I can’t put it more plainly than that headline. If you’re in a position to hire (and fire) marketing personnel, don’t think for one second that AI is your payroll-slashing solution.
I’ve spent the past year testing and teaching the generative AI platforms ChatGPT, Claude, and Gemini on a weekly basis. I’ve similarly used other AI programs specializing in design, illustration, and video. Most recently, in March I chaired a seminar about AI in marketing, tapping four individuals deeply experienced in coding and/or using specialized AI agents to share insight. These four professionals confirmed what I discovered organically. AI is phenomenal at some tasks, but it’s also really bad at others when it lacks human guidance. They also confirmed that the human touch is essential, and that they themselves caution companies against reducing headcount. Here’s why.
AI needs specific prompts. As smart as generative AI’s programmers are, you do still need to provide guidance. The first task I ever gave ChatGPT was to write a story about a high-profile charter yacht. That was pretty much the prompt, too: “Write a 300-word blog post about the yacht XYZ.” I didn’t intend to publish the post. Instead, I wanted to understand how it worked its magic. The result, in seconds flat, was deeply disappointing generic fluff. I quickly realized it relied on overly flowerly descriptions from some charter brokers. It took me a while longer, though, to realize my own failure was the cause. I didn’t specify that I wanted a journalistic tone to the story. Whether you use ChatGPT, Claude, or Gemini, always make sure to specify the tone you want: polished and professional, inspirational, funny, etc. If you forget in your initial ask, you’ll likely find that it currently solicits you for the direction.
AI can and still does hallucinate. The CEO of a major publishing company decided last year that every magazine he owned needed to embrace AI. What’s more, he decided that AI could replace freelancers. “Just feed the press releases into ChatGPT,” he said. Editors pushed back, sounding the alarm that personal, long-term relationships were key to uncovering the stories behind the stories. Some of them were fired. The magazines suddenly began publishing multiple articles written by “staff,” not specific editors or freelancers. Imagine the surprise of readers—and advertisers—in finding glaring errors and made-up facts in those stories. Imagine what would happen with a major ad or marketing campaign. .
AI “improves” things not needing improvement. Onen of my symposium speakers shares a story that should make everyone sit up and pay attention. A company asked for her help in adopting AI. Good move. She provided advice on proper prompting, double checking facts, and more. An even better move, it turns out. One of the employees uploaded a report to ChatGPT about the company’s overall excellent performance, asking it to make the content sound better. ChatGPT proceeded to polish up the writing—and round up all of the sales figures. In its defense, it did exactly what it was told to do. Had it not been for the employee realizing the numbers didn’t match, the results would have been released. Publicly.
Generative AI is a phenomenal tool. But at the end of the day, it’s just that, a tool. A hammer needs a literal guiding hand to be effective. The same is true of a welding torch, design software, and a host of other things we take for granted. Treat AI the same way.