There is a version of this conversation that is entirely unhelpful: a list of tasks AI will automate, followed by reassurance that human judgement will always matter, followed by a vague call to “adapt.” That is not this post.
The honest version is more uncomfortable and more interesting. AI is not a uniform threat to media buying. It is a specific threat to specific parts of media buying, and an amplifier for the parts it cannot replace. Whether that makes media buyers more or less valuable depends entirely on which parts they have been building their expertise around.
Manual bidding is largely gone for anyone running at scale. Smart Bidding on Google, Advantage+ on Meta, automated budget allocation across campaigns: these are not future capabilities. They are the default. The platforms have spent years training their own models on billions of auction signals, and their bidding algorithms consistently outperform human-managed CPC strategies in controlled tests.
Audience targeting has followed the same trajectory. The ability to layer custom audiences, exclusions, and interest segments was a genuine skill in 2017. It is now less relevant because the platforms’ broad targeting models, fed by conversion signals and first-party data, typically outperform tightly managed audience builds. The craft of audience architecture has been partially replaced by the craft of giving the algorithm enough signal to do it for you.
Reporting automation, anomaly detection, budget pacing alerts: the operational layer of media buying has been steadily absorbed by platform tools, third-party software, and increasingly by AI-powered analytics platforms that can flag performance drops before a human would notice them.
None of this is coming. It is already here. Media buyers who are primarily competing on manual optimisation speed are in the most precarious position.
What AI cannot replicate
The limitations of AI in media buying are not about effort or processing power. They are structural. The platforms can optimise within a campaign structure, but they cannot tell you whether the campaign structure is right. They can improve cost per result against a conversion event, but they cannot tell you whether you are optimising against the right conversion event. They can scale what is already working, but they cannot identify why it is working or whether it will continue to work as conditions change.
This is the strategic layer, and it requires things that machine learning models are not designed to provide: business context, commercial judgement, an understanding of the client’s broader market position, and the ability to reason about things that have not happened yet.
A smart bidding algorithm does not know that your client is about to launch a new product line, raise prices, or shift their customer acquisition focus from volume to margin. It does not know that their creative is burning out before the frequency data surfaces it clearly. It does not know that their attribution model is overclaiming on last-click and that the real incrementality is lower than the dashboard suggests.
A media buyer who understands those things is not competing with the algorithm. They are directing it.
The skills that are gaining value
If automation is compressing the value of manual platform management, it is simultaneously expanding the value of the capabilities that sit around it.
Creative strategy is the most obvious beneficiary. When targeting and bidding are largely automated, creative is the primary variable the human controls. The ability to develop, brief, test, and iterate creative in a structured way, building a body of learning rather than just reacting to performance, is now a core media buying skill rather than a adjacent one.
Measurement architecture matters more, not less, in an automated world. When the platforms are making decisions based on the signals you feed them, the quality of those signals determines the quality of the output. Setting up clean conversion tracking, structuring attribution correctly, deciding what the platform should optimise against: these are strategic decisions with significant downstream consequences.
Channel strategy has become a distinct capability as the paid media landscape has fragmented. The question of how Meta, Google, TikTok, YouTube, and programmatic interact, where demand is created versus captured, how budgets should flow between channels at different business stages, this is not something the platforms will tell you. It requires a view across the whole system that no individual platform’s automation can provide.
The media buyers who will be displaced
The honest answer to the original question is that AI will displace some media buyers, specifically those whose primary value was in the operational execution of tasks the platforms now handle themselves. If your expertise is in audience segmentation, manual bid adjustments, and campaign structuring, and you have not developed the strategic layer around it, the automation has already eroded the foundation you were standing on.
This is not a failure of effort. The platforms have deliberately commoditised these capabilities because it serves their commercial interest to do so. But the result is that the moat around execution-level expertise is now shallow.
The media buyers who will become indispensable
The ones who will become more valuable are those who were always doing more than optimisation: the ones who were asking why a campaign structure made sense before building it, who were treating creative as a strategic variable rather than a production task, who were reading business performance rather than just platform metrics, and who were building measurement systems that told clients something true rather than something flattering.
Those capabilities have always mattered. They just matter more now that the platforms have automated everything around them.
The future of media buying is not about surviving AI. It is about being genuinely, demonstrably better at the parts of the job that AI cannot do, and being honest with yourself about whether you have built that expertise or assumed it.
0 Comments