Adobe

Adobe Shipped Its Agentic Stack. Someone Still Has to Run It.

At Summit in April, Adobe stopped positioning AI as something that helps your team work faster and started positioning it as something that does the work. In June, it shipped.

That shift gets discussed almost entirely as a software story. We want to talk about the part that wasn’t on a keynote slide: an agentic stack changes who you need on the team.

What Adobe actually shipped

On April 20, 2026 — the opening day of Summit’s main program in Las Vegas — Adobe introduced Adobe CX Enterprise, which it describes as an end-to-end agentic AI system bringing together AI agents, agent skills and Model Context Protocol (MCP) endpoints, with an intelligence and governance layer on top.

Announced alongside it:

Adobe named AWS, Anthropic, Google Cloud, IBM, Microsoft, NVIDIA and OpenAI as ecosystem partners, and said more than 20,000 global brands rely on its enterprise applications, with Adobe Experience Platform powering over a trillion experiences annually.

This is no longer a roadmap. On June 10, 2026, Adobe announced the general availability of CX Enterprise Coworker, describing it as an outcomes-based agentic AI solution that coordinates AI agents and workflows across analytics, content creation and journey orchestration — constantly monitoring performance signals, evaluating activities against defined objectives, and adjusting workflows. Anjul Bhambhri, Adobe’s SVP of Engineering for Customer Experience Orchestration, framed it as “reshaping workflows with agentic AI that is grounded in brand, customer and channel intelligence.”

A week later, on June 17, Adobe added Adobe Brand Visibility, pairing Semrush’s AI-visibility intelligence with Adobe’s optimization capability to track how brands appear across ChatGPT, Google AI Mode, Microsoft Copilot and Perplexity. Adobe cited nearly 300 million real-world AI search prompts behind it, and reported that AI-referred traffic to U.S. retail sites rose 1,324% between October 2024 and May 2026.

One clarification worth making, because recaps routinely blur it: Agent Orchestrator was not new this year. Adobe introduced it at Summit 2025. What changed in 2026 is that the orchestration layer grew into a full system — with governance, decisioning, and an execution layer sitting on top of it.

The part that isn’t a software problem

What follows is our read of the market rather than anything Adobe announced — worth stating plainly, because the two get conflated constantly right now.

Licensing an agentic platform is straightforward. Operating one is not. When software moves from recommending to executing, three things become true at once:

None of those are solved by the platform. They’re solved by people — and the people who do this work well have an unusual mix of skills.

The capability set we’re being asked for

Across the Adobe engagements we’re staffing now, demand has concentrated in five areas.

1. Agent orchestration engineering

Practitioners who understand MCP and A2A as integration standards rather than buzzwords — who can define agent skills, wire endpoints to real systems, and reason clearly about what an agent is permitted to do. This is the closest thing to a genuinely new role, and the pool is thin because the standards are young.

2. Context and data engineering

Real-Time CDP now unifies structured and unstructured data. Someone has to model that, keep identity resolution honest, and ensure the context an agent reasons over reflects reality. This is classic Adobe Experience Platform architecture work — with materially higher stakes, because the output is action rather than a report.

3. AI governance and brand safety

Adobe put an explicit governance layer in the stack, which implies someone owns it: approval paths, guardrails, audit trails, and a defensible answer to “why did the system do that?” We see this landing on people who came up through marketing operations and grew into risk.

4. Decisioning and measurement

Engagement Intelligence optimizes toward customer lifetime value. Optimizing toward CLV requires someone who can define it credibly and measure whether the machine actually moved it — a quantitative skill set many CX teams never had to hire for when the headline metric was open rate.

5. Generative engine optimization

With Brand Visibility and LLM Optimizer, Adobe made AI-search presence a product surface. The discipline is close enough to SEO to look familiar, and different enough to punish teams who treat it as SEO with a new name. Structured data, entity consistency, and machine-readable content stop being technical hygiene and start being demand generation.

Why this is hard to hire for

There is no established credential for most of this. The people who are genuinely good at it arrived from adjacent work — an AEP architect who moved into agent design, a marketing technologist who learned governance because someone had to. Titles lag reality by a year or two, which means keyword-matching résumés reliably surfaces the wrong people. We made a version of this argument before the stack shipped, in The Adobe AI Specialist Doesn’t Exist Yet; the 2026 releases turned it from a prediction into a staffing problem teams are handling this quarter.

How we approach it

Focus GTS works at the execution layer of Adobe Experience Cloud — we build on it, not just recruit for it. We’ve open-sourced MCP servers for Adobe Experience Platform, Firefly Services and Edge Delivery Services, so when a client describes an agent integration problem, we have usually hit it ourselves.

That is the difference we would point to: we assess people on the capability rather than the title, because we know what the work involves. If you’re standing up an agentic Adobe practice and trying to work out what to hire before the labels settle, we can help you scope it.

Ready to move faster?

Talk to Focus GTS about the right team for your Adobe, AI, Martech, or Cloud roadmap.

Explore Navigator Talk to us