I replaced my old job with an AI agent

Hello friends. Today I want to talk about agentic media buying. A few points why you might want to consider my perspective. I've been media buying across multiple channels, specifically Meta ads, for over a decade. Previous to that, I had a career in cybersecurity protecting bank and government infrastructure from foreign intrusion. Now I build my own products and work with clients that are looking to grow their businesses. I've been following the trends with AI for a while now. I was a beta tester for GPT-3 several years before ChatGPT was released to the public, and to the best of my ability, I've tested most of the tools and underlying models that have come out over the last five or six years. Over the past month, there was a big release with the official Meta Ads MCP. Between that and the coding abilities provided by GPT-5.5 Codex, I've been able to build out an agent that can essentially replace the role of a media buyer. Let me dig into that. Before AI, the media buying feedback loop looked like this. You would need to do deep research on the product, competitors, the potential customer personas. You might need to put together a brief and some ideas for what some ads might look like. You or someone else on the team would need to write copy. You would need to have images and or videos created. Then all of that would need to be sent to the media buyer to assemble in Ads Manager. In addition to that, there was a significant amount of backend setup that needs to be done. Unless you are a developer, the chances of this being set up correctly can be challenging. For example, let's say you need new images. That used to take anywhere from a couple days if you have an in-house team to perhaps a week. Same thing with copy. Either you as the founder or media buyer have those additional skills, or you need to wait. If you do have the skills, you have to spend your own time. The other thing would be reporting. You'd have to go into ads manager and sort through sometimes hundreds of thousands of dollars worth of data, thousands of transactions, to understand what are your best ads, why are they your best ads. Then you would have to take that information and rerun the process of creating copy, creating new images, angles etc. That time and effort in terms of raw costs adds up quickly. Agents have been an idea for several years now, and what would happen with the earlier versions of agents, and I don't mean this just with media buying agents but in general, and I'm sure those of you that have used AI deeply will understand this, the agents would break. You would generate an image, you would try to iterate on the image, and at some point it would be easier to go into Canva and just do it yourself. But now we've reached a point where the agents work effectively. It's faster to iterate. At least for now, and we have to take into account that the tokens we're spending are heavily subsidized so that these AI companies can acquire customers. In the long run, we'll have to see how the costs relate to reality. But at least for now, I can run an entire Meta ads account for less than $20 a month. The other thing I've seen over the past 5 years is a lot of resistance to these technologies. I remember posting on Reddit and saying something about copywriting, and there was a lot of copywriters that jumped in and said they couldn't be replaced. Now it's the same thing with developers. Don't get me wrong, development, especially right now, is a superpower. But I think it's a bit naive to think that AI, aren't going to take a chunk out of the margins out of certain types of businesses (especially freelancers) With this media buying agent that I put together, I can run reports, I can do in-depth analysis, and I can do that report in just a few minutes where it would have taken me or another media buyer a couple hours. Same thing with creating ads. Now I can create 30 ads and have them deployed into the account in the same amount of time that it would have taken me to create one ad previously. That includes the images, includes the copy. It includes research directly in Facebook Ads Library. It can determine who potential personas are, and then create images and copy specific to those personas. The next layer of this is creating landing pages that are congruent with those persona based ads. What I can say is, it works. I can message this agent from Telegram. I can check my stats. I can update a dashboard, and at least so far, it's incredibly accurate. Another thing I have it doing is UTM and accurate naming conventions. If you've run Facebook ads before, if you've ever audited a Facebook account, you'll see a lot of dash copies. If you're doing any sort of significant granular tracking and passing those as UTMs, it can get ugly pretty quick in your analytics. I can have the agent go back and rename campaigns, I can have it place UTMs, I can have it pull reports based on what I want to see, and then I can ask it questions like I would ask a media buyer. For me, it's interesting to see an agent that can finally do what I've been doing. It doesn't mean that it replaces me completely as a human. But it does allow me to operate from my zone of genius, which is creative strategy. I dare say it's made media buying and building shit more fun for me. The disclosure is, I have long experienced anxiety and pretty challenging ADHD, and at times execution felt overwhelming. Now it feels relatively effortless. How hard is it to talk to an agent? I know that there are a lot of skeptical people, and I wanted to prove that this stuff actually works in real life not just in theory. One of my clients, pretty much everything we deployed this year was AI. The landing page was AI. The ads are AI. The copy is AI. It's converting. It's converting into sales. We've generated approximately $65,000 in sales in the last couple months starting from scratch. What a time to be alive. There's definitely a part of me that thought we would get to a point where the agents wouldn't continue to get better. But now I'm seeing the feedback loops speeding up, for lack of a better word. This is something I've been thinking about for 5 years. I remember reading a paper a couple years ago about AI's impact on business feedback loops but the impact didn't land until this latest iteration of tools. The big takeaway is this. You don't have to use AI. But you are going to have to compete against people that are using AI. If you're producing 35 ads a week with the old feedback loop versus somebody that can deploy 30, 50, 100 ads in a day, one thing is the speed, but then also the cost. If we add up all the costs of human produced assets and we put that same person against somebody who's leveraging AI, all of the additional margin that they save on time can be applied to buying your customers. I'm not saying this is good. I'm just saying that this is, to the best of my ability, accurate and true about what I'm seeing in real time. I want people to be able to make accurate decisions for their life. If you have any questions about this topic feel free to reach out and if you have already built your own agent lets trade notes. Cheers, Mike