Why you can't vibe code your brand

Kevin LiKevin Li
May 19, 2026

Brand is an institutional decision, not a prompt. Kevin Li explains why scaling enterprise content needs an agentic CMS — not a vibe-coded workaround.

Here's a number that should bother you: the average enterprise content production cycle takes weeks. A brief, a writer, a review, a round of edits, a legal sign-off, a publish, a wait for results. Weeks. For a single page.

Meanwhile, somewhere in your organization, a marketer just proposed building 2,000 individualized landing pages for your next ABM push. And everyone in that room laughed. Because of course you can't do 2,000 pages. You can barely do five.

That's the impossible I want to talk about.

The thing about impossible

"Impossible" is a moving target. It used to be impossible to stream a movie. To get in a stranger's car. To have a real conversation with a machine at 2am about why your landing page copy isn't converting.

In the CMS world, impossible has always meant the same thing: too many people, too many handoffs, too much coordination to turn a good idea into a live experience before it loses its edge.

I think about it like Thanksgiving. You've got 20 people coming over. The day before, you run out to buy everything. Something's running low, you forgot something, etc. You preheat the oven. You grab the fancy dishes from the attic. You make food for 40 because you can't risk running short, and then seven people actually show up. All of that — every bit of it — is sludge. The reviews, the approvals, the handoffs, the guesswork. The actual meal gets made eventually, but you've spent more time managing the process than cooking the food.

That's what content production looks like at most enterprises right now. And it has nothing to do with a shortage of good ideas.

This is what we decided to fix. Not with incremental improvements to an existing workflow. With a genuinely different architecture.

In the CMS world, impossible has always meant the same thing: too many people, too many handoffs, too much coordination to turn a good idea into a live experience before it loses its edge.

Kevin Li|SVP, Product, Optimizely

The answer isn't just speed — it's pairing AI's scale with the institutional knowledge of what 'good' actually looks like for your business.

Why we went agentic, and what that actually means

When we talk about an agentic CMS, we're not talking about a writing assistant. We're not talking about a smarter headline suggestion or an AI that reformats your copy for mobile.

We're talking about agents that use software the way it was always meant to be used: at scale, at speed, without the sludge in between.

Jensen Huang at NVIDIA made a point that stuck with me. In the future, it's not primarily humans using software. It's agents using software. More efficiently. At greater scale than any human team can. Our job is to make sure that when those agents show up, they have the right raw ingredients to work with. And here's the thing: you can have all the right market conditions, all the AI momentum in the world — but if you don't have the ingredients, it's going to be a pretty crappy dish.

We have the ingredients. A mature CMS. An agent orchestration platform that operates as a genuine standalone layer, not a set of bolted-on features. A delivery infrastructure that closes the feedback loop between what you publish and how it performs. A planning and workflow layer that keeps humans in the loop when it matters.

That combination is what makes the impossible achievable. Gartner predicts that 40% of enterprise applications will be integrated with task-specific AI agents by the end of 2026, up from less than 5% in 2025. The infrastructure to support that isn't optional. It's the whole game.

40%

of enterprise applications will be integrated with task-specific AI agents by the end of 2026, up from less than 5% in 2025

The 800-page problem

Let me give you a concrete example of what "impossible" actually looks like in practice.

We had 800 organizations attending a major industry summit. Everyone agrees that ABM-personalized pages drive better engagement. Nobody disputes that. The disagreement is always about what's feasible.

The old math goes like this: 2,000 target accounts is too many pages to build. So, you bucket them by industry. Thirty-seven industries is still too many. Top five? Maybe you can manage that. And you end up with five landing pages that are maybe 15% more relevant than a generic one.

The new math: 800 organizations. 800 individualized pages. Each one researched and built at a quality level that would take a human content writer hours to replicate per account.

We didn't generate text and call it done. Agents assembled structured content into actual pages using our Page Builder. Brand guardrails built in. Account-specific information pulled automatically. Published through a workflow a human can review, approve, and push live. The content creation and the page creation happened together, in the same system, at a scale that was previously a punchline in a planning meeting.

Page creation at scale is what this is. The difference matters.

...The content creation and the page creation happened together, in the same system, at a scale that was previously a punchline in a planning meeting.

Kevin Li|SVP, Product, Optimizely

Give every visitor a 1:1 experience, every time

Just let agents run and watch conversions grow.

Why you can't just vibe code it

At some point in every conversation about agentic content, someone asks the obvious question: can't I just use a general-purpose AI and build whatever I want?

It's a fair question. And the honest answer is: for some things, yes. For an enterprise brand trying to produce what I just described, no. For a few very specific reasons.

1

Brand is an institutional decision, not an individual one.

Here's a definition I've been using lately: a brand is a set of opinions for how an institution wants to carry and express itself. When content is on brand, it means the opinions are consistent with how the organization wants to represent itself. When it's off brand, it means someone (or something) made a unilateral call they weren't authorized to make.

You can't vibe code brand. An individual might have strong opinions. A general-purpose AI will make its best guess. But brand is governed by the institution. That's why we have brand teams, approval workflows, structured content types, and Page Builder templates that bake those guardrails in at the infrastructure level. When agents are creating pages at scale, they're operating within a system that carries the brand's opinions — not improvising from a blank prompt. Without that layer, you're not doing content at scale. You're doing content chaos at scale.

2

Web performance is a physics problem.

I had a conversation with one of our engineering leads not long ago. I said: 200 milliseconds — that's the round-trip expectation for a browser to load a page. How close can you get me to generating a fully assembled, personalized page in that window? He said 30 seconds. And I said: do you know how big of a difference that is? Have you ever waited 30 seconds for anything on a screen? That is an eternity. That's someone closing the tab.

Getting personalized pages to load at the speeds modern experiences demand, for humans and AI retrieval agents alike, requires owning the delivery layer. You can't patch that together from outside tools.

3

Collaboration doesn't happen in a single-player interface.

Anyone who's tried to share an AI-created artifact for review inside a general-purpose AI tool knows what I mean. There's no private sharing. No proper approval routing. No workflow that takes agent-created content from draft to reviewed to published inside a structure your legal, brand, and content teams trust. If your agentic CMS story depends on a third-party platform solving that problem for you, you're waiting in line behind every other priority that platform has decided matters more.

The dependency question

This brings me to the point I think is most important for anyone evaluating CMS vendors right now.

Some vendors have decided that their agentic story is really someone else's story. They're broadly compatible, their demos run beautifully inside third-party AI interfaces. That's a fine integration story. But it's not an agentic CMS story. When a vendor's intelligence layer lives inside a platform they don't control, they've outsourced their destiny.

Think about what that means in practice. If you need a collaborative review workflow for AI-generated content, you're waiting for that third-party platform to build it. If you need content assembled and published through an approval chain your organization actually trusts, you're dependent on a roadmap you have no influence over. If the third-party AI deprioritizes your specific use case — and trillion-dollar companies have a lot of other things to build — your agentic CMS capabilities stall.

I'll put it plainly: you cannot say "I am an agentic CMS" when you rely entirely on someone else to be the agent. It's the same move headless vendors pulled a few years ago. They gave up delivery and said they were still a CMS. But why don't you all chop off your heads and tell me if you're still human? I'm still human — I'm just a headless human. It doesn't work. Once you give up the sum of the parts, you can't claim the whole.

The vendors who went all-in on that path? They're a bit like getting a bag of Lego pieces with no box and no instructions. Great individual pieces, but no picture on the box. In an agentic world, the picture on the box matters more than ever.

We built our own agent orchestration platform because that dependency is not one we're willing to accept on behalf of our customers. When an agent creates a page, our workflow routes it for review. When a team needs to collaborate on AI-assembled content, our platform handles that. When brand guardrails need to govern what an agent can and can't produce, our system enforces it.

Open standards and broad compatibility matter. We're building for an interoperable world. But interoperability is different from dependency. You can connect to everything and still own your own intelligence. That's the position we're building from.

I'll put it plainly: [a CMS vendor] cannot say "I am an agentic CMS" when they rely entirely on someone else to be the agent.

Kevin Li|SVP, Product, Optimizely

What the market is getting wrong

Most CMS vendors are making incremental moves. Adding AI writing assistance to existing editors. Bolting translation features onto existing workflows. A headline suggestion here, a summarization tool there. Evolutionary steps dropped into a paradigm that hasn't fundamentally changed.

What they're missing is that agentic AI isn't an incremental opportunity. It's a step-function shift in what's architecturally possible. Headless CMS vendors gave up their delivery layer years ago and can't reclaim it without rebuilding from scratch. That layer is where the feedback signal lives. Without it, you're a backend that can't learn.

Adding AI features to a fundamentally constrained architecture is like slapping wings on a car. Maybe it gets a few inches off the ground, but it won't be a jet.

I look at where some of these vendors are sitting right now and I think about dinosaurs looking up at the sky going: is it just me, or is the sun getting bigger? Architecturally, there's nothing they can do about what's coming.

The market data reflects this urgency. Deloitte's 2026 State of AI in the Enterprise report — based on a survey of 3,235 business and IT leaders across 24 countries — found that just 23 percent of companies are currently using agentic AI at least moderately. Within two years, that figure is expected to reach 74 percent. The window to build the right architecture is not staying open.

Adding AI features to a fundamentally constrained [CMS] architecture is like slapping wings on a car. Maybe it gets a few inches off the ground, but it won't be a jet.

Kevin Li|SVP, Product, Optimizely
74%

of companies are expected to be using agentic AI at least moderately by 2028

From managing content to driving outcomes

Here's the shift that I think changes everything.

CMS vendors have always sold tools. We give you the platform. You figure out what to do with it. That made sense when the constraint was the tool. It makes less sense when the tool can now do almost anything.

The real question is: what business outcome do you actually want? For most B2B companies, the website has three jobs — book meetings, drive target account engagement, convert event registrations. That's it. Everything else is infrastructure in service of those outcomes.

The smarter play is selling the outcome, not the platform. Target account engagement. Meetings booked. Events filled. The CMS is infrastructure. The outcome is the product. Think about it this way: if I came to you with a basement of squirrels who could solve your demand generation problem — meetings booked, accounts engaged, events filled — and I said I'll handle the squirrels, you just pay one number, you would not ask me how the squirrels work. You'd say yes. Because the outcome is the thing.

That's the direction this market is moving. An agentic CMS doesn't just help you manage content. It helps you pursue outcomes directly: 800 personalized pages for 800 accounts at a summit. Individualized landing pages for a seven-million-person contact database. Content that didn't exist until the moment a specific person requested it, assembled at the edge in real time, on-brand, fully governed, and live.

Gartner's 2026 Strategic Technology Trends identified multiagent systems as one of the defining enterprise priorities, noting that adopting modular, specialized agents gives organizations a practical way to automate complex business processes and create new ways for people and AI to work together. That's not a prediction about what's possible. That's a description of what's already being built.

That's what we built for. That's the bet we're making. The impossible brief isn't so impossible anymore.

Why agentic CMS is a yes

Built for marketers to create and developers to scale. Move faster with an agentic CMS.