What interests me most is how much faster AI video makes the testing loop. I’m currently experimenting with AI-generated video using GPT Image 2 for stills and Kling 3.0 motion control for movement. From what I’ve seen, that workflow gets better results than going fully AI-native from scratch. Pure AI video can still fall into the uncanny valley, especially with realistic human content, but animated styles, claymation, and stylized ad creative are already performing really well. I’ve been seeing a lot of that type of content work in paid social, especially on Facebook. The exciting part for me is not just scale. It’s control.

In a normal UGC play, if we want to test a new hook, outfit, pacing change, visual angle, or script variation, we have to brief the creator, wait for them to film, edit, post, and then analyze the result. AI video cuts out a lot of that delay. You can test small creative variables much faster and get a clearer read on what is actually moving performance. I don’t think AI replaces taste or strategy. If anything, it makes those more important. But it gives a social team way more shots on goal. You can generate variations, learn what works, and scale the winners much faster than a traditional creator-only workflow.

I stay organized by treating social like an analytics and systems problem, not just a posting task. I’ve worked in creator/social for 2+ years and have helped drive 600M+ views through creator content, so the biggest thing I’ve learned is that good social ops comes down to clear tracking: what formats are working, which hooks are getting retention, what posting windows are converting, and what should be scaled or killed. For day to day organization, I’d keep a central content calendar, use a cross-posting/scheduling tool like Postiz for anything that should go across platforms, and track each account by status: ideas, scripts, generated videos, scheduled posts, live posts, results, and next iterations. I’m also a big believer in using AI agents/MCP connections to keep the system updated automatically. For example, I use Hermes-style agents to monitor performance, summarize what’s working, and flag what needs attention so I’m not manually checking every platform all day. Social can look fluffy from the outside, but once you’re managing multiple pages, it becomes very analytical. AI is especially useful there because it can help spot patterns across hooks, formats, timing, and audience response faster than a person can manually. My role is to combine that with taste and experience: knowing when something is just a one-off spike versus when it’s a format worth scaling.

In June 2025, I led W’s influencer outreach program (@getw). We gifted around 250 creators, with 170 of them posting content. This generated over 11 million views and an estimated earned media value of $539.3K. The campaign’s primary goal was brand awareness, so our key success metric was reach rather than direct engagement - and we were able to achieve significant exposure through organic creator posts.