Adobe retired Experience Cloud. On April 20, 2026, at Adobe Summit, the company replaced its flagship marketing platform with a new product called CX Enterprise. The architecture is agent-first. The headline capability is a persistent AI agent class called Coworkers. The name reflects the design.

A Coworker, in Adobe’s model, is a persistent agent that runs continuously, triggers from signals or schedules, learns from outcomes over time, and orchestrates work across multiple systems and other agents. It operates toward goals. You tell it what you want to achieve. It figures out how to get there.

This is the company that runs marketing operations for roughly 75% of the Fortune 100. When it rebuilds its core product around agents, that signals where the operating model of enterprise software is heading.

What Actually Changed

Experience Cloud was a suite. It had products for analytics, audience management, campaign execution, content management, and personalization. You bought the tools. You staffed teams to run them. The tools executed what humans designed.

CX Enterprise is structured differently. It organizes around three business outcomes: brand visibility, customer engagement, and content supply chain. The platform deploys agents that work toward those outcomes autonomously within defined guardrails.

Adobe built two oversight models into the system. “Human-in-the-loop” requires a person to approve an agent’s work before it executes. That is the model for campaign planning, where judgment and creativity still matter. “Human-on-the-loop” lets agents operate autonomously within guardrails while humans monitor results. That is the model for consumer-facing interactions where speed matters more than pre-approval.

The platform is built on open standards: Model Context Protocol (MCP) and Agent-to-Agent (A2A) protocol. That means Adobe’s agents can communicate with agents from other platforms. Reference architectures already exist for Microsoft Copilot, ChatGPT Enterprise, Claude, and Gemini Enterprise. The five largest agency holding companies, including WPP, Publicis, and Omnicom, are integrated at launch.

This creates an agent network across platforms.

Why the Name Matters

Adobe could have called them assistants. Or copilots. Or automations. They called them Coworkers. That word choice reflects a specific architectural decision: these agents participate in shared workflows. They have persistent memory. They learn. They coordinate with other agents and with human team members toward shared goals.

The distinction between a tool and a coworker changes how you design work.

When you think of AI as a tool, you assign it tasks. One at a time. You check the output. You move on. The tool has no memory of what it did yesterday. It has no understanding of what you are trying to accomplish this quarter. Every interaction starts from zero.

When you think of AI as a coworker, the architecture shifts. The agent holds context. It knows the campaign history, the brand guidelines, the audience segments, the performance benchmarks. It triggers actions based on signals it observes. It escalates when something falls outside its parameters. It improves as it accumulates experience.

Adobe’s 2026 AI and Digital Trends Study found that 75% of organizations cite data integration and quality as their top AI implementation challenge. The Coworker model addresses this directly by giving agents persistent access to unified data across the platform, so humans can work from shared context across tasks.

What This Means for the Rest of the Market

Adobe joins other enterprise software companies shipping AI agents. Salesforce has Agentforce. ServiceNow has AI Agents. Microsoft has Copilot Studio. OpenAI shipped workspace agents into ChatGPT for Business on April 22.

Adobe retired the old product and replaced it with the agent-first version, rebuilding the platform around a different way of working.

That is a stronger signal than any feature announcement. It makes the agent model structural.

For business leaders watching this unfold, consider how many software platform transitions you will need to manage in the next 18 months as your vendors make the same move Adobe just made.

Gartner estimates that 40% of enterprise applications will embed AI agents by the end of 2026. MCP is already running on more than 10,000 enterprise servers with 97 million SDK downloads. The infrastructure layer is set. The platform vendors are rebuilding on top of it. The operating model is shifting from tools-you-use to agents-that-work-alongside-you.

If your team is still thinking about AI as something you add to existing workflows, Adobe just showed you what the replacement looks like. The workflows are being redesigned around agents that hold context, learn from results, and coordinate across systems.

The companies that figure this out first will gain speed and operate on a different model than the ones still running the old architecture.