AI Works Best When the “Human in the Loop” Commits to Nonstop Error-Testing

I recently published an article about taking on writing assignments for marketing projects in 2011. Getting a content-heavy project assignment on Friday that’s due on Monday was the beginning of what would be a labor-intensive weekend, with notebooks, multiple printouts, insane amounts of coffee, and email requests begging for references.

Fifteen years later, content creation is radically different. Timelines are accelerated; research options are fast, vast, and literally served up to us on a golden digital platter; and there are new platforms that address every workflow need you could possibly have.

My Typical Workflow for Marketing Content Creation

My typical workflow combines Claude (Anthropic); Gemini (Google); Napkin, NotebookLM (Google); Opal (Google), and sometimes ChatGPT (OpenAI). I use Claude to search and validate references; Notebook to teach myself what I’m working on; Gemini for vetting and optimizing content and generating images; ChatGPT for general questions and certain tasks; and Opal if I need to set up an agent to handle the same task repeatedly. Napkin is for quick generation of conceptual graphics and background guidance.

I then cross-validate across platforms and iterate repeatedly. At one point, I walk away from AI and move into 100% Human-in-the-Loop mode. This is where expertise and experience are invaluable.

Certainly, the process is less time-consuming, but not any less intellectually challenging if your goal is to create something of value, versus AI slop that is obvious, unredacted, and ultimately lazy.

AI is Not a Substitute for Not Knowing

Is it really that easy?  Has AI made it easier to create content that you don’t understand?

Does it come down to using great prompts to coax out polished content (or code) even if you don’t understand the larger context of what you’re interrogating for?

The answer is an emphatic “no.” There is no deus ex machina moment with AI where everything comes out clean and perfect.

AI makes processes easier, faster, and much more far-reaching, but it does not eliminate the need to understand what you’re working on.

In fact, using AI increases the need to be discerning and to double- and triple-check output.  Each missed mistake that stays in the workflow can easily compound and compromise the reliability of your work. We all have access to the same tools. It comes down to using them strategically, with an appreciation for the productivity boost and extreme caution when it comes to error-testing AI’s output and ensuring full integrity in your utilization and disclosure.