Someone still has to spot the opportunity, find the right content, review the data, make the update, route it for approval, publish it, and then figure out whether it worked.
That is where Agentic CMS starts to feel different.
It's more than isolated AI tasks
This is not just about using AI to write a headline, summarize a page, or suggest metadata. Those capabilities are helpful, but they are still isolated tasks.
It is a little like improving one station on an assembly line. The work at that station may move faster, but the product still has to pass through every other step before it reaches the customer.
Agentic CMS starts connecting the entire line. It can help identify what needs attention, recommend what should happen next, prepare the work, support approvals, and help teams improve the experience over time.
Where Mark comes in
And this is where Optimizely's Agent Platform, Mark, gets really exciting.
Mark is not simply a writing assistant sitting next to the content editor. It helps connect the work happening across Optimizely, from content and data to experimentation, personalization, and analytics. It can help teams create, analyze, coordinate, and move from an idea to an actual experience much faster.
That is a much bigger deal than generating another content variation.
The real problem: disconnected instruments
Think about a typical content optimization process. A team may need to review search performance, identify pages losing visibility, compare competitor content, improve metadata, update copy, add structured data, confirm brand alignment, publish the changes, and monitor the results.
Today, that work may span several teams, tools, spreadsheets, and tickets. The problem is rarely that the organization lacks technology. More often, the data, ownership, processes, and goals are disconnected.
Most organizations already have the instruments. They have a CMS, analytics, customer data, experimentation, personalization, recommendations, and talented teams.
The challenge is that those instruments are not always playing from the same sheet of music.
One team may be optimizing content while another is building a personalization experience. A third may be running an experiment, while the analytics team is measuring something slightly different. Everyone may be doing good work, but the pieces are not always moving toward the same outcome.
Mark has the potential to act as the conductor, helping those capabilities work together around a shared goal.
Why the workflow has to come first
But adding AI to a disconnected process does not automatically fix it. It may only help the organization create disconnected work faster.
That is why the workflow matters first.
With Mark, teams can start with the business problem and then use agents where they are most useful. One agent might identify declining content. Another could evaluate SEO and generative engine visibility. A third could recommend changes based on audience needs and brand standards. Another could prepare those updates inside the CMS, while human reviewers still control approvals and publishing.
Over time, the workflow could monitor performance and use those results to guide the next recommendation.
What Agentic CMS looks like in practice
That is what Agentic CMS looks like in practice. It is not one magical agent doing everything. It is a connected set of capabilities helping teams move through the work with less manual effort.
It can help teams identify outdated or underperforming content, research audience and search needs, recommend updates, generate approved variants, apply metadata and tagging, prepare content in the CMS, route work through governance, monitor performance, and recommend the next action.
Beyond content, across the ecosystem
The best part is that this does not have to stop with content.
Because Mark works across the Optimizely ecosystem, those workflows can connect CMS, experimentation, personalization, analytics, customer data, recommendations, and campaign operations.
Imagine an agent recognizing that a high-value audience is engaging with a certain topic, finding the most relevant content, identifying a gap, proposing a personalized experience, and recommending an experiment to validate the idea.
The team is still setting the strategy. The team is still deciding what should reach the customer. Mark is reducing the research, coordination, and repetitive execution required to get there.
Here's what that looks like at Velir
We built an Automated Cohort Segmentation workflow to help solve one of the hardest parts of personalization: figuring out who to personalize for and pulling together the data needed to make that decision. Every 24 hours, agents review enriched customer profiles in ODP and other connected sources. They look at behavioral activity, firmographic and demographic data, CRM context, content interests, topical alignment, and data from the warehouse.
From there, the workflow scores different audience cohorts and gives the marketing team a clear view of which audiences are showing the strongest signals and where there may be an opportunity for personalization or experimentation. The marketing team still reviews the recommendations and decides what should move forward. Once approved, those audiences can be created in ODP and used within Web Experimentation and personalization.
The biggest benefit is the time saved. Instead of asking marketers to manually dig through multiple sources to find possible audiences, the workflow helps surface the audiences that are already showing meaningful engagement or intent. It can also help identify the content and topics most relevant to those audiences, making it easier to decide what experience to test or use to nurture them.
A better GPS, not a self-driving car
I think of it less like a self-driving car and more like a much better GPS.
The team still chooses the destination. People still decide which route makes sense and whether the recommendation is appropriate. Mark helps surface obstacles, identify faster paths, and adjust when the conditions change.
That distinction matters because the best use of AI is not to remove people from digital experience creation. It is to give them more time for the work that requires context, creativity, empathy, prioritization, and judgment.
You can't automate a broken process
Agentic CMS is also not a switch that organizations can flip to fix broken content operations.
Automating a broken workflow is like paving a cow path. The work may move faster, but it is still following the wrong route.
If ownership is unclear, data is unreliable, approvals take six weeks, or teams do not agree on the goal, adding agents will not solve those problems automatically.
Organizations still need governance, brand instructions, data permissions, quality standards, review processes, and clear accountability for agent actions.
Measuring what actually matters
They also need the right success measures.
The number of assets generated or tasks automated may show activity, but it does not prove customer or business value. Agentic CMS should be measured through faster time to market, stronger content quality, improved visibility, more relevant experiences, higher engagement, increased conversion, and less manual effort.
The shift that matters
The CMS has always helped teams publish.
With Mark, it can also help them understand what is happening, decide what to do next, and continuously improve what they put into the world.
That is why I think Agentic CMS is such an important shift. It is not about handing the website over to AI. It is about giving teams a smarter way to identify opportunities, complete repetitive work, and learn from what happens next.
AI should not just help us create more.
It should help us create better, act faster, and make every next step more informed.
See what agentic CMS looks like for your team
The shift from disconnected tools to connected, agent-powered workflows does not happen overnight, but a conversation is a good starting point.
As an Optimizely Platinum Solution Partner, Velir works with marketing and content organizations every day to help them figure out where agents can have the most impact, what workflows are worth connecting first, and how to build toward it without disrupting what is already working.
If any of this resonated, we would love to talk.