AI in Media Buying: What the Smart Money Is Doing
AI is being integrated into media buying and planning at a pace that most marketing teams are not ready for. The platforms have been doing it quietly for years, the holding companies are investing heavily, and a growing number of brands are restructuring how they allocate budget, build forecasts, and make real-time bidding decisions based on machine-led signals rather than human intuition alone.
This is not a future-state conversation. The shift is already priced into how Google, Meta, and the major DSPs operate. The question for any serious media buyer or planning director is not whether to engage with AI-driven media infrastructure, but how to stay in control of it.
Key Takeaways
- The major platforms have embedded AI into their bidding and targeting infrastructure so deeply that opting out is no longer a realistic option for most advertisers.
- The companies seeing the best results are not the ones handing everything to automation. They are the ones maintaining tight control over inputs: creative, audience signals, budget constraints, and conversion data.
- AI in media planning is strongest at pattern recognition and allocation at scale. It is weakest at brand judgment, strategic context, and knowing when a number is wrong.
- The risk of full automation is not that the machine makes bad decisions. It is that no one notices until the damage is done.
- Media agencies that are not actively restructuring their planning and buying workflows around AI will lose ground to those that are, regardless of the quality of their human talent.
In This Article
The Platforms Got There First
Before we talk about what companies are doing with AI in media buying, it is worth acknowledging who built the infrastructure they are working within. Google, Meta, Amazon, and the major programmatic platforms did not wait for the industry to catch up. They embedded machine learning into the core of how their ad systems work, and they did it incrementally enough that most advertisers barely noticed.
Smart Bidding on Google Ads is AI. Meta’s Advantage+ campaigns are AI. The way a DSP dynamically adjusts CPMs across an open exchange in real time is AI. Google has been refining how it uses contextual query signals to shape ad delivery for well over a decade. The AI conversation in media buying is not new. What is new is the degree to which it now operates above the waterline, where planners and buyers can actually see it and interact with it.
I spent years managing significant paid search budgets at the agency level. In the early days, you could build a campaign structure, set your bids manually, and have a reasonable expectation that what you built was what ran. That level of control has been systematically reduced. Not because the platforms are adversarial, but because their systems genuinely outperform manual bidding in most high-volume environments. The math is not close.
If you are working in paid advertising and want broader context on how the channel landscape has evolved, the paid advertising hub at The Marketing Juice covers the strategic layer across search, social, and programmatic.
What the Holding Companies Are Building
The agency holding groups, WPP, Publicis, Omnicom, IPG, and Dentsu, have all made significant public commitments to AI-driven media infrastructure. Publicis built Marcel. WPP has invested in its own AI operating layer. These are not PR exercises. They represent genuine restructuring of how planning, buying, and measurement workflows are organised at scale.
What they are building, broadly, falls into three categories. First, AI-assisted planning tools that can model audience reach, frequency, and channel mix across a media plan faster and with more data inputs than any human planning team. Second, real-time optimisation engines that adjust in-flight spend allocation based on performance signals. Third, AI-driven creative testing frameworks that connect ad performance data back to creative decisions much earlier in the production cycle.
The independent agency market is more fragmented. Some shops have built their own proprietary tools. Most are working with a combination of platform-native AI features and third-party software layers from companies like Albert, Smartly, Skai, or Basis Technologies. The sophistication varies enormously, and the gap between what the holding groups can build and what an independent agency can access is widening.
I ran an agency that grew from around 20 people to over 100 during a period when programmatic buying was becoming the dominant model. The operational pressure that came with that growth was real. You are hiring faster than you can train, you are managing more client accounts than your team can give proper attention to, and you are constantly asking yourself which decisions can be systematised and which ones genuinely need a human in the loop. AI in media buying answers some of that question. Not all of it.
Where AI Is Genuinely Adding Value
There are specific areas where AI-driven media buying is delivering measurable improvement, and it is worth being precise about what they are rather than making sweeping claims.
Bid management at scale is the clearest win. When you are running thousands of keywords or audience segments across multiple platforms, the number of variables involved in setting and adjusting bids in real time is beyond what any human team can manage manually. AI handles this better. It processes more signals, adjusts faster, and does not get tired or distracted. Platform-level quality scoring mechanisms have also become more sophisticated over time, which means the AI bidding systems are working with better underlying data than they were five years ago.
Budget allocation across channels is another area where AI tools are adding real value. Traditional media planning relied on historical benchmarks, category norms, and planner judgment. AI-assisted planning can model a much larger set of scenarios, factor in cross-channel interaction effects, and update recommendations based on in-flight performance data. This does not replace strategic judgment, but it does give planners a better analytical foundation to work from.
Audience targeting has also improved materially. The ability to identify high-value audience segments based on behavioural signals, lookalike modelling, and predictive intent scoring has made targeting more precise in environments where first-party data is available and properly connected to the media stack. The conversion rate differential between well-targeted paid placements and broader reach is significant enough that improvements in audience precision translate directly into measurable efficiency gains.
Early in my career, I launched a paid search campaign for a music festival at lastminute.com. The campaign itself was not complicated, but the speed at which it generated revenue, six figures within roughly a day, made it clear that the combination of intent-based targeting and the right offer at the right moment was genuinely powerful. What AI does is compress the time it takes to find that combination, and it does it across a much larger surface area than a human team can cover manually.
Where AI Creates New Problems
The honest version of this conversation includes the failure modes, because there are several that are not getting enough attention.
The first is the black box problem. When an AI system is making thousands of micro-decisions per day about where to show your ads, at what price, to whom, and in what context, the audit trail becomes difficult to follow. Most advertisers cannot explain why their CPAs moved in a given week. They can see the outcome, but not the reasoning. This is a real problem when something goes wrong, when brand safety issues emerge, or when a client asks a reasonable question about why spend shifted the way it did.
Attribution and campaign analytics were already complicated before AI-driven buying became the norm. The combination of opaque platform algorithms and multi-touch attribution models means that the data story a media team presents to a client is often a confident-sounding approximation of what actually happened, not a precise account of it. AI makes this worse, not better, at least for now.
The second problem is over-automation without strategic oversight. I have seen this play out multiple times. A brand hands its campaigns over to platform automation, removes the human review layer to save on agency fees, and watches performance drift over months as the system optimises for the signals it can measure rather than the outcomes the business actually cares about. The machine does not know that your brand has a positioning problem. It does not know that the conversion you are optimising for is not a good proxy for customer lifetime value. It optimises what it can see.
The third issue is data quality dependency. AI systems in media buying are only as good as the data they are trained on and the signals they receive. If your conversion tracking is broken, if your CRM data is not connected properly, or if your attribution model is attributing credit in a way that does not reflect reality, the AI will optimise confidently in the wrong direction. Garbage in, garbage out is not a new principle, but AI makes it a more expensive mistake.
A useful framing here: the landing page experience is a good example of where AI optimisation hits a wall. A platform can optimise your bids and audience targeting with impressive precision, but if the page someone lands on is not converting, the system will eventually exhaust its options and performance will plateau. The AI cannot fix what happens off-platform.
The Companies Getting It Right
The organisations seeing the best results from AI integration in media buying share a few characteristics that are worth naming.
They treat AI as an execution layer, not a strategy layer. The planning decisions, the channel mix, the audience strategy, the creative brief, these remain human decisions informed by AI outputs rather than replaced by them. The AI handles the volume and speed of execution that humans cannot match. The humans handle the judgment calls that machines cannot make.
They invest seriously in data infrastructure before they invest in AI tooling. Clean first-party data, properly connected to the media stack, is the foundation everything else runs on. Companies that skip this step and jump straight to AI-driven buying tools find that the tools underperform because the underlying data is not reliable enough to train on.
They maintain human oversight at the campaign level, not just at the reporting level. There is a difference between reviewing a weekly performance report and having someone who understands the media mechanics actively monitoring what the system is doing and why. The former catches problems after they have already cost money. The latter catches them earlier.
They also test and validate AI recommendations rather than accepting them as authoritative. When an AI planning tool recommends shifting 30% of budget from one channel to another, the best media teams ask what assumptions that recommendation is based on, whether those assumptions hold in their specific context, and what would have to be true for that recommendation to be wrong. This is not scepticism for its own sake. It is professional discipline.
When I was judging the Effie Awards, the campaigns that stood out were not the ones with the most sophisticated technology stack. They were the ones where a clear strategic idea had been executed with discipline across every touchpoint. Technology was in service of the idea, not the other way around. That principle applies directly to AI in media buying. The tool is not the strategy.
What This Means for Media Agency Structure
The structural implications for media agencies are significant and not fully worked through yet. If AI systems can handle bid management, audience optimisation, and budget allocation at scale, the value of a large team of junior buyers executing manual optimisations is reduced. The value of senior strategists who can set the right objectives, interpret outputs critically, and make brand-level judgment calls is increased.
This is already reshaping hiring. The agencies growing their capability fastest are recruiting people who can sit at the intersection of data science, media strategy, and commercial thinking. Pure media planners who cannot engage with data infrastructure are increasingly exposed. Pure data scientists who cannot translate outputs into business decisions are equally limited.
The client relationship is also changing. Clients are asking harder questions about what they are actually paying for when so much of the execution is automated. The honest answer from agencies needs to be about strategic oversight, creative quality, data architecture, and the human judgment layer that keeps the machine pointed in the right direction. That is a harder value proposition to articulate than “we have 40 people working on your account,” but it is the more defensible one.
There is more on the mechanics of paid channel strategy across the broader landscape in the paid advertising section of The Marketing Juice, including how to think about performance measurement and channel selection as the media environment continues to shift.
The Honest Assessment
AI in media buying is not overhyped in the sense that the capabilities are real and the performance gains in the right contexts are genuine. It is overhyped in the sense that many companies are treating it as a solution to problems that are actually strategic or structural, not executional.
If your media strategy is unclear, AI will execute it at greater speed and scale. If your creative is weak, AI will serve it more efficiently to the wrong people. If your measurement framework is broken, AI will optimise confidently toward the wrong outcomes. The technology amplifies what is already there, good and bad.
The companies that will build durable advantage from AI in media are the ones treating it as a capability to be integrated thoughtfully into a well-run operation, not a shortcut to a well-run operation. That distinction matters more than any specific tool or platform.
Early in my agency career, I was handed a whiteboard pen in a brainstorm for a major brand when the founder had to leave the room unexpectedly. The internal reaction was something close to panic. But the discipline of having to structure thinking clearly, defend it to a room of experienced people, and connect it to a commercial outcome, that is exactly the skill that AI cannot replace. It can process more data than any human. It cannot tell you whether the thinking is right.
About the Author
Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.
