Listening Is a Skill AI Can’t Replicate

How storytelling became the most important thing I do with AI

Halfway through a client call last week, I stopped a woman mid-sentence to ask her a question she wasn't expecting.

She'd been telling me about a trip she took. Her family had gone for a friend's milestone celebration, and while they were there, she and her husband connected with a local organization that supports survivors of sexual violence.

And then my client did something she hadn't planned. She asked them about their dreams.

I interrupted: "What made you ask that question?"

She paused. Thought about it. Said she didn't want to just sit with the reporting. She wanted to shift the room. She knew she had the capacity to create some kind of transformation, even a small one, even in that moment, and asking about dreams was how she did it.

I said: "That's the thread." That's not something AI can do. AI is skilled at pattern recognition, yes, but it won’t stop you mid stream to pull a thread when you’re still in the life of the story you’re telling. It might be able to do it after you’re done, but then how easily will you be able to get into that flow?

And this premise is exactly why the way I work with AI is both different, and more impactful for most people.

I work at the intersection of two skills that most people treat as entirely separate. I'm a storyteller, which means I know how to sit with someone, create safety, and pull the specific, sensory, too-human-to-fabricate details out of their experience. And I use AI, which means I can take those details and build from them at a speed and scale that would have been impossible five years ago.

The combination is simple to describe and difficult to replicate, because the value isn't in either skill alone. It's in the sequence.

Here's what I mean.

Before this call I’m talking about, my client had already tried AI. She'd pasted her interview questions into ChatGPT and gotten back perfectly structured answers. Her own notes next to them read: "Chat crap." "Bogus." She could feel that the responses were crap. She just didn't have the language to explain why, and she didn't know how to fix it.

The problem was that the input she gave ChatGPT could never give her the output she wanted. She gave it questions and expected answers. She got answers. Clean, professional, interchangeable answers that could have belonged to any woman entering any contest for any cause.

She needed to give it the swimming pool. Let me explain the swimming pool.

During our call, I asked about a song she'd written. She told me about how she had written it at a time she felt creatively blocked. Then one day, in the pool, a song started coming through her. She almost went right past that moment, but I leaned in and motioned with my hands for her to keep going. I didn’t want to interrupt her flow (pun intended). What came out because she received that visible cue from me was something ChatGPT never could have pulled out of her.

That’s because AI doesn't shut up. It generates. That's its job. But generating from thin input produces thin output. Always.

After the call, I sat with the transcript and did something I've done hundreds of times as a writer and editor, but never could do at this speed before AI.

I mapped the story architecture before going to AI. I recorded a separate transcript describing my own reflections and things that felt important to me from the conversation. Where were the convergence points? Where were the through-lines? What moments stopped me in my tracks that I hadn’t gotten a chance to pull the thread on in a 60 minute call.

Then I gave all of it all to AI with an idea of the types of deliverables I wanted and it gave me things that if done manually would have taken me two days of research and a full day of drafting. And the output would have been good. But it wouldn't have been better. It would have been the same quality at three times the cost and five times the timeline.

I think a lot of people in the creative and consulting world are operating with one of two assumptions, and both are wrong.

The first: AI replaces the human skill. It does not. It replaces the labor. The competitive analysis you used to do by reading 30 articles over a weekend. It replaces the things that were always just work, not art.

The second: the human skill makes AI unnecessary. It does not. The human skill, alone, is slower than the market now demands. A client who needs contest answers, magazine responses, a private editorial analysis, social campaign strategy, and follow-up questions... that's a week of work for a solo practitioner. It's a day when the extraction is human and the amplification is machine.

The diagnostic is human. The extraction is human. The synthesis is human. The amplification is machine. The calibration is human again.

My client ended our call by telling me she felt safe telling me her stories. That I have a gift for listening and pulling things out.

I do. But the gift isn't just in the pulling. It's in knowing what to do with what comes out.

Five years ago, I would have taken that hour of raw material and spent a week turning it into deliverables. Beautiful deliverables. Handcrafted. Slow.

Now the listening goes further. The stories reach more surfaces. The turnaround makes it possible to serve people who couldn't have afforded a week of my time.

The AI didn't replace the listening. It gave the listening somewhere to go.

That's the part I wish more people understood. We all know that AI is powerful. But that the thing which makes it powerful, the thing that separates extraordinary output from chat crap, is still the same thing it's always been.

Someone has to sit in the room and ask the right question and then someone has to hear the answer.