In thirty years of building digital experience platforms, I've watched vendor after vendor make the same mistake: betting everything on their own tech stack when the landscape shifts beneath them every six months. Now that landscape is shifting faster than ever and impossible to pick a winner.
Sitecore just made an AI pivot. And it may be the smartest platform decision I've seen from a DXP vendor in years.
At Symposium 2025, Sitecore didn't just announce another AI assistant bolted onto a content management system. They unveiled SitecoreAI with Agentic Studio at its core, running on Microsoft Azure and leveraging Microsoft's Agent Framework for orchestration. Mo Cherif, Sitecore's VP of AI, put it plainly: the partnership gives Sitecore "the flexibility to power fully non-deterministic agentic orchestrations and the reliability to run more deterministic, repeatable agents" while benefiting from Microsoft's enterprise-grade observability and telemetry.
That's not marketing speak. That's an architecture decision with real consequences.
In a recent webinar, Mo Cherif was refreshingly candid about what Sitecore had built with the initial release of Agentic Studio: "Too complex for marketers, but too simple and not extensible enough for developers." That's the trap most DXP vendors fall into when they try to build AI capabilities in-house. Microsoft Foundry levels Sitecore up immediately by providing enterprise-grade orchestration without forcing them to reinvent infrastructure that Microsoft has already solved.
The Problem With Building Your Own AI Stack
Most DXP vendors right now are doing one of three things, and all of them create problems down the road.
The first approach is building proprietary RAG systems tightly coupled to specific models. This works until the model you've built around gets superseded or the provider changes pricing. Then your customers are stuck with yesterday's intelligence powering tomorrow's content operations.
The second approach is creating deep integrations with a single AI provider. Adobe and Firefly. Optimizely and their AI Orchestration wrapper on Google Gemini. These work within their ecosystems but leave customers locked in when the next generation of models arrives from a different provider.
The third approach, and perhaps the most dangerous, is treating AI as a feature checkbox. Add some content generation here, a chatbot there, call it "AI-powered" and move on. This approach ignores that agentic systems require fundamentally different architecture than simple prompt-response integrations.
Sitecore's move to Microsoft Foundry sidesteps all three traps.
What Microsoft Foundry Actually Brings
Let me be clear about what Microsoft Foundry is, because the name change from Azure AI Foundry has created some confusion.
Foundry is Microsoft's unified platform for building, deploying, and governing AI agents at scale. It evolved from a model experimentation environment into what Microsoft now calls an "agent factory" where models, tools, policies, and orchestration work together under a single runtime.
The Foundry Agent Service manages conversations, orchestrates tool calls, enforces content safety, and integrates with identity and observability systems. For enterprise deployments, this means agents that are secure, compliant, and production-ready from day one.
More importantly, Foundry supports multiple model providers. Azure is now the only cloud platform offering both OpenAI's GPT series and Anthropic's Claude models side by side. The Model Router can automatically select the best or most cost-effective model for each task. This is the opposite of vendor lock-in.
The multi-agent workflow capabilities allow developers to design processes involving multiple specialized agents, either through visual tools or YAML definitions. One agent handles content analysis, another manages personalization rules, a third coordinates publishing workflows. They collaborate with persistent state, error recovery, and context sharing.
Why Extensibility Matters More Than Features
Here's what concerns me about most DXP AI implementations: they ship with a fixed set of capabilities determined by the vendor.
Sitecore launched with 20 pre-built agents covering common workflows like campaign planning, content migration, and bulk content generation. But the real story is Sitecore Studio, which lets customers and partners build their own agents.
At Symposium, organizations had already built a few simple baseline solutions using Studio for a hackathon. Horizontal Digital won the Partner Challenge with their Sense Agent, which analyzes customer reviews and maps sentiment against brand voice guidelines. One North created FigCore, which automates converting Figma designs into Sitecore components. Fairly simple implementations as a starting point, but shows the extensibility.
This extensibility model acknowledges something important: no vendor can anticipate every workflow their customers need. The organizations using Sitecore span retail, healthcare, financial services, and dozens of other industries. Each has unique processes, compliance requirements, and integration needs.
By providing the foundation for custom agent development rather than just pre-built tools, Sitecore enables solutions that actually fit enterprise reality.
The Composable Architecture Finally Delivers
I've been skeptical of composable architecture claims for years. Too often, "composable" meant "we have APIs" without real thought about how the pieces fit together.
SitecoreAI unifies what were previously separate products—XM Cloud, DAM, MRM, CMP—into a single platform with a shared data model. Roger Connolly, Sitecore's CPO, described the connected stack as feeding "what an agentic future actually could look like and help us orchestrate it."
The practical result is that agents can work across the full digital experience lifecycle without jumping between disconnected systems. A governance agent can audit assets in the DAM while a brand agent checks content in the CMS while a personalization agent adjusts targeting rules in the CDP. The Agentic Flows orchestrate these multi-step processes with full visibility for marketing teams.
This is what composable was supposed to mean: specialized components that work together intelligently, not a collection of tools that each require separate management.
What This Means for Sitecore Competitors
The enterprise DXP market just got more interesting.
Adobe has invested heavily in Firefly and their proprietary AI stack. This gives them control but also means customers are buying into Adobe's specific vision of AI-powered creativity. If the market moves toward open protocols and multi-model architectures, Adobe will need to adapt.
Optimizely recently overtook Adobe in Gartner's Magic Quadrant, partly on the strength of their unified data model. CEO Alex Atzberger emphasized that "the data model and the orchestration of the workflow are the two things that are going to be so, so important in the future." He's right about orchestration, but Optimizely's Opal is built around Google's Gemini. With Anthropic releasing Claude Opus 4.5 and OpenAI pushing ChatGPT 5.2, you have to question whether single-model architectures can keep pace with the relentless innovation cycle. Orchestration through an open framework like Microsoft Foundry may prove more adaptable than betting on any one provider.
Magnolia, Contentstack, and the headless CMS players have been building AI integrations, but most follow the RAG-plus-chatbot pattern rather than true agentic architectures. The gap between "AI-assisted" and "AI-orchestrated" will become more apparent as enterprise expectations rise.
The Real Test Is What Happens Next
I've seen enough platform announcements to know that roadmaps don't always match reality.
Sitecore's implementation is promising, but the proof will come over the next 12-18 months. Can the Agentic Studio scale to handle enterprise workloads? Will the partner ecosystem produce genuinely useful agents or just variations on the same themes? Will Microsoft Foundry continue evolving in ways that support Sitecore's architecture?
The XM Cloud customer upgrade path is seamless—no migration required, full data continuity, instant access to Agentic Studio. That's a strong start. But enterprise buyers will be watching how the platform handles the messy realities of large-scale content operations.
What I'd Recommend
If you're evaluating DXPs right now, the Sitecore announcement changes the conversation.
For existing Sitecore customers on XM Cloud, explore what Agentic Studio can do for your specific workflows. The pre-built agents are a starting point, but the custom agent capabilities through Studio may deliver more value for complex operations.
For organizations considering a platform change, ask every vendor about their AI architecture. Not features, architecture. How do they handle model selection? What happens when new models emerge? Can you build custom agents for your specific workflows? How do agents integrate with your existing systems?
For anyone building AI into their content operations, pay attention to open standards. MCP adoption is accelerating. Microsoft Agent Framework, Semantic Kernel, and the tools being developed under the Agentic AI Foundation represent the infrastructure layer for the next generation of enterprise AI.
The Foundation You Choose Today
I've watched platforms rise and fall for three decades. The ones that endure share a common characteristic: they adapt to change without forcing customers to start over.
Sitecore's bet on Microsoft Foundry and open protocols positions them to evolve as the AI landscape continues shifting. They're not building a walled garden. They're building an extensible foundation.
For enterprise organizations making platform decisions, that difference matters. The content management system you choose today will be running for five, maybe ten years. The AI capabilities of 2026 will look primitive by 2030. The question isn't what your platform can do now—it's whether the architecture supports what you'll need to do then.
Sitecore seems to understand that. Whether they execute on the vision remains to be seen. But the architectural choices they've made are, in my experience, the right foundation for what's coming next.
