I was scrolling through X and saw yet another post marketing a $5,000 course about how to use Veo3 to replace your video team. The kinds of courses you saw 6 months ago for N8N automations (and before that, Notion templates).

Veo3 came out in May and is Google’s newest video generation model. Using it is like working with a social media manager that drops 50 “viral” posts per hour at the cost of a Starbucks latte but they’re all about different kinds of cheese… but  you don’t sell cheese.

Sure, Veo3 is the best video model on the market right now. But having spent 10 hours on it, I’m not convinced. 

Here’s what I tested:

Verdict: Veo3 won't replace your video team. But it could actually be useful for B-roll and filler content. At times I’ve also seen it work well for totally fake hooks.

Think: animating a static product shot, creating background footage for existing videos, or basic 3-second clips you'd normally source from stock footage. The key word is augment, not replace.

Case in point: Shane Heath (CEO of MUD/WTR) used AI to create an example ad for his brand. Even a talented creative had to combine his own product footage with 5 different tools to make it work.

The result was better than I expected:

But…it also felt disjointed and not-quite-right. Kind of like walking into a Walgreens at 11pm.

I predict 12 months from now Veo3 (or a different model) will be a meaningful part of your workflow. Here’s why:

Bottom line: Still closer to overhyped, but at the pace AI video is moving, put it in the lab. Just don't waste $5K on courses promising to replace your entire video team.

Picture this: It's 9pm. I'm sitting in my office, literally arguing with an AI about how to close a popup while my wife asks from the other room if I'm okay. 

"Yeah honey, just teaching a robot how to defeat popups so it can buy me a wallet!" 

My wife:

Welcome to agentic shopping.

In this case, via Perplexity’s new and agentic Comet browser, that can do anything a (cognitively impaired, at times) human can do online.

I had Comet:

The last one broke everything.

You’d think finding me the best wallet from Ridge.com with a $100 budget would be simple, right?

Wrong.

First attempt: Ridge's website broke the agent before it could even start shopping. The moment Comet hit their email popup, game over. It couldn't figure out how to close it. (Can you blame it? It takes me forever to figure out how to close them, too.)

Second attempt: After some coaching, I finally got Comet past the popup (though it still took forever). It then researched the products, looked for sales, looked at reviews, found the best product, and added it to the cart so it was ready for me to checkout.

Here's what hit me: Ridge's website nearly lost a sale to a newsletter signup form. If this was a real buying situation, it might not have gotten a second chance with the agentic shopper.

We’re at a point now 60% of US searches are AI-powered. Agentic shopping isn’t some far off future that may or may not happen, like your friend’s crypto gains. It’s actively happening right now. The same way ChatGPT now has 800M daily active users.

Ridge’s website wasn’t designed for AI agents. Neither is yours, I imagine. But the brands that optimize for this first will start to capture more sales both from LLM search, and from agentic shopping as it comes.

I wouldn’t invest in this today. But I would absolutely spend 20 minutes with Comet this weekend. It will give you real experience of a new way to shop that’s coming.

And… that’s it for this week’s edition. AI video isn't ready to replace your team, but AI shopping is about to change everything. Start thinking about how agents will interact with your site. Because ready or not, they're coming. And for the love of all that is holy, fix your popups before the robots revolt.