← Blog / AI
AI did not replace media buyers. It replaced their dashboards.
30 June 2026 · 2 min read · by Obscale
LOG / ai-and-media-buying
operating notes
Every year since 2023 someone declares the media buyer dead. Meanwhile our paid team spends more than ever and performs better than ever. Both things are true, and the explanation says a lot about where AI actually lands in marketing.
What AI took over
The mechanical layer of media buying is gone, and nobody misses it. Bid management went first, then placement selection, then audience construction. Google’s Performance Max and Meta’s Advantage+ made the auction a black box and the machine bids better than any human ever did.
The reporting layer went next. Nobody on our floor exports CSVs to find out what happened yesterday. Anomaly detection flags the account that broke, budget pacing corrects itself, and the morning review is a list of exceptions, not a spreadsheet ritual. This is where AI earned its keep: not by being brilliant, but by being awake at 4am.
What AI did not take over
Three things stubbornly remain human, and they are now the entire job.
The offer. No model decides that your bundle is wrong, your price anchors poorly, or your market is saturated. The biggest wins we had in the last two years came from offer changes, not campaign changes.
The creative angle. AI produces variations at industrial volume, but the insight that a certain objection, spoken plainly in the first three seconds, would flip a cold audience, that still comes from someone who talks to customers. Volume is automated. Judgment is not.
The measurement question. Deciding what to trust when the platform says one thing, attribution says another, and finance says a third. That is an epistemology problem, and epistemology does not autoscale.
The new shape of the team
In 2021 a paid team was buyers plus a data analyst. In 2026 ours looks like operators plus infrastructure: fewer people, each with full access to live attribution, each responsible for a number rather than a channel. The tools got smarter, so the humans moved up a level of abstraction.
The teams struggling with AI in marketing are mostly trying to keep the old org chart and sprinkle AI on top. It goes the other way: automate the layer that should be a machine, then let the people do the part that never was.
We wrote the tools we use for that transition, and two of them, ScaleTrack and ScaleCRM, are open to operators running the same playbook.