Meta Just Handed Ad Management to AI Agents. Here's What That Actually Means for Your Budget.

Meta integrated Manus AI into Ads Manager in February. More than 4 million advertisers now have access to an autonomous agent that can analyse campaigns, build reports and optimise spend on their behalf.

A month earlier, Threads ads went global. New inventory. New AI to push it. That timing tells you everything you need to know about who this was built for.

My first thought when I heard about it: Meta is not a platform that innovates for advertisers. The Ads Manager dashboard in 2025 was essentially the same one you used in 2015. Ten years of the same interface. So when they suddenly invest $2 billion in an AI agent that sits inside that dashboard, the question isn't "what can it do?" It's "why now, and for whom?"

Meta has a lot of inventory. Not all of it is good. Not everything is Reels. Not everything is Stories. Some of it is lead generation ads appearing in Facebook Marketplace because the algorithm decided it "made sense." It doesn't.

An AI agent that distributes spend across all of that inventory, automatically, is a way to monetise placements that most competent media buyers would never select manually. Scale that across 4 million advertisers and the maths works beautifully. For Meta.

To be fair, there's a reasonable argument on the other side. Meta sees millions of badly managed accounts. Advertisers who pick the wrong conversion event, set nonsensical targets and burn money through their own incompetence. An AI agent that raises the floor for those accounts is a genuine improvement. If you're currently running campaigns with no idea what you're doing, a robot will probably do it better than you.

The problem is that the same system also applies to advertisers who do know what they're doing. And for them, it's not raising the floor. It's lowering the ceiling by spreading spend into placements they'd never choose and reducing the visibility they need to make sharp decisions.

Meta's median return on ad spend sits around 1.93 to 2.19x depending on which study you look at. Google Ads averages 4 to 8x. That's not a marginal difference. That's a different category of performance.

And it's not just the top-line ROAS. The quality of everything underneath it is worse. The conversion tracking is less reliable. The traffic quality is lower. The attribution is murkier. You pay less per click on Meta, sure. But you get what you pay for.

I worked with an advertiser spending around $120,000 a month. Roughly $90,000 of that was on Meta and $30,000 on Google. We cut Meta to $25,000 and moved the balance into Google. Same sales outcome. The business had been struggling with low quality leads and, surprise, that problem correlated directly with the volume of Meta spend. When the Facebook coverage went up, the lead quality went down. Cut it back and the pipeline cleaned up.

That's one account. It's not a universal truth. Meta's Advantage+ shopping campaigns have genuinely outperformed manual setups for a lot of ecommerce brands. If you're selling a product people buy on impulse and your creative is strong, Meta can work. The platform knows how to find buyers in a scroll. But the idea that Meta is a high-performing platform by default, across all business types, is something the data doesn't support. And it's a particularly dangerous assumption for lead generation businesses where the quality of the enquiry matters as much as the volume.

Let me be clear about something. Meta's AI agent will almost certainly understand patterns better than a human media buyer. It will process more data, react faster and find optimisation opportunities that a person sitting in front of a dashboard would miss. That part of the pitch is correct.

But the question was never "can AI optimise better than a human?" The question is: optimise for whom?

Your job as a media buyer, whether you're doing it yourself or paying someone, is to be the advocate for your business in a relationship with the advertising platform. The platform wants you to spend more. You want the best outcome from the least spend. Those interests overlap sometimes and conflict often. A good media buyer holds the tension between those two things. An AI agent built by the platform doesn't.

Trusting that a multi-trillion dollar company has built an AI agent with your interests at heart is, I'll be polite, a bit ambitious. And this is a company whose track record on protecting user data isn't exactly spotless. They built a better salesperson. An incredibly effective one that works 24 hours a day across 4 million accounts and never pushes back on expanding placements or increasing budgets. If I was Meta, I'd have done the same thing. It's a smart business move. Just don't confuse it with a favour.

Threads ads went global in January. Manus AI rolled into Ads Manager in February. Threads has over 400 million monthly active users and a brand new advertising surface that needs to generate revenue.

An AI agent that automatically expands your campaigns across placements is the perfect distribution mechanism for a platform that just opened up a massive new channel. You probably won't notice Threads appearing in your placement mix. The AI will handle that for you. How convenient.

There's also a go-to-market logic here that makes sense. Roll it out to smaller self-serve advertisers first, where Meta provides almost no account management anyway. In Australia, you need to be spending around $8 million a year before you get a dedicated rep, and even then they're based out of Singapore. I've been trying to do something as basic as transferring a credit line between entities for a client. Four months. Still not resolved. That's the level of human support you're working with. For most Australian advertisers, the AI agent isn't supplementing a rep. It's replacing the absence of one.

The changes are coming regardless. You're not going to stop Meta from rolling this out. So the practical move is to understand exactly where you stand before the AI starts making decisions on your behalf.

Know what you're spending on Meta right now. Know what you're getting back. Not the numbers Meta's dashboard tells you, but the actual business outcomes. Revenue. Lead quality. Customer acquisition cost measured against lifetime value. If you can't answer those questions today, an AI agent isn't going to fix that. It's going to obscure it.

Make sure you have the mechanisms to identify whether automation is improving performance or just redistributing spend into places you can't see. Set your benchmarks before the agent starts running. Because once it's optimising, you need a baseline to judge it against.

And diversify. Meta is one channel. Google is another. We've seen TikTok outperform Meta in several recent campaigns. That's anecdotal, not a trend. But it's a reminder that there are multiple places to buy attention, and the businesses that get hurt by platform-level AI aren't the ones who use it. They're the ones who depend on a single platform and don't have the measurement in place to know when it stops working.

The AI isn't the problem. The problem is the same one it's always been: not knowing your numbers well enough to tell whether someone is helping you or selling to you.