George Carlin once said in an extended diatribe about white celebrities who play the blues, “It’s not enough to know which notes to play. You gotta know why they need to be played.”
The same can be said for the way companies are deploying LLMs for content creation: applying AI to the wrong parts of the process lowers quality and squanders expertise without significantly speeding things up, turning compelling insights into average content.
For instance, I had a conversation last week with a former colleague who has been an editor for more than 25 years. He remarked that gen AI has dramatically changed his job over the past six months—just not in the way he had expected.
Previously, he would have started with an outline or PowerPoint deck, reviewed the research and insights, and had a kickoff call with the authors to discuss the narrative before writing an initial draft.
Now, author teams are working directly with Claude to create a draft they believe is ready for publication. His role: polishing the draft as a last step before moving to production.
There are a few issues with this approach:
Collaboration at the beginning of a project elevates quality and sharpens insights
Author teams that skip this step risk overlooking what’s most distinctive in their research and analysis—a common challenge when you know too much and are too close to the insights. Author teams that work with Claude or ChatGPT will deliver an average draft not as a starting point but what they consider the finished product. The combination of an expert author team and experienced editor, particularly one who understands the current conversation and what target audiences care about, will produce more-compelling content.
Engaging writing that connects with audiences isn’t gen AI’s forte.
That requires an understanding of the specific challenges clients are facing rather than a synthesis of previously published content on a subject. As important, if you’re trying to connect with real people, finding that more human approach to storytelling can be the difference between true engagement or a quick discard. In addition, people seeking authentic insights are becoming hypersensitive to AI-generated content, which has the potential to erode trust when it matters most.
Focusing gen AI on drafting brings cutting-edge tech to the least onerous part of the process
For instance, say it takes eight weeks to go from kickoff call to publication. Developing the article may take two to three weeks; the rest is consumed by syndication, risk and legal reviews, production, and web staging. Starting with an AI-generated draft might—MIGHT—save a week at the front end, but it does little to speed up the rest of the process.
Don’t think I’m against the well-considered use of gen AI in other parts of the editorial process. It’s proved to be a great tool for brainstorming, conducting competitive research, assessing the novelty of drafts, and handling some of the more tedious, time-consuming tasks, such as writing alt text.
Keep in mind: if your company has spent decades building its reputation, average will undermine your brand and underwhelm your target audience—if they care to engage at all.
Want to optimize your publishing operations with AI the right way? I’d be happy to share what we’ve learned through hands-on experimentation.














