“There’s never been an easier time to look like you are doing marketing, but you are actually flaming up cash,” said Murphy during a discussion at Fortune’s Brainstorm Tech conference in Aspen this week. “The message is not getting into the brain, it’s flicked away in 200 milliseconds, and it is as forgettable as most AI slop actually is.”

To be clear, like most companies Liquid Death is using AI to an “extreme level” behind the curtain, Murphy said, and employees celebrate it in workplace messaging platform Slack and joke that Anthropic’s Claude is their new direct report. However, true “zero to one thinking,” said Murphy— referring to investor Peter Thiel’s concept of creating something entirely new—isn’t where AI excels relative to creative human beings. 

He cited Liquid Death’s collaboration with Spotify as a quick study of the issue. The drink company’s band of “crazy” artists and comedians—including some who previously wrote for satirical publication The Onion and Cartoon Network’s “Adult Swim”—dreamed up the world’s first bluetooth-enabled urn. The reason? So people can listen to music after they’re dead, said Murphy. It cost “a couple hundred grand,” which Spotify covered, and earned the company 6 billion earned media impressions, he added.

“We spent a very little amount of money, and we’re able to trace that back to real-world results, awareness that peaks, sales that go through the register, traffic that comes to life with that,” said Murphy. “Our God metric, FYI, for our content that we put out there, is shares. That’s it—is it compelling enough?”

Vishal Sood, president of R&D at AI brand and creative management platform Typeface, agreed to an extent, noting portions of creative work are simply out of reach for AI. 

“Taste and judgment is not AI’s game right now,” said Sood. “You have to leave it to the people who really understand.” He cited a quote from notable AI researcher Andrej Karpathy, who has said repeatedly, “You can outsource your thinking, but you can’t outsource your understanding.”

Where AI excels, said Sood, is in taking an existing idea and creating new versions, images, storyboards, and drafts. The repetitive production work is where AI can earn its keep, he said, particularly in business-to-business marketing. He cited one customer that lifted its email click-through rate by two percentage points through AI personalization, almost tripling its previous rate. He also pointed to the task of taking a campaign’s “key visuals” and resizing them into the hundreds of formats digital media requires that sometimes takes months at large companies.

“All of that is totally AI game,” said Sood.

The distinction is in determining which slice of the work is irreplaceable and the main purview of humans, and which slices can be delegated to AI. 

“That requires taste and judgment—you do not outsource to AI,” Sood said. “You’ve got to know where your secret sauce is.”

Stacy Simpson, chief marketing officer of athenahealth, said her team uses AI tools throughout its operational work, and it gets campaigns to market faster and shrinks the time between steps. That said, she keeps AI away from the creative core processes. 

“We use it in our creative process,” said Simpson. “We do not use it in our creative ideation.”

Simpson’s litmus test is whether the tools are solving a real problem. “Just because you can, doesn’t mean you should,” Simpson said.

For Caitlin Allen, president of marketing at Simbe Robotics, the reason for the steady stream of what people think is slop is that there’s a lot of output, and far less input.

“There’s a difference between creation and listening,” she said. “The reason that AI slop exists is because there’s so much creation.” When companies tackle imaginative, artistic work, said Allen, they tend to over focus on what they want to say, rather than on how to communicate with an audience and what people want to hear. The more promising use, in her view, is for AI to automate “the repetition of listening” so marketers can determine what people really care about.

Ben Gammell, president of fintech startup Brex, said the company views AI as “an accelerator for the people you have,” not a reason to cull roles. 

“We don’t think about AI taking our jobs, we don’t think about restructuring our workforce because of AI,” he said. “We’re really thinking about, ‘How can AI accelerate what you’re doing?’”

Ultimately, AI tools can be the most effective and well-designed in the world, but two people can use the same tool and get to vastly different results, said Simpson. Two people can run the same model and produce wildly different results, Simpson noted.

“That is crap, and that is amazing, and they have the exact same tools,” she said. The distinction will always come down to good judgment, well-informed context, and the skill and experience too know what’s worth creating and what’s best to delegate.