What changes when your CMS stops dictating how and what you build
Teams on Adobe have put up with a rigid, locked-in platform for years. Agents are bringing it to a head.

Hollie Aghajani
Staff Product Marketing Manager
Published
A product manager at a large telecom enterprise spent the better part of three months trying to get agents working against Adobe Experience Manager (AEM). He wanted an agent to take a ticket, fix a layout bug, run the tests, and confirm the fix held. This wasn’t a novel idea, as his developers were already handing that kind of work to agents in other parts of the stack.
Well this didn’t work, and it wasn’t the agent’s fault. The agent had two ways in and both were dead ends. It could scrape the rendered page, which burned through tokens and showed it the result rather than the thing that produced it. Or it could dig into the repository underneath, which burned just as many and made less sense. What the agent needed was a way to hook into the CMS, change one component and confirm the change took. In the product manager’s experience, AEM kept control of layout and presentation, so the agent had no clean way to do either.
In his summary: he spent a lot of time and tokens and had nothing to show for it. Then he ran the same loop against the
The workarounds that came first
Teams have been working around the same limit for years. But from the inside, it looked more like a process problem, not an architecture one.
Before moving to Sanity, a web content manager at a national bridal retailer would update the homepage every week and built new pages whenever a store opened. Each product feature was a title, an image, and a link she typed in by hand, with nothing tying it to the product itself. When a price or name changed, she found every page it appeared on and did it again.
Tata Digital hit the same limit on the developer side. Building one app out of 20+ brands on AEM, teams modeled content by hand in the UI and passed schemas around as zip files, which got harder to govern as the team grew from 20 to 100 developers. After a year, the app still wasn’t ready.
In each case a person stood in for something AEM couldn’t hold, and after a few years the adaptation became the process.
Start from the content, not the page
When AEM came out of the early 2000s, a website meant pages and the only reader was a person with a browser. Putting the page together was the job, and it did that well enough. The side effect is that content ends up organized by where it shows up, not by what it is.
Sanity starts from the other end. Your products, regions, and brands are the content itself. A page is one of several ways to show them. Your team builds the app or frontend however it wants, in whatever framework it prefers. Your
The real cost of running AEM
Each of those workarounds has a price, and most of it won’t show next to the license. The license is only one part of what AEM costs to run. Specialist skills come next: when routine changes depend on a small pool of developers who know the platform, every change costs more and takes longer.
The hardest cost to see is time. A schema passed around as a zip file or a price retyped by hand won’t appear on an invoice, but each one pushes back the next launch, AI experiment, or new market rollout.
Decide what you’re rebuilding
Teams on Adobe are weighing an upgrade or a move, and most of the effort goes to the part that’s easy to plan: moving pages. That part is manageable. What the content looks like after the move decides whether an agent can use it.
If the new setup still organizes content around the page, the platform keeps deciding how your content is shaped, and the next agent project will hit the same wall the product manager did. So the question before you scope anything is what you’re rebuilding: the same page-first structure, or content modeled by what it is, with agents able to read and change it directly.