AI won't make your content better just by getting it live faster

Raff Di MeoRaff Di Meo
Aug 18, 2026

AI tools are helping teams publish more, faster. But faster noise is still noise. Raff Di Meo on AI that improves a marketer's job — not just their output rate.

Every marketing leader I sit down with says the same thing. They're not worried about publishing too slowly. They're worried about AI publishing more of the wrong thing, faster. The industry has built exactly the tool they're afraid of. Every roadmap has agents in it. Multiple analyst reports throughout 2025 — including Gartner's Hype Cycle for Digital Marketing, 2025 — warned about content saturation, yet the industry remains obsessed with speed. We're building tools to publish faster without asking if we're publishing better. The conversation has shifted from "can AI help us create?" to "can AI just do it?" And the answer, increasingly, is yes.

So the question I keep coming back to is: what does AI need to help us be faster at? And how do we know those are the right tasks to be doing faster?

I lead the Experience Creation design team at Optimizely. We work across CMS, CMP, and DAM. And I've noticed that the hardest part of AI-powered products isn't the AI. It's understanding the people using it. Especially when we're enabling a new entity to decide on someone's behalf, understanding what people are trying to do and how becomes the product itself.

[Marketers] aren't worried about publishing too slowly. They're worried about AI publishing more of the wrong thing, faster.

Raff Di Meo|Senior Manager Product Design, Optimizely

What marketing teams want

If you work in tech, you know what everyone is talking about. The message is loud and clear: go faster. And it's landing on marketing teams every day. More content, more channels, more output. The pressure is real.

But when I actually sit down with those teams, speed isn't the thing they lead with. Their top-of-mind problem is quality. The fear most leaders talk about is AI publishing more of the wrong thing, faster. There's a pressure to move quickly, but it needs to be without losing the brand voice, the accuracy, the judgement that makes content worth reading.

That tension is what I keep coming back to, and what I urge design to think about. Our goal isn't to help marketing teams go faster for the sake of it. It's to help them do their jobs better. And the only way we can be confident we're working on the right problems, and solving them in the right way, is by staying in constant conversation with them.

Axiom Play: testing the future inside the product

One of the things the design team has built that I'm most proud of is something called Axiom Play. It's the fruit of great work from Benjamin Holme, who ran with a simple idea and created a product that can now be used by every other designer within Optimizely.

Axiom is our design system. Axiom Play is what happens when you use that design system to build a fully functional prototype that looks, feels, and behaves exactly like the real Optimizely CMS. Not a wireframe. Not a clickable mock. Something a customer can sit down with and use, without feeling like they're in a research session. It feels and behaves the same as the real CMS, enabling customers to explore what a future version of the product could feel like.

The reason for creating and using something like Axiom Play is to create a space where we can test future features and workflows in a context that feels real. The closer the prototype is to the actual product, the more honest the feedback becomes. People stop performing and start doing their job. When users interact with a clickable mock, they're testing our design. When they use a functional prototype like Axiom Play, they're testing their own ability to get work done. That's where the performance ends and the real work begins. And by observing how people perform specific tasks, seeing where the "Wow" happens and where the "Mmmh" happens, well, that's gold. That's when you find out what's actually working, what's worth investing in more, and what can wait. It's a design team initiative. Built by the team, owned by the team. And it's changed how we think about validation.

What happens when you show people

The work we've done with customers through Axiom Play is starting to show up in the product. Two examples stand out.

1

The first is Optimizely's agent platform. We've been designing a set of agents that can be triggered within the CMS, either manually or configured to run autonomously on specific jobs, such as proofreading, WCAG compliance, and brand standards checks. When we showed customers how this worked, the reactions were immediate. People said things like "if we only had that, our jobs would be so much simpler." That kind of response doesn't come from a feature announcement. It comes from seeing customers discover the feature in a toolbar, click on it, and exclaim "wow". All we have to do is observe and take notes.

2

The second example is more instructive. Search management is a powerful feature, but it becomes super complex when you start diving into it, and it didn't land cleanly the first time we showed it. We're building something that lets marketing teams configure how search behaves on their websites — controlling pinned results, synonym handling, and query weighting. It's powerful. But in early sessions, customers didn't quite get the labelling, and some of the complexity we were proud of sat on the surface of the interface. Watching a seasoned marketer squint at our dashboard and hesitate for ten seconds before clicking a search-weighting slider told us everything we needed to know. Our goal is to make complex configuration feel seamless, regardless of what's happening underneath. So, our team went back and refined it, and it's now launched in a new beta form: much simpler and much closer to what our customers needed. That feedback didn't feel like a setback. It felt like the system was working exactly as it should. That's the difference between building with customers and building for them.

The architectural difference

There's a version of the future where every CMS vendor just sprinkles more AI on top of their product. Better autocomplete. Faster generation. A chatbot in the sidebar. And maybe that's useful. But it doesn't fundamentally improve a marketer's job.

What makes it better is AI that understands the work. Not just the task it's been given, but the context it's operating in. The standards the content must meet before it goes anywhere near a customer.

That's what Optimizely's agentic CMS is being built to do. Not to be a general-purpose AI layer on top of Optimizely, but to think like a marketer and act inside the CMS like one. The agents don't just generate. They participate. They make changes for you. And the only reason we can build that with any confidence is that we're close enough to our customers to know what "participating well" looks like in practice.

You could use a general-purpose AI tool for a lot of this. But most general-purpose tools don't have native access to your brand, your site's content structure, or your data model. Optimizely's agent platform, built on top of a CMS that understands all of that, can work with that context by default rather than by workaround.

That's not a feature difference. That's an architectural difference. And it starts with staying close to the people the product is meant to serve.

Most companies won't do this. Not because they don't care about their customers, but because it's slow. Sitting with someone while they try to use a prototype, watching where they hesitate, listening to the feedback that sends you back to the drawing board.

The teams that skip it will ship faster. For a while. And then they'll wonder why adoption is flat, why the AI features aren't landing, why their marketing customers keep asking for things that feel obvious.

Customer closeness isn't a differentiator you can copy by adding another AI model. It's built over time, in the moments where you choose to go back and fix something instead of shipping and moving on.

We're choosing that. And I think, over the next few years, it's going to matter more than any of us currently expect.

An agentic CMS built for marketers

AI shouldn't sit next to your CMS. It should operate inside it. Now your teams stop waiting and ship more.