When Terra interviewed for her role at Good Circle Marketing, she told me she preferred to write herself.
I was looking for people who were comfortable with AI. That answer could have ended things. Instead, it had a lot to do with why I hired her.
Terra manages the strategy side of our marketing work. She’s also the human who checks our AI output before it goes anywhere. And if you asked her today, she’d still call herself anti-AI. She monitors the environmental conversation. She read about DeepSeek and pointed out they could have done this whole thing with 17 times less energy if they’d waited a year. She pushes back on me regularly.
She is one of the most valuable people on our team.
I sat down with her for a new episode of A Good Pour’s Summer of Good AI series because I think what we’ve built together is a real-world example of something most businesses are still trying to figure out: what does it actually look like when a skeptic and a true believer have to work it out?
What a skeptic actually brings
There’s a version of AI adoption where enthusiasm is universal. Everyone’s prompting. Output is moving fast. Nobody’s questioning it.
That’s not how we work. And I think our work is better for it.
Terra catches things. A factual error in client content nearly went out recently. The information had been accurate when we first built the skill, but something had changed. She flagged it before it got anywhere. Not because she immediately knew the right answer, but because something felt off and she said so.
That’s not something AI does. AI doesn’t feel off about things.
She said it plainly in our conversation: “Experts have accountability that AI doesn’t have. If Claude lets something through, that’s Claude. He doesn’t get fired.”
The people doing the work carry responsibility for it. That accountability is part of what makes the output trustworthy. When we remove humans from the loop entirely, we lose that layer. And the errors don’t disappear. They just go out unchecked.
The conversation we had to have
Terra wanted to be honest on LinkedIn. She’d see posts about AI and want to comment. But she was nervous. She thought if I saw her being openly skeptical, I’d question why she was on the team.
So she came to me directly and said she didn’t like this. Could we work something out?
We talked through it and put some things in place. We don’t use AI for images because we want to protect artists and creators. We still hire videographers and photographers. She handles the human review at the end of every workflow.
That conversation made us better. Not because we resolved every tension, but because we named what we each cared about and built something around it.
If someone on your team is nervous about having that conversation, that might be the most important signal you’ve gotten all year.
The human-AI sandwich
I talk about this framework often, and Terra and I both use it. It’s not complicated.
Start with something real and human. A business owner telling their story on video. A conversation where someone says what they actually mean. AI comes in the middle, taking that authentic input and working with it. Then a human comes back at the end to check it, shape it, and make sure it sounds like a person made decisions about it.
Human. AI. Human.
Starting with AI and finishing with AI skips the part that matters most. That produces content that technically functions but doesn’t feel like anything. Terra described her role at the end: “I push back on it a lot. I’m like, that doesn’t make any sense. Nobody’s talking about that. Try again.”
That’s the job. The human at the end of the sandwich doesn’t just sign off. She interrogates.
What the skeptic protects
Something I didn’t fully anticipate when I hired Terra: having someone who questions AI constantly keeps the team more creative, not less.
There’s a pull in AI workflows toward letting the tool decide. It’s subtle. You get a draft, it’s fine, you move on. Over time, you’re accepting fine at a rate that slowly replaces the work of actually thinking.
When someone on the team keeps asking “do we really need to do that?” or “that was my human idea, let’s make sure it stays that way,” the team has to keep articulating why. The reasons don’t get lazy. The creativity doesn’t go quiet.
A built-in devil’s advocate is worth a lot right now. Not someone who’s combative, not someone who just says no. Someone who has a genuine point of view, stays informed, and pushes from a place of values.
What we can do while we wait for guardrails
National regulation on AI isn’t moving fast enough. I’ve had conversations with politicians about this. They’re trying. But this technology moves faster than legislation ever has, and we can’t afford to wait.
Terra made a comparison to the sustainability movement: when national policy lags, businesses have the opportunity to be more responsible before they’re forced to be.
So she sets her own limits. If AI doesn’t get something right after three tries, she does it herself. She knows what she’s good at, writing especially, and she protects that rather than letting AI take it over. She knows when a problem is above what AI should handle and calls in the right person.
Those aren’t policy guardrails. They’re personal ones. And right now, personal ones are the most practical place to start.
The version of AI worth building toward
We talked about IKEA in this episode. When they increased efficiency with customer bots, they didn’t fire 8,500 people. They retrained them. That opened a new billion-dollar revenue stream.
That’s the version of AI I want businesses to move toward. Not “how do we do this without humans,” but “what can we do with humans now that some of the other work is handled?”
I believe we were created to create. When AI takes on the parts of the work that don’t require our full attention and creativity, the question is what we do with what’s left. That’s the conversation most businesses aren’t having yet.
And honestly, your skeptic is probably the person most likely to ask it.
Try this today
Think about one person on your team who has been slower to adopt AI, pushed back on something, or asked an uncomfortable question about it.
Ask them this: “What would need to be true for you to feel good about how we’re using this?”
Write down what they say. Not to debate it. Just to understand what they’re trying to protect, and see if any of it belongs in how you work.
You might find your guardrails already exist. You just haven’t named them.
Listen to the Full Episode
You can hear the full conversation with Terra on A Good Pour wherever you listen to podcasts. It’s one of the most honest episodes in the Summer of Good AI series because it’s told from the inside, by someone who isn’t a convert and isn’t pretending to be.