These people made an internal podcast. AI helped them reach the whole team
Published on October 01, 2025
Amanda Cupido doesn’t speak Spanish or French, but she and her team made a podcast in both languages all the same. Amanda is an audio producer and the founder of a production company called Lead Podcasting. One of her clients is a global nonprofit with over 35,000 employees—and not all of them speak English. So she made them a pitch: what if they added AI into the mix?
Amanda had already been experimenting with generative AI voice tools when she pitched her client on using the tech to make their internal podcast more accessible to as many employees as possible. They would make the podcast in English, and then use AI to translate it into Spanish and French—with a lot of human oversight, of course. Driven by a desire to use these tools for good, the goal was never to replace people, but to reach more people, and it worked.
On season two of the Dropbox podcast Working Smarter, Amanda shows what it’s like—and what it sounds like—to make a podcast with AI that’s still human at its core.
You can read an excerpt of our conversation with Amanda below.
~ ~ ~
This was your first time pitching generative AI for a podcast. Did you know what you were getting into?
No, no. I was just going with my gut. Sometimes you say AI and, depending who you pitch it to, they're very against it. And I can respect that. But as soon as I said the words, I didn't see a grimace. I saw them lean in, like “Oh, tell me more.” And so I started just saying what I knew, which was: There's technology that can make this much easier. The workflow is going to be instantaneous. We still have local language producers on the team overseeing it and ensuring that the quality is up to par. And the client said, “All right. Get me a price list and a formal estimate, and we'll see.”
So you've pitched this idea to the client. You have this tool. You’re ready to create an AI-generated version of the podcast in a couple of different languages. How does this process actually play out?
We’re still iterating as we speak. You can choose one of two ways to generate a voice. You can either create a new voice profile using the guest’s audio, or you can go to a database of voices and choose a pre-established voice that sounds similar to the guest. In the beginning we tried generating new voice profiles using the guest’s audio, but the translations were not to the level we needed them to be. What we realized was we needed to have a step in-between—so first to tweak the translation, and then create the voice. But by episode two, we decided it was actually faster and better to use the database voices instead of generating a new voice profile. There was more consistency, flexibility, and creative control using a pre-established voice.
Dropbox Dash: Find anything. Protect everything.Hearing you describe all of the steps involved, was there ever a point in this process where you or your team were banging your heads against the wall and almost regretting your decision to pitch this in the first place?
Definitely. Yeah, the whole time. The reality is, I didn't account for the amount of time it was going to take to fix the translation. And there were definitely times when we were frustrated trying to build the new voice profiles—hence why we ended up switching tactics. But I gave real power to the team. I just said, “Go with your gut. Do what you think sounds best.” They were getting nuanced with the accents—is it a Colombian or Venezuelan Spanish accent?—to make sure we were honoring it as closely as possible. And they were able to do that better with the pre-established ones. So I thought, okay, no-brainer.
It sounds like your process still required a significant amount of human involvement. Is that correct?
For sure. Especially when it comes to the nuance of translation, language, and accent. That is deeply personal, deeply human, and that human oversight was important. And so that's why the key point-people on my team who were running the show before [we adopted AI voices] were all still there doing their thing. We just changed what their job involved. And once we got over the learning curve, the workflow was more seamless. But definitely lots of human fingerprints all over it.
Why did you feel it was important to maintain that degree of human involvement in the process?
From a content agency perspective, it's hugely important to still have that human oversight. That's what allows us to create excellent content. And so I don't ever see that going away. We're always going to have that human oversight to ensure we’re getting what we want from this tool. It ended up being so good that when we delivered the first episode, the client said, “Oh, but I thought you were using AI for these?” She thought it was human. To me, that means it was a success.
A tool alone isn’t going to have the instincts or the experience or all of the other knowledge that you've brought from all of the other clients that you've worked with in the past. Is that what sets you apart from people who think they can just plug a transcript into a voice generator and make a podcast?
Exactly. You nailed it. Any person can go clickity-clack some buttons and pump out a whole thing and not have that level of oversight and care. You can't replicate that.
Where do you land on the usage of AI more broadly in the podcasting space?
I really believe it's going to help with globalizing podcasts. That's one of the good news stories that's going to come out of all of this. I think it's going to be a click of a button on the back end of your hosting site. You choose which languages you want to put it out in, and then it's out. And that's great because there's so many incredible Spanish podcasts that I don't understand—and vice-versa, English podcasts that others don't understand. Then we can see, like, what is the number one podcast? Truly, what is the number one? Because right now it's skewed. It's not accurate.
That is such a fascinating thing to think about.
Right? And then, from a production side, I think that simple editing is going to be made even simpler—like, the cleanup edit as we call it in the industry. Just taking out the sneezes, false starts, the rambling. There's so many people who want to make a podcast but don't have access to the tools, and that's going to be really easy.
Do you think that leaning too much on something like AI voice generation takes something away from podcasting, or from the people who work in this space?
If you were to say to the photographer who was in the dark room dipping their film in a bunch of different chemicals, “We got a digital camera for you,” and they're like, “No, it's ruining the art of the dark room”—we know how this story ends. We’ll look back on the time where we had to do all of this manually, just like we look back on literally cutting tape. But this is just the evolution of technology, and there will be a new art that emerges that I would encourage people to be open to exploring.
This interview has been edited and condensed. For more interviews and past episodes, visit workingsmarter.ai
The Author
.png/_jcr_content/renditions/hero_square%20(1).webp)
.png/_jcr_content/renditions/hero_wide%20(1).webp)
.png/_jcr_content/renditions/hero_square%20(3).webp)
.png/_jcr_content/renditions/blog%20(1).webp)
.png/_jcr_content/renditions/hero%20(1).webp)
.png/_jcr_content/renditions/hero_wide%20(1).webp)
.png/_jcr_content/renditions/1080x1080%20(1).webp)
.gif)
.png)
.png)
.png)
.jpg)
.jpg)
