By Eric Brackmann, Vice President of Commerce Media at Koddi
With consumer spending under pressure, commerce media networks have emerged as one of the most important growth engines in digital advertising. McKinsey projects the U.S. commerce media market will surpass $100 billion by 2027, growing faster than display, connected TV, and even search. That growth is not just about inventory or scale. It is increasingly driven by how intelligently networks apply AI across their technology stack.
As the market has expanded, it has also become crowded. There are now more than 250 commerce media networks, many offering similar formats, familiar integrations, and similar performance narratives. In response, AI has become a default talking point. But in practice, most platforms still apply AI in narrow, disconnected ways. That distinction matters more than ever.
In a market where a small number of networks capture the majority of advertiser spend, performance compounds for those with better intelligence. Platforms that coordinate AI across decisioning, optimization, and measurement create multiplicative value. Platforms that rely on isolated AI generate incremental gains that plateau quickly.
Point AI is table stakes; coordinated, end-to-end AI is decisive
No single AI model or agent can solve commerce media on its own. Effective networks require specialized intelligence across relevance, bidding, pacing, inventory management, measurement, and incrementality. The difference between leading platforms and the rest is not whether they use point AI solutions. It is whether those solutions operate as a system.
Many commerce media tools deploy AI to optimize a single dimension, most commonly relevance. These solutions predict which ad or product a shopper is most likely to engage with. Relevance matters, but on its own, it solves only part of the problem.
When relevance models operate without bid intelligence, they cannot manage cost efficiency or scale. When bidding is not informed by pacing and inventory signals, budgets misfire. When measurement sits downstream and does not feed back into optimization, learning arrives too late to change outcomes. Each AI solution may perform well in isolation, but the network underperforms as a whole.
Coordinated, end-to-end AI changes that dynamic.
Within a unified platform, relevance models inform bidding models. Bidding agents incorporate advertiser goals, budget constraints, and inventory availability. Pacing intelligence ensures spend is distributed intelligently over time. Measurement and incrementality models feed real performance signals back into decisioning loops. Each model and agent remains specialized, but none operate blindly.
That coordination is where AI becomes multiplicative rather than additive.
Why relevance-only optimization breaks down at scale
Relevance-only platforms often look strong in controlled environments. They surface ads shoppers are likely to click. But commerce media is not a relevance problem alone. It is an economic system.
Without bid optimization, relevance does not account for price. Without pacing intelligence, efficiency trades off against consistency. Without inventory awareness, spend continues behind unavailable products. Without real-time measurement, optimization is reactive rather than proactive.
This is why relevance-centric tools often struggle with enterprise advertisers. As budgets grow and complexity increases, these platforms cannot balance efficiency, scale, and control simultaneously. Performance flattens. Manual intervention increases. Trust erodes.
Commerce media requires real-time decisioning across the full equation. What to show, how much to bid, when to spend, where to allocate budget next, and how to prove incremental impact. AI must support all of it.
Real-world experience shapes better AI
Coordinated, end-to-end AI cannot be built in theory alone. It requires deep exposure to how commerce media networks actually operate.
Live networks are volatile. Products go out of stock. Promotions shift daily. Category rules conflict. Advertisers change goals mid-flight. Shopper behavior evolves continuously. AI models and agents must absorb this friction without degrading performance or breaking trust.
Platforms built without real operating experience tend to underestimate this complexity. Their models perform well on clean data and controlled assumptions, but struggle when edge cases pile up. Human workarounds fill the gaps. Over time, those gaps become structural limitations.
AI shaped by real-world commerce media operations behaves differently. Its models and agents are trained on imperfection. Feedback loops are designed to handle change, not avoid it. Intelligence becomes resilient, not fragile.
System intelligence outperforms local optimization
In fragmented tech stacks, AI improves individual components but cannot optimize the system. One model maximizes click probability. Another chases efficiency. Measurement validates outcomes after the fact. Each function succeeds locally, but the overall network leaves value on the table.
In a coordinated AI platform, every signal improves the system. Performance in one channel informs decisions in another. Measurement is not an endpoint; it is an input. Incrementality modeling sharpens where budget flows next.
This system-level intelligence is what allows networks to scale performance sustainably. It reduces operational burden while increasing precision. And it allows platforms to evolve as advertiser expectations and market dynamics change.
Measurement that improves outcomes, not just reports them
Measurement has become table stakes. Using it to drive optimization is the real differentiator.
According to the Association of National Advertisers, 40% of brands struggle to get timely data from commerce media networks. When insights arrive late, opportunities are missed, and confidence suffers.
In a coordinated AI system, measurement operates in real time. Performance shifts trigger automatic adjustments. Bids respond to demand and inventory changes. Waste is reduced before it occurs. Incrementality analysis improves as models and agents observe the full system rather than isolated touchpoints.
This transparency builds advertiser trust. And trust drives larger, longer-term investment.
The bottom line
Commerce media has moved into a new phase. And, the reality is that most networks simply aren’t delivering at the level of sophistication expected. According to a commerce media benchmark study conducted by Forrester Consulting, 42% of commerce media operators describe themselves as advanced, yet only 13% actually meet the criteria.
Point AI solutions are essential, but only when they work together inside a single platform. Isolated models and agents optimize locally and plateau. Coordinated AI compounds value across relevance, bidding, pacing, and measurement.
Platforms that focus narrowly on relevance solve part of the problem. Platforms built without real operating experience struggle when theory meets reality. The networks that win will be those powered by coordinated, enterprise-grade AI shaped by real-world execution.
Commerce media is a growth engine. But it is system intelligence, not isolated optimization, that determines how powerful that engine becomes.
