AI-generated music—which tells you everything you need to know about where we are in this conversation.

When CBS Sunday Morning covers something, it means your parents are now thinking about it over coffee and crossword puzzles. AI music synthesis has been around for years, but now it’s officially reached “mainstream cultural anxiety” status.

And honestly? That’s when things get interesting. Because it sent me down a rabbit hole that ends somewhere between “this changes everything” and “someone please stop the machines before they ruin music the way they ruined donuts.”

Stay with me here.

THE KIDS ARE ALREADY LIVING IN THE HOLOGRAM

Let’s get one thing straight: the “will AI replace real musicians?” debate is already over, and we lost—or won, depending on your perspective.

Here’s why: The generation inheriting the music industry doesn’t care about your definition of “real.”

They grew up watching VTubers. They spent actual money on Fortnite skins. They treat Hatsune Miku—a literal hologram voiced by synthesis software—as a legitimate artist who sells out stadiums. To them, she IS real. She has personality, lore, a community, drama. That’s the mythology. That’s what matters.

So yes, entire AI-generated music labels are coming. Probably already here, lurking in your “chill study beats” playlist like musical kudzu. And honestly? The economics are too compelling to resist: zero studio time, no tour bus disasters, no rehab stints, no arguing about who gets credit for the bridge. Just pure, frictionless content generation optimized for algorithmic distribution.

The prompt engineer is already here. They’re just called “music supervisors” or “creative directors” right now, figuring out which combination of parameters creates something that doesn’t sound like a robot having an existential crisis.

BUT HERE’S WHERE IT GETS INTERESTING

This isn’t about AI replacing human creativity. It’s about AI requiring a completely different KIND of creativity that might actually be more complex than what came before.

Think about it: When synthesizers showed up in the ’70s and ’80s, people lost their minds. “That’s not REAL music!” Same pearl-clutching that happened when electric guitars “ruined” music, when drum machines “killed” session drummers, when sampling “destroyed” originality.

Every generation’s authenticity crisis becomes the next generation’s baseline.

I spent a decade as a DJ—remixing, sampling, curating, entertaining. That was already a form of agentic creativity. I wasn’t creating every sound from scratch. I was:

Curating (choosing what matters from infinite options)
Contextualizing (knowing what works for THIS room, THIS moment)
Intervening (reading the crowd and adjusting in real-time)
Orchestrating (making disparate elements cohere into something greater)

Sound familiar? That’s exactly what the future of AI music looks like—except instead of managing records, you’re managing narrative architectures, transmedia storylines, and parasocial relationships at scale.

ENTER THE SALT & VINEGAR DONUT

Here’s the problem: Agentic AI is incredible at optimization within parameters. It’s terrible at knowing when the parameters themselves are insane.

I wrote about this phenomenon when a fictional CPG company decided Salt & Vinegar donuts were a good idea. Clearly, nobody with functioning taste buds was in the room to say, “Hey, maybe we pump the brakes on this one.”

The AI looked at the data:

Sweet/savory mashups are trending ✓
Salt & Vinegar is a popular chip flavor ✓
Donuts are beloved ✓
Conclusion: MAKE SALT & VINEGAR DONUTS

It’s not wrong based on the data. But nobody asked if they should.

THE FUTURE NEEDS CREATIVE DIRECTORS, NOT JUST CREATORS

The same thing will happen with AI music labels. The algorithms will:

Analyze trending sounds ✓
Generate hooks optimized for 15-second TikTok clips ✓
Create “personality” based on engagement metrics ✓
Build storylines that maximize parasocial attachment ✓

But who’s there to ask the important questions?

“Is this storyline getting too dark?”

“Are these lyrics technically coherent but meaningfully empty?”

“Have we optimized ourselves into creating the musical equivalent of Salt & Vinegar donuts?”

You need humans with taste. Not just technical knowledge—judgment. The ineffable thing that makes you say, “I understand WHY the algorithm suggested this, but trust me, we’re not doing it.”

Because algorithms optimize for engagement, not for good. They’ll chase metrics right off a cliff. They’ll make Salt & Vinegar donuts all day long because the data supports it, even though anyone with functioning taste buds knows it’s a crime against breakfast.

IT’S MARVEL, BUT MAKE IT MUSIC

Building an AI music label isn’t really about the music generation itself—that’s table stakes. It’s about worldbuilding at scale.

You need:

Transmedia storytelling chops (the Instagram story that connects to the TikTok that references the Discord easter egg)
Community management (fans will build their own lore; you need to guide it)
Narrative architecture (deciding when your AI artist enters their “experimental phase”)
Someone who understands parasocial relationships at a PhD level

It’s the Marvel Cinematic Universe model applied to music. It’s professional wrestling with better production values. Everyone knows it’s staged, but if the storyline is compelling enough, who cares?

This might actually be MORE human-intensive than traditional music—just different humans. You’re not paying for vocal coaches and session musicians. You’re paying for writers, strategists, community managers, and someone with the taste to say “no” when the algorithm suggests something insane.

THE BANDWIDTH PROBLEM (OR: COMING SOON TO ELVIS DURAN)

Here’s where it gets really wild: Human celebrities have a bandwidth problem. AI celebrities don’t.

Taylor Swift can’t do Elvis Duran at 8am, Good Morning America at 9am, and ten podcast interviews simultaneously. She has to sleep, travel, get sick, have bad days, cancel when she’s exhausted. Her publicist has to carefully ration her appearances to maximize impact while preventing burnout.

But AI Taylor? She could theoretically do a hundred interviews at once, each perfectly tailored to that show’s audience, with the “spontaneity” dialed to whatever level feels authentic for that format.

Is there a day coming when Elvis Duran interviews an AI celebrity? When morning show hosts, podcasters, and radio DJs are chatting with synthetic entities that can maintain dozens of conversations simultaneously?

And more importantly: Would anyone care?

The honest answer is: it depends on how the fourth wall gets handled.

If the AI artist tries to hide the simultaneous appearances, someone will notice. The continuity errors, the impossible scheduling, the moment they reference something that hasn’t happened yet—it’ll shatter the illusion, and the backlash will be swift.

But what if being openly AI IS the positioning? What if the interview questions become: “So what’s it like existing in 47 places simultaneously?” That’s actually more interesting than asking another human celebrity about their workout routine.

The smarter play, though, might be something more subtle: Use the bandwidth advantage for depth, not breadth.

Don’t do every show simultaneously. Maintain the illusion of scarcity with traditional media (three carefully chosen interviews per week). But use that infinite bandwidth for fan engagement at scale—responding to DMs, creating personalized content, showing up in Discord servers, making thousands of fans feel individually seen.

Because here’s what matters: Distribution and PR aren’t about ubiquity. They’re about building belief.

The AI celebrity that appears everywhere simultaneously becomes background noise. The AI celebrity that seems exclusive and hard to get, while somehow making each fan feel special? That’s the one that wins.

It’s the same challenge human celebrities face, just with different constraints. You still need:

Strategic media placement (not just “everywhere”)
Narrative consistency across platforms (harder when you’re in 47 places at once)
The ability to go viral authentically (algorithms can detect try-hard desperation)
Community management that doesn’t feel like community management

The bandwidth advantage doesn’t eliminate these challenges. It just means you can’t use “we’re too busy” as an excuse anymore.

THE BOTTOM LINE (UP FRONT, BECAUSE YOU’RE BUSY)

AI isn’t killing music. It’s creating a new discipline that sits somewhere between showrunner, brand strategist, and creative director.

The skill isn’t “can you play guitar?”

It’s “can you maintain narrative coherence across platforms while preventing your synthetic artist from accidentally becoming a cautionary tale?”

It’s “can you read the room—not just the data—and know when to pull the plug on a bad idea before it becomes a Salt & Vinegar donut?”

The complexity is perfect for Agentic AI. But the judgment still needs to be human.

Because at the end of the day, someone needs to be in the room to veto the Salt & Vinegar donuts. Someone who understands music—or brands, or culture, or humans—at a fundamental level.

Someone who knows that just because you can doesn’t mean you should.

Quinn Harrington has been saying “no” to bad ideas professionally for over 20 years. As founder of Harrington Design Company, she helps brands avoid their Salt & Vinegar donut moments through strategic intervention disguised as design. She’s also a recovering DJ who still believes the drop should hit at exactly the right moment, algorithm be damned.

Previous PostWhy Hire a Branding Agency? The Question Every Business Owner Eventually Asks

agentic AI AI music algorithmic creativity artificial intelligence brand strategy creative director creative judgment future of music Hatsune Miku music industry music production music technology synthetic celebrities transmedia storytelling VTubers

The Salt & Vinegar Donut Problem: Why AI Music Labels Still Need Humans With Taste

I caught a story on CBS Sunday Morning about AI-generated music—which tells you everything you need to know about where we are in this conversation.

THE KIDS ARE ALREADY LIVING IN THE HOLOGRAM

BUT HERE’S WHERE IT GETS INTERESTING

ENTER THE SALT & VINEGAR DONUT

THE FUTURE NEEDS CREATIVE DIRECTORS, NOT JUST CREATORS

IT’S MARVEL, BUT MAKE IT MUSIC

THE BANDWIDTH PROBLEM (OR: COMING SOON TO ELVIS DURAN)

THE BOTTOM LINE (UP FRONT, BECAUSE YOU’RE BUSY)

Let’s start a project!

Our Reel

Recent Articles

I caught a story on CBS Sunday Morning about AI-generated music—which tells you everything you need to know about where we are in this conversation.

THE KIDS ARE ALREADY LIVING IN THE HOLOGRAM

BUT HERE’S WHERE IT GETS INTERESTING

ENTER THE SALT & VINEGAR DONUT

THE FUTURE NEEDS CREATIVE DIRECTORS, NOT JUST CREATORS

IT’S MARVEL, BUT MAKE IT MUSIC

THE BANDWIDTH PROBLEM (OR: COMING SOON TO ELVIS DURAN)

THE BOTTOM LINE (UP FRONT, BECAUSE YOU’RE BUSY)

Related Posts

Why Hire a Branding Agency? The Question Every Business Owner Eventually Asks

Someone Needs To Veto The Salt And Vinegar Donuts

Let’s start a project!

Our Reel

Recent Articles