Performance marketers are turning to AI to navigate rising costs, data loss and increasingly complex shopper journeys – and it’s working. Creative is more personalized, targeting is sharper and optimization happens in real time.
AI has proven remarkably effective at identifying patterns, predicting outcomes and driving gains across key metrics. Yet many marketers are still asking a harder question: If AI is improving efficiency, why aren’t we seeing the same lift in business results?
The answer often comes down to connection, or the lack of it. Many AI tools are designed to optimize one piece of the performance puzzle – bidding, targeting or measurement – without understanding how the parts fit together. Metrics may improve in isolation, but that doesn’t always translate into outcomes. The missing link is moving from AI that can tell you what to do for a specific case to AI that understands why, so it can be replicated in other similar situations.
To bridge that gap, marketers need AI models that truly understand performance, which requires two dimensions.
The first is integration to connect data, tools and decisioning across the full shopper journey. The second is domain expertise that gives AI a foundational understanding of how commerce actually works, including how products move, shoppers behave and intent becomes purchase.
When those two layers come together, models evolve from task optimizers to performance drivers.
Fragmented tools limit AI’s impact
Every performance marketer knows the pain of fragmented tools. Bidding platforms, creative optimizers and measurement systems each tell part of the story, but without shared context, teams spend valuable time reconciling dashboards instead of accelerating results.
Disconnected systems slow optimization, obscure performance drivers and limit even the smartest AI models.
When data, decisioning and activation are trained together, AI can finally see the full picture. It can learn how a price change impacts add-to-cart rates or how creative performance differs by channel and audience. That closed feedback loop transforms optimization from reactive to proactive.
But here is where general-purpose AI falls short.
Why you need commerce-literate AI
Even when insights are connected, most AI systems still lack real-world commerce experience. They are trained on limited data sets, such as one brand’s campaign history or CRM data, without the broader context of how products, shoppers and media interact at scale.
They are brilliant learners but lack experience. Predictions plateau without exposure to the dynamic mechanics of commerce media. Elements like price elasticity, category trends and cross-channel shopper behavior change by the hour. The ability to predict what matters next depends on AI that truly understands those underlying market dynamics.
Here’s the difference between general-purpose AI and commerce-literate AI in action. A general-purpose model might recognize that a shopper has viewed a blue shirt. But an experienced model understands that it’s a cotton-linen blend trending in coastal markets, its price elasticity is tightening and video ads outperform static images for this SKU on weekends.
That level of fluency requires contextual experience to connect granular product attributes with live shopper behaviors and media dynamics. That’s where AI models move from optimizing for engagement to truly understanding why people buy.
Deeper data builds deeper experience
The most effective AI systems are built on a foundation of commerce-grade intelligence: real purchase transactions, unified product catalogs across retailers, SKU-level attributes and multimodal signals from across the open web. This foundation is enhanced using machine learning for scale, including pattern recognition across billions of interactions, and LLMs to enrich and normalize data. A few good examples include generating consistent product tags, synthesizing customer review insights or connecting variant SKUs across retailers.
Building and maintaining that depth of data is hard work. It requires engineering pipelines that ingest and normalize structured and unstructured data, trusted partnerships across advertisers and publishers and the use of AI tools – both machine learning and LLMs – to keep signals fresh and accurate.
But the hard work is worth it. It’s what transforms AI from a capable optimizer into a true commerce expert – one that can make faster, more accurate and more cost-efficient decisions about who to reach, when to engage, where to appear and what to show.
How AI learns commerce expertise
The path to domain-specific expertise mirrors how humans develop mastery through layered learning:
- Learn the market: Before optimizing for any one client, a model should be trained on broad, multimodal commerce data, such as product catalogs, prices, availability, reviews, creative performance and real shopping behaviors. This builds a foundational understanding of how commerce moves.
- Learn the client: Once that foundation is set, a model can be fine-tuned for a brand’s specific goals and KPIs. It doesn’t have to “figure out” what drives conversions in fashion versus travel versus retail because it already knows the terrain. The learning is faster, the predictions sharper and the outcomes stronger.
This approach accelerates learning curves and compounds performance. Models that understand commerce from the start can make split-second decisions from Day One, adjusting bids, personalizing creative and reallocating spend to deliver measurable gains in engagement, conversion and ROAS.
The next AI revolution: Understanding commerce
AI has already proven it can learn fast. The next evolution is helping it understand the world more deeply.
For performance marketers, that means moving beyond fragmented tools toward connected, commerce-literate systems. They need AI that sees the entire shopper journey and understands how every signal, every channel and every decision connects to real outcomes. We’re talking about AI that doesn’t just spot patterns but grasps the pulse of commerce, from what drives a sale and when intent shifts to how demand moves.
The next generation of performance marketers won’t ask their AI to optimize; they’ll ask it to understand – and those who make that leap first will gain a lasting advantage.
For more articles featuring Wilfried Schobeiri, click here.
