Every CMS vendor in 2026 wants you to believe their platform is “agentic.” The word has become the industry’s favorite adjective, applied to everything from genuinely autonomous workflow engines to what amounts to a chatbot bolted onto a content editor. The difference matters, because one of these things changes how your team operates and the other just changes your vendor’s positioning slide.
Let’s sort out which is which.
Why “Agentic” Became the Buzzword of the Year
The timing isn’t accidental. Gartner expects 40% of enterprise applications to include task-specific AI agents by the end of 2026, up from less than 5% in 2025. McKinsey reports that 62% of organizations are already using AI agents in some capacity. The market signal is clear: enterprises aren’t interested in AI that waits to be prompted. They want systems that can observe, decide, and act.
CMS vendors read the same reports. Within roughly 18 months, Kontent.ai declared itself “the world’s first Agentic CMS.” Optimizely launched Opal, an AI agent orchestration platform embedded across its product suite. Sitecore rebranded to SitecoreAI and introduced Stream as an orchestration engine with specialized agents. Adobe built an Agent Orchestrator into Experience Platform, deploying purpose-built agents across AEM for content production, experience modernization, and development. Storyblok introduced Strata and FlowMotion. Kentico launched ARIA and KentiCopilot.
Meanwhile, platforms like Sanity took a different path entirely, building an MCP server and Agent API that let external AI agents interact with structured content through an open protocol rather than embedding proprietary agents into the CMS itself.
Everyone is claiming the agentic label. But not everyone means the same thing by it.
What “Agentic” Actually Means in Computer Science
Before evaluating vendor claims, it helps to establish what the term actually describes. IBM defines agentic AI as systems designed to “autonomously make decisions and act, with the ability to pursue complex goals with limited supervision.” MIT Sloan describes AI agents as “a new breed of AI systems that are semi- or fully autonomous and thus able to perceive, reason, and act on their own.”
The academic literature identifies several characteristics that distinguish genuinely agentic systems from AI features that simply respond to prompts.
Autonomous goal pursuit. An agentic system receives a high-level objective and determines its own path to achieve it. It doesn’t need step-by-step instructions. It breaks complex goals into subtasks, sequences them, and executes.
Environmental perception. The system observes its operating environment, gathering data about current state, available resources, and constraints. In a CMS context, this means understanding what content exists, what’s missing, what’s outdated, and what workflows are active.
Continuous operation. Unlike prompt-and-response tools that activate only when a human asks them something, agentic systems run persistently. They monitor, detect, and act on predefined triggers without waiting for manual intervention.
Adaptive learning. The system refines its behavior based on outcomes. If an action doesn’t produce the expected result, it adjusts its approach. Over time, it becomes more effective at achieving its goals within its specific operating context.
Multi-step reasoning. Rather than handling isolated tasks, agentic systems chain actions across multiple steps and systems, maintaining context and state throughout a complex workflow.
This is a high bar. Most of what CMS vendors call “agentic” in 2026 falls short of it.
The Three Tiers of AI in Content Management
When you strip away the marketing language, CMS AI capabilities in 2026 sort into three distinct tiers, and only the third one is genuinely agentic.
Tier 1: Generative AI Features (AI-Assisted)
This is where most platforms started and where many still primarily operate. The CMS includes AI-powered features that respond to user prompts: generate a draft, suggest a headline, summarize a document, translate content, optimize for SEO. The human initiates every action. The AI produces output. The human reviews and approves.
This is valuable. It accelerates individual tasks. But it’s not agentic. It’s a tool. The distinction matters because a generative AI feature doesn’t observe the environment, doesn’t pursue goals autonomously, and doesn’t chain operations across systems.
Tier 2: AI Workflow Automation (AI-Augmented)
At this level, AI is embedded into workflows with predefined triggers and rules. Content published in English automatically triggers translation into three other languages. A new product page automatically generates associated FAQ content. Content that hasn’t been updated in 90 days gets flagged for review and routed to the appropriate editor.
This is meaningfully more capable than Tier 1. The automation runs without manual initiation. But most implementations are still rule-based orchestration with AI performing specific tasks within each step. The system follows a predetermined workflow. It doesn’t decide what the workflow should be.
Tier 3: Agentic Content Operations (AI-Autonomous)
At this level, the AI system operates with genuine autonomy within defined boundaries. Given a goal like “ensure our product documentation is current across all markets,” the system determines which content is outdated, prioritizes updates based on traffic and business impact, drafts revised content, routes it through the appropriate approval workflow, and monitors the results. If translation quality drops in a particular market, it adjusts its approach.
The system isn’t following a script someone wrote in advance. It’s pursuing an objective and making decisions about how to get there. Humans set the boundaries, define the goals, and review critical outputs. But the operational decision-making sits with the agent.
Evaluating the Vendor Claims
With that framework in mind, let’s look at how the major CMS platforms actually map.
Kontent.ai: The Loudest Claim
Kontent.ai has been the most aggressive in claiming the agentic label, declaring itself “the world’s first Agentic CMS” when it launched its AI Agent in October 2025. In March 2026, it added Expert Agents that it describes as purpose-built agents running continuously across content workflows.
The claimed capabilities are real and worth acknowledging. Sixty organizations are actively using the platform, with reported results like 70% reduction in draft creation effort and 134 content pieces in two languages produced in a single day. The system operates through two layers: a Main Agent that allows natural language control of the platform and Expert Agents that automate specific workflow patterns.
Where does this land on the three-tier framework? Kontent.ai appears to genuinely operate at the boundary between Tier 2 and Tier 3. The Expert Agents run continuously on predefined triggers and execute multi-step operations, which moves beyond simple workflow automation. The platform’s emphasis on governance and permissions shows architectural thinking about agent boundaries. But much of what’s demonstrated publicly still centers on content creation acceleration rather than autonomous goal pursuit across the full content lifecycle.
The “world’s first” claim is marketing. The underlying capability is legitimately further along than many competitors.
Adobe Experience Manager: The Enterprise Heavyweight Approach
Adobe has taken the broadest approach, building an Agent Orchestrator into Experience Platform that powers specialized agents across the entire Experience Cloud. For AEM specifically, the Brand Experience Agent comprises three sub-agents: an Experience Modernization Agent for migrations, an Experience Production Agent for content updates and form creation, and a Development Agent for troubleshooting and build automation.
Adobe’s approach sits firmly in Tier 2 with elements of Tier 3. The agents are purpose-built for specific operational categories, and the orchestration layer coordinates between them. The migration agent in particular shows genuine environmental perception: it analyzes existing site structure, identifies content patterns, and proposes migration strategies. But the system still requires significant human configuration and oversight for each deployment.
The distinguishing feature of Adobe’s approach is scale. These agents operate across a massive platform ecosystem with deep integration into Creative Cloud, Workfront, and the broader marketing stack. For organizations already embedded in Adobe’s ecosystem, the agentic capabilities extend across more operational surface area than any competitor.
Optimizely Opal: The Marketing Operations Focus
Optimizely’s Opal is positioned as a conversational AI interface embedded across the entire Optimizely One suite. Recent demonstrations show it executing research, analyzing documents, orchestrating tools, and automating workflows across CMS, experimentation, analytics, and content marketing.
Opal operates primarily at Tier 2, with its strongest agentic characteristics appearing in experimentation workflows where the system can autonomously design, run, and interpret tests. The conversational interface is well-executed but fundamentally prompt-driven, meaning human initiation is required for most operations. Where Opal genuinely differentiates is in connecting AI actions across content creation, testing, and analytics in a single operational flow.
SitecoreAI: The Orchestration Bet
Sitecore’s rebrand to SitecoreAI in November 2025 signaled a wholesale commitment to AI-first positioning. Stream, the new orchestration engine, uses a “Brand Kit” RAG architecture to maintain brand consistency across AI-generated outputs. The Agentic Studio concept envisions centralized agent management.
SitecoreAI is currently between Tier 1 and Tier 2 for most production implementations. The architectural vision is ambitious, and the brand consistency approach through RAG is smart engineering. But the platform is still in the middle of a major transition, and many organizations are evaluating SitecoreAI alongside a migration from XP. The agentic capabilities are more roadmap than production reality for most customers today.
Sanity: The Protocol Approach
Sanity took a fundamentally different architectural bet. Rather than building proprietary agents into the CMS, Sanity built a Content Operating System with an MCP server, Content Agent, Functions, and Agent API that allow any AI agent to interact with structured content through an open protocol.
This is a philosophical distinction with practical implications. Sanity’s approach says the CMS shouldn’t try to be the agent. Instead, the CMS should be the structured content layer that agents of all kinds can read from and write to, with full awareness of the content schema. External agents in tools like Claude Code, Cursor, or custom workflows can query content, manage releases, patch documents, and deploy schemas because the structured content model makes the data inherently machine-readable.
Agent Context, Sanity’s latest feature, compresses the content schema so agents understand the data model itself, not just individual documents. This means agents can translate natural language questions into precise queries against the actual data structure.
Where does this sit in the tier framework? Sanity enables Tier 3 behavior without embedding Tier 3 agents. The platform provides the structured foundation and open protocol that genuinely agentic systems need to operate, while letting organizations choose their own agents and orchestration layers. It’s a bet that the content layer and the intelligence layer should be decoupled.
The Architecture Question Nobody Is Asking
Here’s what most vendor evaluations miss: the agentic capability of a CMS matters far less than whether its content architecture can support agentic operations.
An AI agent can only be as effective as the data it has access to. If your CMS stores content as page-shaped HTML blobs without semantic structure, defined relationships, or machine-readable metadata, no amount of agentic AI bolted on top will fix the fundamental problem. The agent can’t reason about content it can’t parse. It can’t make decisions about content relationships it can’t see. It can’t maintain governance over content structures it doesn’t understand.
This is why content architecture and agentic capability are inseparable concerns. A CMS with a rich, typed content model, structured text, and explicit relationships between documents gives any agent, whether built into the CMS or operating externally through a protocol like MCP, the foundation it needs to actually do useful work.
A CMS that stores everything in a single rich-text field and then adds an AI chatbot to the editing interface has an “AI feature.” It does not have agentic content operations.
Five Questions to Ask Before You Buy the Hype
If you’re evaluating platforms and every vendor is claiming to be agentic, these questions will help you separate genuine capability from positioning.
Does the AI operate continuously, or only when prompted? A truly agentic system monitors, detects, and acts on conditions without someone typing a request. If every AI interaction starts with a human asking for something, you’re looking at a Tier 1 feature, not an agent.
Can the system pursue goals across multiple steps and systems? Ask for a demonstration of the AI completing a multi-step objective that spans content creation, governance, publishing, and measurement. If the demo shows individual AI-powered tasks but not connected autonomous workflows, the integration work still falls on your team.
What happens when the AI makes a mistake? Agentic systems need governance guardrails. Ask how the platform handles agent errors, what approval gates exist, what actions are autonomous versus supervised, and what the audit trail looks like. If the vendor can’t clearly articulate the boundary between agent autonomy and human oversight, the system isn’t production-ready.
Is the content architecture structured enough to support agent reasoning? Ask to see the content model. If the agent is operating on unstructured rich-text blobs, its reasoning capability is fundamentally limited. If it’s operating on typed, fielded, relationship-aware structured content, it has a foundation for intelligent action.
Are you locked into the vendor’s agents, or can you bring your own? Some platforms embed proprietary agents that only work within their ecosystem. Others provide open protocols that let you use any AI agent. This is a strategic question about vendor dependency and long-term flexibility, especially as the agent landscape is evolving rapidly.
What Actually Matters Right Now
The honest assessment of the CMS market in April 2026 is this: genuinely agentic content operations are still early. The most advanced implementations are demonstrating real value in specific workflow categories like draft acceleration, translation automation, and compliance checking. But the vision of fully autonomous content operations that reason across the entire content lifecycle, learn from outcomes, and continuously optimize without human intervention? That’s still being built.
What you can do today is ensure your content architecture is ready for it. Structured content models, typed fields, defined relationships, open protocols for agent access, and clear governance frameworks are the foundation that every form of agentic AI needs to function. Without that foundation, no amount of vendor branding will make your content operations intelligent.
The platforms that will win the agentic era aren’t necessarily the ones with the flashiest AI demos today. They’re the ones whose content architecture was designed to make content understandable to machines from the ground up. Because the best automation doesn’t fight the data structure. It works with it.
