Adobe connects Coworker to AI platforms for marketing workflows

Adobe is extending access to CX Enterprise Coworker beyond its own interface through integrations with AI platforms used in businesses. The 4 November announcement includes ChatGPT, where the service is offered as a plugin, Claude, Databricks Genie, Gemini for business and Microsoft Copilot. The system coordinates marketing and customer experience activities using an organisation's data, tools and rules.

According to Adobe, the integration lets teams start those processes from the AI platform where they already work. Availability in Amazon Quick, IBM and Snowflake is presented separately as a future stage. The expansion concerns access points and system connectivity, while Coworker itself had been announced as generally available in June 2026.

A marketing objective can start a process across several applications

Coworker is built for processes in which one request involves several operations: analysing performance, defining an audience, preparing a campaign and coordinating actions across connected systems. Adobe describes an orchestration layer that selects the appropriate skills and tools, executes the stages and checks the outcome against the objective it received.

The context used includes company terminology, customer and channel data, brand rules and internal policies. That information is separate from the wording of the request itself. A campaign process needs both the objective expressed by the team and access to the data and tools through which it can be prepared or executed.

Adobe presents the system as a connection between analysing a situation and taking action in company applications. For a marketing team, this can mean reviewing a campaign's performance, examining an audience segment and preparing an intervention in the customer journey. Actions are coordinated according to the defined objective and the capabilities connected within the organisation.

The orchestration mechanism is called CX Enterprise harness and represents an evolution of Adobe Experience Platform Agent Orchestrator. The provider describes using specialised instructions, business context, memory and connected tools at each stage. Its capabilities include retrying operations that encounter errors and dividing complex tasks into parallel workflows, whose outputs are reunited in the process coordinated by Coworker.

Connectors use standards for tools and agents

The architecture uses Model Context Protocol, known as MCP, and Agent2Agent, abbreviated A2A. The former connects AI applications to tools and data sources, while the latter supports collaboration between agent systems. Under this structure, the platform where a user starts and the application where an operation is performed can be different components.

Alongside the expansion, Adobe is announcing preconfigured MCP connectors for partner systems. The list includes Amazon Ads, Databricks Genie, Demandbase, an ads server for Meta campaigns, Microsoft Work IQ, RainFocus and Snowflake Cortex AI. The connectors provide access paths to connected tools; a service appearing in that list is distinct from the Coworker interface being available within that provider's platform.

The customisation documentation distinguishes four components. Skills standardise repeatable workflows, Integrations connect systems, Plugins bundle skills and MCP servers into installable packages, and Memory retains preferences and working styles across conversations. They are different configuration layers within the same solution, each with a role in adapting it to the team's work.

Current compatibility is specified by product and edition

Effective feature access is tied to the module used, licensing, organisation-level enablement and available connectors. In its current compatibility description, Adobe lists the B2C editions of Real-Time CDP, Journey Optimizer and Customer Journey Analytics. The company also identifies an available connector for Marketo Engage.

In the same description, Experience Manager, Workfront and B2B editions are marked as coming later. The broader Projects component is also described as planned. Those boundaries distinguish workflow examples presented across the product from capabilities enabled in the configuration available to a particular organisation.

The expansion into external AI platforms does not itself grant access to every Adobe application. A team uses the capabilities associated with its enabled products and integrations. In this context, application integration and workflow automation involve defining connected systems, permitted operations and user responsibilities.

Data analysis retains the user's permissions

The Coworker Chat documentation describes connections to Customer Journey Analytics data views and Adobe Analytics report suites. Users can access only the views, dimensions, metrics and segments available to them in Analysis Workspace. The connection to the AI tool preserves those permissions rather than automatically opening every organisational dataset.

Selecting a data source matters to the meaning of the analysis. A Customer Journey Analytics data view determines how data is interpreted, while an Adobe Analytics report suite contains data collected from websites and applications. A preference for a particular source can be retained in memory, allowing later conversations to use the same working context.

Once a visualisation has been generated, it can be opened in an Analysis Workspace project for detailed examination. The documented workflow allows a move from a conversational answer to analysis in the reporting tool, where the team can modify the visualisation and verify the data. The project opens in Customer Journey Analytics or Adobe Analytics, depending on the source of the information used.

Approvals and the activity record remain part of the process

Adobe describes a governance layer that inherits an organisation's roles, data policies and approval procedures. Configurable checkpoints can hold execution until the responsible stakeholder approves a stage. According to the product description, the audit trail records actions, policy decisions and human approvals.

Access through an external AI platform therefore keeps company systems and rules central to the workflow. The announced expansion provides new starting points for marketing processes, with explicit differences between platforms connected now, compatible Adobe products and components scheduled for later availability.