Adobe Revolutionizes Marketing with Agentic AI

Adobe Unveils a New Era of Marketing with CX Enterprise

At the 2026 edition of its Summit, Adobe introduced its latest agentic AI system, named CX Enterprise, which promises to transform the landscape of digital marketing. This event not only featured numerous announcements but also highlighted an industry thesis that extends far beyond the scope of this publisher.

Three Years After Generative AI, the Industry Takes a Turn

Generative AI is now three years old. During this time, publishers have explored the idea that every employee could become capable of creating and manipulating their own agents. However, Frédéric Cavazza, a digital transformation expert, describes this vision as purely utopian. The Adobe Summit 2026 revealed another direction: that of orchestration.

Adobe observed a 269% increase in AI-generated traffic on U.S. e-commerce sites in just one year. Anil Chakravarthy, president of Adobe's Customer Experience Orchestration division, emphasized that 80% of companies recognize the need to understand the proliferation of channels through which they must interact with their customers. This convergence of approaches among giants like Google, Microsoft, and Adobe sends a strong signal of adoption.

For brands, this change is tangible. Lionel Lemoine, head of consulting solutions at Adobe for the Western EMEA region, noted that 25% of traffic and information discovery now occur via large language models (LLMs) and conversational interfaces. It is in this context that Adobe introduced CX Enterprise, laying the groundwork for a new era in digital marketing.

AI Operates, Humans Govern

From AI Tool to AI Operator

During the opening keynote, Anjul Bhambhri, senior vice president of engineering for Adobe's Customer Experience Orchestration division, encapsulated the thesis of this edition in one sentence: "For years, humans have operated the software. In the new model, AI operates it. And humans? They define the strategy."

This shift is not just a simple change of slogan. According to Bhambhri, it is about redefining the relationship between work and software. "The question changes: we no longer ask where we can use AI, but how AI can use what we use. This is precisely what distinguishes the agentic generation," she emphasized.

Generative AI, Agentic AI: What Sets Them Apart

Generative AI is capable of producing content, whether text, images, or videos, based on a request. In contrast, agentic AI goes further: it interprets a goal, plans actions, invokes tools, and executes a complete task, with or without human supervision. The industry is focusing on this second wave, integrating generative layers into agentic ones as needed.

Adobe places this change within a timeline that supports its positioning. "Twenty years ago, Adobe helped define digital marketing. Ten years ago, it refined customer experience management. Today, Adobe is redefining the very fabric of customer experience that AI builds," Bhambhri continued. The transition from CXM (Customer Experience Management) to CXO (Customer Experience Orchestration) is the formula adopted by the publisher for this new phase.

Humans Move Up a Level

This inversion is materialized in CX Enterprise Coworker, the "super-agent" presented as the control center of the new system. The distinction from a classic agent is conceptual. "We tell it less about what we want to do; we tell it what we want to achieve," summarizes Lionel Lemoine. The marketer defines a business objective, for example, increasing cross-selling by 3% on a given product, and the Coworker assembles the necessary agents, proposes a plan, awaits validation, and then executes it.

Humans are not excluded from the process. They move up a level, taking on a strategic supervisory role. Responsibility also shifts. "Responsibility is not with the algorithm, it is not with the agent. It is really with the person who did not critically assess the validation," reminds Lionel Lemoine. On stage, Anjul Bhambhri summarized the division of labor in three words that could become the formula for this transition: "Humans govern, AI operates."

Beyond Generative, Much Deterministic and Ecosystem

The Architecture of CX Enterprise in Six Layers

Adobe presented CX Enterprise as a stack of layers, each with a specific function:

• Three business pillars on the surface: Brand Visibility (Adobe Experience Manager and Agentic web), Customer Engagement (AEP Native Apps), and Content Supply Chain (Adobe GenStudio)
• A cross-layer of analytics: Adobe CX Analytics
• An agentic orchestration layer: CX Enterprise Coworker, Adobe and third-party agents, skills, tools, and MCP servers
• Two intelligence layers: Adobe Brand Intelligence and Adobe Engagement Intelligence
• A foundational data platform: Adobe Experience Platform (AEP)
• Native integrations with major AI providers: Anthropic, OpenAI, Google Cloud, Microsoft, AWS, and Nvidia

Brand Intelligence, More a Harness than a Model

Some of the presented layers are less visible but very telling. Contrary to the dominant discourse on AI, what Adobe promotes in the customer experience part is not predominantly generative. "Customer experience involves a lot of conversion journeys, a lot of analytics, and predictive elements. We are more focused on statistical calculations than probabilistic ones," analyzes Frédéric Cavazza. This nuance applies to the CXO layer. Because on the Creative Cloud side, generative remains more central to the system, with Firefly and its ecosystem of creative agents.

One of the flagship announcements of this edition illustrates this. Adobe Brand Intelligence, presented as a continuous learning engine for a brand's identity, does not function as a generative model. "It should indeed be seen more as a harness," explains Lionel Lemoine. In practice, the intelligence layer is trained on a brand's existing content (about thirty assets are enough to start) and then learns from cycles of validation and rejection to refine its understanding.

Aaron Finegold, Head of Product Marketing for Customer Experience Orchestration, confirms this in more technical but similar terms: "The validation capability relies on specialized computer vision models driven by a reasoning engine. It is not generative." He adds: "Many of my clients are becoming very, very selective about where they use generative capabilities. No model is good enough. And even when it is, it is not always reliable, because you launch a second prompt and get something else." Generative has not disappeared from the offering; it can still be found in the GenStudio chain, but it is now positioned as an option among others in largely deterministic workflows, as Aaron Finegold detailed in our interview.

Adobe Accepts Not to Be the Sole Orchestrator

The second notable positioning is strategic. Adobe has structured its offering to accept that it does not necessarily have to be the orchestrator for the companies that use it. "Perhaps some companies have decided to bet everything on Anthropic.