Yes, we use AI heavily. No, we don't use it to write content. The distinction is the entire post.
This is the question we get on every first call. "Do you use AI?" The honest answer requires unpacking what "use" means, because the way we use AI is different from the way most content agencies use AI, and the difference is the most important thing about our methodology.
What AI is good at
AI is excellent at three things our workflow needs.
It's excellent at research summarization. Point it at a topic and it can produce a useful overview of the existing landscape in minutes. What's been written, what positions have been taken, where the debates sit. This compresses what used to be hours of background reading into something fast.
It's excellent at organizing transcripts. A 90-minute Expert Interview produces roughly 12,000 words of raw transcript. AI can pull out themes, group related anecdotes, and surface contradictions much faster than a human reading the transcript linearly.
It's excellent at surfacing adjacent angles. If we have a framework an executive has built, AI can quickly identify the related ideas in their field. What concepts are nearby, what counterarguments exist, what's been said and not said. This is research-grade work, and it's research that used to take a day and now takes an hour.
All of this is AI as a research assistant. None of it is AI as a writer.
What AI is bad at
AI is bad at producing content that sounds like a specific person with a specific point of view.
The reason is structural, not skill-based. Large language models are probabilistic text generators trained on vast amounts of text. When prompted to write about a topic, they gravitate toward the consensus position, the average of what's been said about the topic across the training data. That's not a flaw to be engineered around. It's how the technology works.
This works against everything content programs are trying to achieve. The value of a senior executive's content is the position. The position is, by definition, not the consensus. It's what the executive thinks that not everyone else thinks. The moment AI is in the loop generating final content, the position gets averaged out. What survives is the consensus version of the executive's idea, which is exactly the version that doesn't differentiate them from any other voice in their feed.
There's a second issue. AI-generated content carries the patterns of AI-generated content. The parallel-structure cadence, the certain phrases that show up everywhere. Sophisticated readers can spot it. They downgrade the source. A 2025 study found that 42% of readers trust AI-generated content versus 68% for content known to be human-authored. The trust gap is real and it's growing.
How we actually work
Our workflow uses AI at three points. None of them involve generating final content.
At the research stage, before writing begins, we use AI to map the territory around the executive's positions. What's been written? What angles haven't been covered? What's the prevailing view we'll be pushing against?
At the structuring stage, after the Expert Interview, we use AI to help organize the transcript. Themes, recurring threads, examples that connect.
At the editing stage, we sometimes use AI to flag issues. Passive voice, redundancy, places where the argument flow breaks. The writer decides whether to act on the flags.
At every stage, the writing itself is human. The position-taking is human. The voice calibration is human. Every word that ships under a client's name is written by a senior writer and edited by a senior editor.
What the data says
The argument for human-led writing isn't just about audience trust. It's also about output quality.
A 2025 Graphite study found that AI-assisted workflows with strong human oversight perform 4x better than fully automated AI workflows on actual content performance metrics. The difference isn't about whether AI is involved. It's about whether humans are still doing the judgment work. When humans are doing the judgment and AI is doing the research, the work performs. When AI is doing both, the work converges to the mean.
We agree with the data. So do the executives we write for.
The question to ask any agency
If you're evaluating a content agency in 2026, ask the AI question directly. Ask how AI is used at each stage of the workflow. Ask whether final drafts are AI-generated or human-written. Ask whether they have a written policy.
Watch for hedging. Hedging usually means more AI is involved than the agency is comfortable saying out loud. Our position is on the record because we've made the choice deliberately. Most agencies haven't.