Optimizely

I work on AI Innovation here at Optimizely. For the last 6.5 years, we've always been at the forefront of using our own products as "customer zero" — that's what we call ourselves — implementing Experimentation, CMS, and AI into our daily lives. I like to say I have the best job as a marketer in the world, because I get to use all the fun stuff without having to do any procurement. Occasionally it's frustrating. Often, it's hard. But it's always been rewarding. Many of the challenges we face internally are the same ones our customers grapple with. I just don't pay for anything. I'd love to share some of those learnings with you. The work I stopped doing was invisible until I stopped doing it The most surprising thing about last year isn't what we started doing — though we did a lot more. It's what we stopped. When we were challenged internally with "AI adoption" following a productivity mandate, we scratched our heads. Do we get everyone chatting more with AI? Build a

The AI gets blamed for the part that wasn't its fault. When agentic CMS workflows fail in production, the failure looks like an AI problem. It almost never is. It's a content model problem. A governance problem. A taxonomy problem. An escalation problem. An ownership problem. The agent is the thing that exposes the gap, not the thing that created it. The teams that succeed with agentic content workflows aren't the ones with better models. They're the ones who fixed the operating model before they connected the agent. I've worked on content systems from every chair in the room — agency builds, brand-side platform governance, and now leading knowledge and education programs at Optimizely. I've also spent three years teaching software lifecycle and AI to postgrad students. I've watched all five of these failures happen. I've caused a few of them myself. Here are my thoughts on the five failure modes and some conversations with my past self about what I should’ve done better. H

I'll be honest with you: walking into a room where most of the participants are skeptical about AI, and you're the person who's supposed to get them excited about it, is a little nerve-wracking. But that's exactly where we found ourselves a few weeks ago in Atlanta, running a hands-on Opal workshop, in partnership with Optimizely. And what happened over the course of that day reminded me why I love this work. Here's one big thing I learned: Most people aren't resistant to AI. They're just uneasy about it. Nobody learned to swim by watching a presentation The majority of participants who joined us were new to AI tools and, understandably, a bit cautious. There's a lot of noise out there about what AI can and can't do, what's safe and what isn't, and whether the output is actually trustworthy. That uncertainty doesn't go away by watching a presentation, it goes away by doing. So that's what we did. Instead of leading with slides and theory, we built the day around actual experimentation.

The assumption I couldn't let go of I have to admit, I have been reluctant, and a bit stubborn, about using AI in my day to day life. Apparently, I am not alone in that. While Gen Z is often called digitally native and quick to adapt (I mean, my whole life is documented through Facebook): Gallup found that 38% of Gen Z'ers feel that AI would "be harmful rather than helpful to their abilities to come up with new ideas on their own." Like many others, I have the same question: Why do I need a robot to help me write content when I went to college to learn how to be a better writer? Graduating from college in a pandemic gives me a particular perspective on the AI boom — we had Quizlet and Chegg instead of ChatGPT and Claude. I take pride in coming up with my own ideas. Whether that means being able to write a 30-page research paper on my own (not to brag, but I do have a master's degree) or helping my best friend come up with an Instagram caption for her vacation photo dump — I

When AI first arrived in marketing, the reactions ranged from "this changes everything" to "what exactly is this changing, and should I be worried?" Either way, most teams did what they do best: they jumped in, figured it out, and learned fast. That was then. AI for marketing has matured significantly since the early days of prompt-happy copy generation and one-off experiments. The tools are smarter. The expectations are higher. And the conversation has shifted decisively from "Should we use AI?" to "Are we using it in the most strategic, scalable way possible?" More than that — we're now firmly in the agent era. It's not just about using AI to speed up individual tasks. It's about building AI agents that orchestrate entire workflows: planning, executing, analyzing, optimizing... often without a human in the loop for every step. Which raises a genuinely important question: where does your marketing team actually sit on the AI maturity curve? Use this self-assessment to find out.

"Why would anyone want a dumber response?" That was the pushback I got when I argued that sometimes the right answer is to think less. I work on AI products that marketing teams use at scale every day, which means I feel latency and token cost problems before most teams even notice them. We had spent months building an enrichment pipeline we were genuinely proud of, and for good reason. When Opal pulled in brand guidelines via RAG, searched the knowledge base, loaded specialized tools, and consulted episodic memory from prior conversations, the responses were measurably better. So, why would anyone skip that? The thing is, we were already making smart decisions about enrichment. We weren't blindly fanning out across every context source on every request. The system checked whether relevant knowledge bases were available, whether the conversation state warranted certain lookups, and what model the user had selected. The enrichment pipeline had conditionals. It had early exits. It was a

When it comes to personalization , we're in competitive times. We're in hard-to-please times. We're in attention-span-of-a-goldfish times. Everyone knows personalization matters; that's not news. But how do you deliver truly relevant experiences that drive conversions without wasting resources? That's where contextual bandits come in. By selecting relevant user attributes, contextual bandits automatically learn which experiences work best for different audiences, providing valuable insights while maximizing conversions throughout its runtime. What is a contextual bandit? The term " multi-armed bandit " comes from the classic slot machine analogy (the "one-armed bandit"). Imagine a casino with multiple slot machines. Which one do you play to maximize your winnings? That's the basic challenge. Contextual bandits take this to the next level by factoring in who's pulling the lever. They leverage user data to make better algorithmic decisions and deliver 1:1 personalization . The machine le

There's a version of the AI conversation in financial services that goes like this: leadership sends a memo, a tool gets approved, someone books a training, and six months later nothing has really changed. The productivity gains haven't materialized. The enthusiasm has faded. And the team is quietly using ChatGPT for emails while pretending they're not. We've heard this story enough times that we decided to do something about it. Earlier this year, we ran the first cohort of Opal University for Financial Services,  a five-day live program built specifically for marketing and digital leaders working inside one of the most compliance-conscious industries in the world. Fifty seats. One hour of live training followed by one hour of hands-on agent building, every day, for a week. A global cohort — participants joined from the US, the UK, and Europe — all working through the same challenges in the same room. The goal wasn't to teach people what AI is. It was to get them buil

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. Th

Like most content marketers using AI, I've been training my tools to write in my tone of voice. Great news : it worked.  Bad news : apparently I say "actually" approximately 400 times per doc. My AI has absorbed this completely. It now opens sentences with it more confidently than I do, which I didn't think was possible, and yet... here we are. Also, food metaphors. Everywhere. Do I constantly write copy when I'm hungry? Because it bloody well seems like it. The em dash conversation (/ heated debate ) Okay, so here's the one that really got me. I love em dashes. I have always loved em dashes. They've been in my content guidelines at more than one job, and part of that is professional convention, but in all seriousness... part of it is that I think punctuation is beautiful and I like how it makes a sentence look. In my totally humble (and perhaps a lil bit content-nerdy) opinion, the em dash is a very useful and very pretty piece of punctuation. Alas , then the great em dash witch