# Show Me Your AI — Episode 12 Full transcript: How To Build An Agentic Workflow For Content Production. --- *Cold open.* *"Coming at you with some fresh content."* *"Is your content past its expiration date?"* *"We are flowing today."* *"Make your content fresh. Like me."* --- **DAN:** Welcome to Show Me Your AI with Dan Gilbert. Today we're talking about workflows, specifically a workflow for content production. Casey, Jamie or AI? **CASEY:** Yeah, for sure. So today I'm sure that we're all familiar with what workflows used to look like. I don't know if you want to run through this. ## The old workflow **DAN:** Yes. I mean, it's different in every company, but let's just say that you want to produce some new content. Someone spots a gap. The CMO says, "We're not showing up on ChatGPT." They email the team. The team emails the agency. They produce some analysis. They brief the SEO agency. They make some content, do a few drafts of it, send it back to the internal team, rewrite the content, and so on and so forth. Obviously there's a thousand different variations of this in different companies, especially with where the LLMs are ranking. They're looking at freshness and volume of content. So you rank by producing lots of content quite quickly. You can't produce one piece of content a month. Show us how we can do that a bit better, Casey. ## The new workflow **CASEY:** Yeah, for sure. So now we're going to run through what a workflow actually looks like in the age of AI, and how this can be immediately leveled up. In this example we can see a Notion task is created: "I want to do my weekly content refresh." And we're also showing an example where you go into Claude, but this is very flexible and can be done anywhere. That could have been automatically kicked off, and I don't actually even need to go into Claude. I'm just showing you the most thorough example taking off. And so I go in for this fake brand, Pipperwick. I want to run a content refresh for them. First it confirms that setup and picks the article that should be refreshed, which we can see here. Then it pulls the keyword cluster and volume, and it's automatically linking with Ahrefs, so we can automatically pull all of the keyword rankings. Then it's making the brief and topics. We're able to get the headings and everything for the content refresh, and this is all being done. I don't actually need to touch my computer at all. Then we're making the content. Here we can see the updated content that has been created in 30 minutes instead of 30 days. It actually also wrote a video pack, so we're automatically creating a video in this workflow as well. --- *I hope you're enjoying this video. Every like, subscribe or comment adds to Danbot Gilbot's training data and does its own part in helping Danbot Gilbot take over the world. Thank you. Now back to the video.* --- **CASEY:** And then finally, I'm approving in here. It's also being sent to my Slack, where I can approve it as well, so there's lots of flexibility on where you can actually approve this content. And finally we have it hooked up to Vercel. So automatically, when I publish this to Vercel, you can see that it included the video it made, and here is the updated content refresh. ## What changed **DAN:** Yeah. Before, maybe this took six weeks and a million steps that have been forgotten, and this took 30 minutes to run in Claude. It could have been done entirely autonomously as well, but you can also choose whichever steps you want to have a human in the loop to approve. Magic. **CASEY:** Well, it's not really magic at all, is it? There's loads of engineering that goes behind all of this being possible, including connections with the various parts that need to happen, specific instructions around things like the briefing documents, connectors into the actual web team, into Slack for approval, and so on and so forth. We cover a lot of that in the episode we did on how to set up workflows. But of course, across all of the different episodes that we've recorded, there's tons of engineering that goes into making this possible. **DAN:** The example that you gave, if it wasn't already obvious, and I think you mentioned it already, is that you can adjust any step of this. Humans can do some parts, and you can plug in different agents to do other parts. So on that research piece, we showed Ahrefs as one of the sources, where we have an MCP, an excellent tool. You could plug into Profound, or it could be a manual research process. It doesn't have to kick off in Slack. It can be Notion. You can create the content manually. You can send it to an agency. You can use agents to do that for you. When you're reviewing it, it can come up in a web app. It can be a pure human review. You can go to Slack. And then on publication, it doesn't have to be that we deployed to Vercel. There are tons of different ways that you can deploy properties to your CMS. So, interesting. And I think the coolest part, which we mentioned in the other episode, is that it's not necessarily different work, it's the same work that you can do way quicker, which then I think unbounds creativity. People think it's going to replace creativity. Sorry, it's a talking point, but it's quite the opposite. You can be even more creative. Well done, Pipperwick, and good job, Pipperwick, for making all that content. ## Build it yourself **DAN:** If you want to build what we showed today, check out our other video on building agentic workflows, and point your LLM of choice at this video, which has an MD file that will tell your agent exactly what to do to build this for yourself. Thank you.