Adobe brings its marketing agent to ChatGPT and Claude: the agent battle is fought on integration, not the model

By PYMNTS · November 4, 2026.

Adobe has announced that CX Enterprise Coworker, its customer experience orchestration agent, is now available inside ChatGPT, Claude, Databricks Genie, Gemini for business and Microsoft Copilot. According to the company's own press release, it will soon also come to Amazon Quick, IBM and Snowflake.

Adobe says the agent handles data analysis, audience forecasting, customer journey optimization, and product and work planning and management. So that it can connect to other systems without custom development, the company has launched ready-made connectors with MCP (Model Context Protocol, the open standard that lets an agent use external tools and data) servers. They cover Amazon Ads, Databricks Genie, Demandbase, an ads MCP server for Meta campaigns, Microsoft Work IQ, RainFocus and Snowflake Cortex AI.

Adobe is also working with global systems integrators to build industry-specific agentic solutions: Accenture, Capgemini, Cognizant, Deloitte Digital, EY, IBM, Infosys, PwC and TCS. Anjul Bhambhri, Adobe's senior vice president of customer experience orchestration engineering, explained the move by saying that marketers today use AI tools in isolation, without the organizational context or domain knowledge needed to personalize every stage of the customer lifecycle.

Our take: this story isn't about marketing; it's about where the agent lives. Adobe is giving up on getting users to come to its interface and is instead embedding itself in the assistants they already use, including those of direct AI rivals. It's a bet consistent with a view we have long held: in the agentic world, the winner is whoever controls the plumbing (connectors, standards, data context), not just whoever has the smartest model. The fact that ChatGPT, Claude, Gemini and Copilot appear as mere distribution channels shows that models are becoming interchangeable infrastructure, while the differentiator shifts to the brand and customer knowledge each company builds up.

There is also reason for caution. All of the above comes from an Adobe press release: there are no adoption figures or measured results, and "connects tools with brand intelligence" is, for now, a product promise. What will need to be judged is the less glamorous layer: who acts, with what permissions and what access to customer data, and who can revoke it. When a single agent operates from five different platforms and touches Amazon Ads or Meta campaigns, identity and access governance becomes the real product, and also the main risk.

In the short term, this puts pressure on operational marketing roles: analysis, forecasting and planning tasks that used to occupy whole teams move into a conversation with an agent. Broadly speaking, what gets displaced is routine work, not creative judgment or the customer relationship. In the long term, the same logic points to what we care about: automating the repetitive so that people can spend their time on the decisions and creativity that truly define them. For that transition to go well, open connectivity based on standards such as MCP matters more than any single product launch.