Three Ways Agency Leaders Respond to AI Workflows. Only One Works.
By Gabriel Tan | March 2026
Three Responses to AI. Only One Leads Anywhere.
I have shown the same AI press release workflow to a dozen agency leaders. Three responses. Only one leads anywhere.
But first, some context.
Sahil Bloom recently wrote about what he calls the Empty Cup Mindset. He tells the story of Dr. Ignaz Semmelweis, a 19th-century physician in Vienna who discovered that doctors moving straight from autopsies to delivering babies, without washing their hands, were causing an 18% maternal mortality rate. The midwife clinic next door had a 2% rate.
Semmelweis mandated handwashing. The mortality rate dropped immediately to below 2%. The evidence was overwhelming. The results were undeniable.
The medical community ridiculed him. He was dismissed from the hospital and eventually committed to an asylum, where he died at 47.
The doctors who rejected Semmelweis were not stupid. They were accomplished experts. That was the problem. Accepting his evidence meant admitting that their established practices were flawed. There was no room in their cup for that.
Bloom references an old Zen parable where a master keeps pouring tea into a full cup until it overflows. “Come back to me with an empty cup,” the master says. The point: the more you know, the harder it is to accept that some of what you know may no longer be true.
I have been thinking about this in the context of PR agencies and AI.
The workflow I have been showing agency leaders is specific: a shared Claude project with the client’s brand voice document, an anti-AI writing guide, best-in-class samples, a structured intake form, and a QA checklist. The output matches the client’s tone. It does not sound like AI. The senior’s role shifts from rewriting to sign-off.
Here are the three responses.
The first group says “AI cannot write a good press release” and closes the conversation. Their cup is full. They have decided what AI can and cannot do based on an experience they had 18 months ago, or based on a single bad output from ChatGPT. The technology has moved on. Their conclusion has not.
The second group says “we are already using AI for press releases” and assumes there is nothing left to learn. When I ask how they are using it, the answer is usually: someone on the team opens a chat window, types a prompt, and edits whatever comes back. No brand voice control. No anti-AI writing layer. No structured intake. No QA gate. The tool is in use. The system is not.
The third group says “show me how you do it, and teach me.” They are curious. They do not assume they already know the answer. They are willing to sit with the possibility that their current process, even if it works, might not be the best version of itself. Their cup has room.
The leaders in that third group are the ones whose agencies will look different in 12 months. They approached AI with enough openness to build a proper system around it, and that made the difference.
This is not unique to press releases. The same pattern shows up when I walk through monitoring workflows, pitching systems, and client reporting. The leaders who learn fastest are not the ones with the most experience. They are the ones who hold their experience lightly enough to let new evidence in.
Bloom puts it well: the goal is not to be right. It is to find what works.
AI inside a governed workflow produces stronger first drafts in less time. But only if the person building the system is willing to start with an empty cup.
The question for any agency leader reading this: when was the last time you sat down with your team’s AI workflow and genuinely asked whether there is a better way to do it? To find out what you might be missing, rather than to confirm what you already believe.
If the answer is never, your cup might be full.
