A behind-the-scenes look at how Aloudable uses AI to run its own content: building repeatable skills, finding topics from real buyer language, and keeping a human in charge.
By Lexie Meskouris
12 August 2026
We are an AI-native podcast agency, so it would be a bit odd if we did not use AI to run our own content. We do, fairly seriously. This is an honest look at how, less a showcase than a description of the actual, slightly repetitive work involved, because the interesting part is not the AI, it is the process around it.
The short version: we have built a set of reusable skills, standing instructions the AI follows, for the jobs we do again and again, and we feed them real, recent buyer language rather than asking a model to invent ideas. A person still makes the calls that matter. The result is more method than magic: the jobs we repeat, done consistently, with the judgement kept human.
We turned repeat tasks into reusable skills
Most content work is the same handful of jobs over and over: research a topic, plan a batch, draft in a particular voice, check that draft, write show notes. Rather than prompt from scratch each time, we wrote each of those as a skill, a detailed set of instructions the AI loads when we ask for that job. It means the hundredth blog plan follows the same thinking as the first, and improvements we learn get written back into the instructions instead of living in someone's head. It is unglamorous and it is the whole point.
We start from real buyer language, not a model's imagination
A model asked to suggest blog topics returns bland, average ideas, because that is what it has. So we do not ask it to invent. We pull what people are actually saying right now, the questions and complaints people are raising in recent posts and threads across the web, and use that as the raw input for topics. The AI helps gather and organise it; the signal comes from real people, not the model. It keeps the content anchored to problems buyers actually have this month, not evergreen guesses.
The voice is the hard part, so we guard it
The thing that breaks first with AI writing is voice: everything drifts towards the same competent, anonymous tone. We treat voice as the main quality problem. We built separate skills for each of our voices, grounded in real examples of how each person actually writes, and a stricter checker skill whose only job is to catch drift, AI tics, and anything that sounds like a costume rather than a person. Nothing goes out until it passes that. When we started, we deliberately kept the volume low to get the voice right before scaling it up.
A person still makes the decisions
The AI does the gathering, the drafting, and the first checking. It does not decide what we publish, which claims we stand behind, or when something is not true to us. Those calls stay with a person, and the most useful moments are often when one of us pushes back and says a draft's central claim is wrong, or would not survive a client reading it. The system speeds up the work around the judgement; it does not replace the judgement.
What we would not claim
A few honest caveats, since this is easy to oversell. It is not hands-off; it saves time on production, not on thinking, and the thinking is still the expensive part. It took real work to build and still needs tending. And it only works because a person who cares about the output stays in the loop; pointed at a team that will publish whatever comes out, the same setup would just produce mediocre content faster. The tools are not the clever bit. The discipline around them is.
In short
Our content runs on AI, but not in the way the phrase usually implies. We built reusable skills for the jobs we repeat, we feed them what buyers are actually saying rather than asking a model to guess, we guard voice obsessively, and we keep a person in charge of every decision that matters. It is deliberate, repetitive, and human-led, which is exactly why it works.
This is the same kind of thinking we bring to clients' podcasts: real process, AI where it helps, judgement kept human. If that is how you would want your own content handled, let's talk.
