By the time Anil Chakravarthy got to the announcement, the framing was already doing most of the work. Customers are arriving through Large Language Model platforms at a pace that dwarfs every other acquisition channel. Marketing teams are running content demand that is rising five times faster than manual production can keep up with. Channel counts are above 50 for Adobe's own marketing team, and attribution has become a measurement crisis. Something had to sit across all of it. That something is Adobe CX Enterprise, and the short version is that Adobe has decided the next unit of customer experience work is not a campaign or a journey. It is an agent.
What Adobe Actually Announced
CX Enterprise is not a new product. It is a new architecture. It binds existing Adobe products, Adobe Experience Platform, Real-Time Customer Data Platform, Adobe Journey Optimizer (AJO), GenStudio, Workfront, Experience Manager, into a single governed agentic layer that covers three use cases: brand visibility, customer engagement, and the content supply chain.
The headline product inside that architecture is the CX Enterprise Coworker. Earlier agentic releases from Adobe were task agents. Invoke the agent, perform the task, return. The Coworker is positioned differently. It is persistent, always on, goal oriented. Give it a target, say a three percent cross-sell lift, and it is designed to decompose that target into a multi-step execution plan, assemble the right audience segments from Real-Time CDP, pull creative assets from GenStudio, get human approval on the plan, run the campaign, and report back against the goal.
Around the Coworker, Adobe introduced two systems of intelligence. Adobe Brand Intelligence is a learning layer that takes brand guidelines, approved assets, review cycle feedback, annotations, rejections, and approvals, and turns them into a machine-readable reference that agents can query before producing content. Adobe Engagement Intelligence is a decisioning engine built around customer lifetime value rather than single-touch conversion, with a governed data model that makes agent actions auditable after the fact.
The brand visibility layer got its own set of specific tools. Adobe LLM Optimizer for content discoverability inside AI platforms, Adobe Brand Concierge as the interface agents use to represent the brand accurately, AEM Sites Optimizer, and the pending acquisition of Semrush, which brings search engine optimization and generative engine optimization capabilities into the stack. On the engagement side, AJO picked up a new extension called Journey Optimizer Loyalty that pulls gamified mechanics and loyalty status directly into personalized journeys. Adobe CX Analytics adds a measurement layer that covers LLM-referred interactions, not just web and app.
Then there is the ecosystem piece. CX Enterprise is built on open standards. Model Context Protocol, the interface specification Anthropic published last year, and Agent2Agent, the inter-agent communication protocol, sit underneath the platform. Native integrations run across Amazon Web Services, Anthropic, Google Cloud, International Business Machines, Microsoft, NVIDIA, and OpenAI. An NVIDIA partnership brings the NVIDIA OpenShell secure runtime and Nemotron open models into Adobe's pipeline for regulated industries.
General availability is "in the coming months." That matters.
How It Works in Practice
The easiest way to understand the architecture is to follow a single workflow from signal to action.
A customer engagement signal arrives. Someone has clicked on a retention email, or has visited a pricing page three times in a week, or has asked a personal AI agent about a competitor's product. That signal lands in Real-Time CDP, which now ingests unstructured data alongside structured profiles. Customer Journey Analytics interprets the signal in context. Engagement Intelligence scores the action against a defined business objective, customer lifetime value expansion, churn prevention, cross-sell probability, and decides what the next best action should be.
The Coworker takes that decision and executes. It queries Brand Intelligence to confirm that the planned creative treatment is on brand. It checks GenStudio for existing assets that match the context, or requests generation if they do not exist. It submits the draft campaign plan for human sign-off. Once approved, it publishes to the target channels through AJO. The audit trail it leaves behind includes the signal, the decision logic, the asset lineage, the approval, and the measurement.
That description is the vendor's picture. The architecture is credible. The operating model required to use it is where most enterprises are not ready.
Brand Intelligence is only as good as the brand documentation it ingests. Most enterprises have a brand guide, a few thousand approved assets, and an implicit governance model that lives in the heads of the creative directors who enforce it. Turning that implicit model into a machine-readable intelligence layer is the real project. It is not a configuration step.
Engagement Intelligence requires a customer data foundation that is complete, clean, and current. Most enterprises have fragments. Decisioning optimized for customer lifetime value also requires the business to have defined what customer lifetime value actually means, by segment, by product, by channel. Most companies have a number in a slide deck, not a working definition that can drive real-time agent decisions.
The Coworker itself depends on the signal quality upstream. Bad signals produce bad next-best-actions, faster.
What Changes for the CMO, the CIO, the CXO
For the chief marketing officer, this is a rewrite of how marketing work gets produced. The model shifts from briefs flowing into a creative queue to goal targets flowing into a persistent agent. Your team stops being the people who produce the next campaign. They become the people who define the goals, approve the plans, and audit the outcomes. That shift requires your marketers to get comfortable reading agent execution plans and intervening when the plan is wrong. It is a skill set most marketing organizations have not trained for.
For the chief information officer, CX Enterprise is a governance project dressed as a marketing purchase. The audit trail across agent actions has to meet the same standard as any other governed system of record. Agent permissions have to be scoped. Brand Intelligence and Engagement Intelligence sit on top of customer data that is already governed by the data team, which means any failure in the agentic layer is a data governance failure the CIO owns. The open-standard posture, Model Context Protocol and Agent2Agent, is genuinely useful here because it means the agent layer can be inspected, logged, and replaced. Closed-standard agent platforms would be harder to govern.
For the chief experience officer, if your organization has one, this is the first vendor platform that treats experience orchestration as a single system rather than a sequence of stitched-together products. That is useful. It is also a deep commitment. Once your customer data, brand guidelines, approval chains, and engagement decisioning are encoded into Adobe's intelligence layers, the switching cost to a different vendor is no longer just the data migration. It is the institutional knowledge your organization has formalized inside Adobe's system.
The joint CIO and CMO governance model that Workday showed at Summit last year is the adoption prerequisite here. Without it, this is a purchase that will not clear its internal bar.
What the CMO Conversation Made Clear
The most useful session I sat through was the on-stage conversation between Adobe Chief Marketing Officer Lara Hood Balazs and Comcast Xfinity Chief Growth Officer Jon Gieselman. This was not a demo. It was two operators describing what happens when a marketing organization that has run for forty years starts running brand intelligence inside its approval loop.
Gieselman was direct about why Xfinity engaged Adobe as a design partner. The category had become promotional and price-driven. Customer relationships had become transactional. Xfinity had, in his words, become a technology company that needed to become a tech company. That distinction is sharper than it sounds. A technology company operates infrastructure. A tech company ships products customers feel. Agentic AI is the tool he is using to close that gap.
The most honest line of the session was about bottlenecks. Gieselman said the thing that keeps him up at night is not the creative he sees in weekly reviews. It is the creative he does not see. When a team is producing hundreds of assets a week across fifty-plus channels, the CMO becomes the bottleneck by definition. Brand Intelligence is the first platform capability I have seen that is designed explicitly to solve that bottleneck. Not by moving the CMO into every review, which does not scale, but by encoding what the CMO would flag into a layer that agents consult before content ships.
Balazs picked up on the harder part. Human judgment and taste are hard to encode. The training loop for Brand Intelligence depends on feedback from actual review cycles, rejections, annotations, approvals. That means the platform gets smarter only if your marketing leadership is disciplined about documenting why a piece of creative was rejected. Most organizations are not disciplined about this. The insight becomes institutional knowledge in the head of one senior creative, and then that person leaves.
Gieselman closed with advice for the marketers in the room. He said most marketing is beige. He tells his creative teams not to walk up to the line and take one step across it. If he pulls them back, they have gone too far, but that is how you break through. That is not a product statement. It is an operating principle. The platform exists to free his teams from mundane production so they can spend their time on creative work that is not safe. If the platform works as intended, the output should become less beige, not more.
That is the test worth watching. If brand intelligence produces safer, more homogenized marketing, Adobe has built a guardrail platform. If it produces braver, more distinctive marketing, Adobe has built something that compounds.
What Comes Next
Three things are worth watching over the next twelve months.
First, whether general availability actually arrives on the timeline Adobe described. "In the coming months" is a wide window, and the Coworker depends on integrations across Real-Time CDP, GenStudio, Workfront, Journey Optimizer, and the two new intelligence layers. Shipping the integrated experience at production grade, not demo grade, is the hard part.
Second, whether the partner ecosystem moves past standardization announcements into actual joint deployments. Six global agencies, dentsu, Havas, Omnicom, Publicis, Stagwell, and WPP, announced that they are standardizing on CX Enterprise. Nine system integrators, Accenture, Capgemini, Cognizant, Deloitte Digital, Ernst and Young, International Business Machines, Infosys, PricewaterhouseCoopers, and Tata Consultancy Services, announced industry-vertical packages. Those commitments matter only if named reference customers show up inside twelve months running real workloads.
Third, whether the LLM Optimizer thesis holds up. The Semrush acquisition signals that Adobe is betting brand discoverability inside AI platforms will become the dominant digital marketing channel. If customers are arriving through AI agents at the rate Adobe's own data suggests, that bet is correct. The risk is that the major AI platforms, OpenAI, Google, Anthropic, Perplexity, Microsoft, have no obligation to expose the retrieval APIs that LLM Optimizer depends on. Some have exposed partial access. Others have not. The optimizer is only as useful as the API surface it is allowed to see.
The Summit context matters too. This was the first Summit since the announcement that CEO Shantanu Narayen is stepping down. The strategic direction is clear. The execution will happen under new leadership. How that transition plays out will shape whether CX Enterprise becomes the platform that redefines customer experience orchestration or the platform that was announced just before the pivot.
Three things I am sitting with after day one.
First, this is powerful, and it is going to move enterprise revenue. Brand visibility inside AI platforms, an agentic coworker that executes against a business goal, and a brand intelligence layer that protects creative at scale are not incremental features. They sit on three of the hardest problems every customer experience leader is working on right now. The business case writes itself for any enterprise that has already committed to Adobe Experience Platform.
Second, I am glad Adobe did not call it Adobeclaw. There is a pattern in this product cycle where every vendor slaps an agentic name on a product to signal newness. Adobe took a harder path. The Coworker is a descriptive label that tells you what the product does. That is a restraint I appreciate and a signal that Adobe is selling the outcome, not the branding.
Third, the adoption curve will not be set by Adobe. It will be set by the speed at which each enterprise can move. Brand Intelligence only works if your brand is documented. Engagement Intelligence only works if your customer data is clean and your definition of customer lifetime value is operational. The Coworker only works if your marketing organization is ready to work alongside an agent instead of around it. Enterprises that have done the preparation work will compound fast. Enterprises that have not will watch their competitors pull ahead and wonder why the same platform produced different results.
Adobe has built something genuinely impressive, and the product narrative deserves a wider audience than the one it currently reaches. The keynote moves comfortably through Adobe Experience Platform, Real-Time Customer Data Platform, GenStudio, Workfront, AEM, AJO, MCP, A2A, and a dozen more acronyms that are second nature to existing Adobe customers. For enterprises already inside the ecosystem, this is a shorthand that saves time. For enterprises outside of it, considering their first commitment, the same shorthand can make the platform feel like an exclusive club where the members already know the handshake.
That matters commercially. The If You Know You Know tone is a real barrier for the mid-market enterprise that is weighing Adobe against a simpler, more self-explanatory alternative. A modest shift, expanding acronyms in customer-facing keynotes, grounding each capability in the business outcome before the product name, would open the story to the buyers Adobe most wants to bring into the fold. The technology is genuinely accessible. The language around it can be, too.
The other suggestion is for the investor narrative. Adobe has a story to tell about how the rise of Large Language Model generative capabilities strengthens its business rather than threatens it, and that story is worth telling more clearly and more often. Consumer-grade generative tools, the ones embedded in ChatGPT, Gemini, Claude, Copilot, and the like, are learning surfaces. They help people experiment, ideate, and develop a feel for what creative production can look like. That is a wonderful thing for the craft, and it expands the total addressable market for professional creative software rather than shrinking it.
Professional creative production, brand-governed content supply chains, enterprise-grade customer experience orchestration, these are different categories of work. They require the governance, the audit trail, the brand intelligence layer, and the deep workflow integration that Adobe has spent decades building. No Large Language Model is currently positioned to replicate that stack, and the ones closest to trying are partnered with Adobe rather than competing with it. Making that distinction visible to the investment community would reduce a headline risk that is larger in perception than in reality, and would help the market value Adobe for what it actually is. Adobe is the professional layer the Large Language Model era depends on, not the incumbent the Large Language Model era replaces.
