It feels like everyone is talking about “going headless” right now. AI is quickly becoming the default user experience. Nobody wants to login to a portal anymore and navigate screens. The expectation is shifting from “go to the system” to “have the system come to you.”
Behind the scenes, things like agent-to-agent (A2A) interactions and protocols like MCP are starting to make this possible. Agents are no longer just answering questions, they can coordinate work across systems, carry context, and most importantly drive outcomes.
Traditionally, the term “headless” meant separating the UI from the platform using a custom front end built in React or another framework. As users become more comfortable with AI agents like Claude and Copilot, the role of the UI starts to change. It becomes less about navigating step-by-step workflows and more about supporting moments where structure and visibility still matter. For example, complex tasks, reporting, dashboards, and audit trails.
But what’s interesting is that for Pega users, this idea isn’t new. Pega’s Center-out architecture has always been about separating the channel from the core logic. Web portals, mobile apps, 3rd party apps, each one is just another way to access the same processes and decisions. AI Agents are no different. They’re simply the next channel. While the technology may be new, channel independence has been a core part of Pega’s strategy for years. It’s just the interface is finally catching up to the architecture.
Totally agree with you here, Chris. Though I think “headless” feels natural to Pega folks because of the workflow layer underneath that keeps everything consistent and predictable. At the end of the day, it’s not just about the channel, it’s about having governed processes that deliver the same outcome no matter how the user engages. That’s the real objective of headless, not just the ability to create some perfect UI/UX, or “Having the system come to you.” Who cares where the system is in relation to them if it doesn’t work well or consistently? Great post, got me thinking!
Imagine I’m using Claude as a front-end system to ask Pega questions (no need to imagine anymore with MCP). The problem I would sometimes have, is not knowing what to ask when the navigation to the information is missing. When I go to a Pega portal, I am instantly reminded I can see work from other users in my team. This can also be implemented with a Web Embed or a React front-end.
The distinction I’m making is that while the latter is also headless, not all headless channels are created equal. My LLM-front end is bound by its context window, and I may quickly lose context if it becomes an infinitely scrollable thread of questions and answers. It doesn’t mean it can create the value I need, but it does mean that I would put the Pega Portal with Web Embed and React Portals in one corner, and LLM-front-ends in the other.
Is AI is quickly becoming the default user experience? Maybe. If it’s becoming an extension of how we use AI personally, it may start to feel more natural in a working environment. Time will tell
I totally agree that being able to interact with Pega processes through MCP and A2A interfaces opens up a whole world of channel interactions!
Like Bas, I am also a bit doubtful if users will really benefit from navigating a complex process through a fully text-based interface.
However, for quick self-service style interactions that do not need a lot of process-context awareness from the user, having the option to offer those as a tool to whatever agent interface the user works with, will be powerful!
