The Future of Paid Media: Why AI Needs Human Guardrails
Chelsey Corbey-West
Paid Media Lead
If you’ve opened your Google Ads or Meta Business Manager lately, you’ve likely noticed that the “manual” buttons are disappearing. They’re being replaced by “Smart” this and “Advantage+” that. The platforms are essentially asking for your credit card and your trust, promising that their algorithms will find your customers for you.
The platforms are becoming more efficient at spending your money, but they aren’t necessarily becoming more efficient at growing your bottom line. If you hand over the keys to the AI without a clear set of guardrails, you aren’t “leveraging” (to use a word we hate) technology, you’re just gambling with it.
The common advice from platform reps is simple: “Consolidate your campaigns, use broad match, and let the machine learn.”
In our experience, following this blindly is the fastest way to tank your ROAS. When you give Google or Meta total control, they tend to go for the “low-hanging fruit” – retargeting people who were going to buy from you anyway. They’ll report a 10x return, but your actual bank account won’t feel the difference because those weren’t incremental sales.
The machine doesn’t understand your margins. It doesn’t know that you’re overstocked on one SKU and running low on another. It just wants to find a conversion at any cost. To win, your job isn’t to manage keywords; it’s to manage the data you feed the machine.
You might think Paid Media updates are just “back-end” technicalities, but they change the entire experience for your customer.
When AI takes over, creative becomes your primary lever. If the algorithm is doing the targeting, your ads are the only thing left to do the “filtering”. If your creative is generic or “salesy”, the AI will serve it to everyone, leading to a disjointed experience where people see products they don’t care about.
A well-run AI campaign should feel invisible to the user. It should show them the right product at the exact moment they need it. When the tech and the creative align, the friction disappears. When they don’t, you’re just another annoying banner ad in their feed.
Interested in Brave but not quite ready to speak to us? Stay in the loop.
Google’s “Recommendations Score” is designed to make Google more money, not you. We’ve seen accounts with 100% optimisation scores that were wasting 30% of their budget on bad traffic. Instead of ticking every box the platform suggests, test them one by one. If “Broad Match” works for your brand, great. If it brings in irrelevant search terms, shut it down.
AI is only as good as the signal you give it. If you’re just tracking “Purchases”, you’re giving the AI a very blunt instrument. Start feeding it “Profit” data or “New Customer” signals. If you tell the algorithm to find people with a high Lifetime Value (LTV) rather than just anyone with a credit card, your long-term growth will look much healthier.
In the old days, we spent hours tweaking bids and match types. Today, we spend that time on creative strategy. Because the AI decides who sees the ad, your ad needs to speak very specifically to your ideal customer. If you want to sell to high-end interior designers, your visuals and copy need to act as a gatekeeper. If the ad looks cheap, the AI will find you “cheap” clicks.
You can no longer “set and forget” a campaign for six months. You need a testing roadmap. This means running split tests on landing pages, trying different video hooks every fortnight, and constantly questioning the platform’s reported results. We use “incrementality testing” to see what would happen if we turned the ads off. If the sales don’t drop, the ads weren’t working, no matter what the dashboard says.
Paid media runs on AI now; the edge comes from who is steering it. Platforms smooth the workflow so anyone can press “go”, and the output is a sea of forgettable campaigns.
If you want to cut through, you become the person who actually understands the maths inside the machine. You watch the numbers, treat automated recommendations as hypotheses rather than orders, and act on the fact that no model knows your brand like you do.
