Introduction
We had two options: either we see a competitor, an AI native CMS, being created in front of us, or we decide to create our own competitor that could become our best ally. The web was mostly built by humans for humans. I think the web is going to be built by agents, mostly for agents and humans, and that is the big shift. Currently, 90% of the code is being AI-coded. Compared to what we used to do with Strapi, which would have taken two weeks, it is now taking two minutes.
I think this is actually great for the role because I don't like to see PMs spending their time doing tasks that can be done by AI. Your real value is about making the right decisions. The market window is right now; it is not tomorrow, and it is not in two or three years. That was Aurélien Georget, who co-founded Strapi in 2013 as a student project. Today, it powers content for the websites of Apple, Adidas, and IKEA.
With $50 million raised and millions of users worldwide, he launched a brand new company called FIMO just two weeks before we sat down together. FIMO.ai is an AI native website builder built on the observation that while every existing tool can build a website, none were designed to let AI maintain and evolve it over time. In this conversation, you will hear about how ten years of running Strapi gave him the insight to see what everyone else is missing. You will also learn what building FIMO in six months with a team of five taught him about what it actually means to be AI native.
How Strapi Was Born: From Student Project to Global Powerhouse
Thank you for having me. It is a pleasure to be here. We started as a student project in 2013, which is where I met my co-founder while doing my master's degree. At first, we were just three freelancers working on different missions, mostly building websites. At the same time, the mobile app started to rise. Our customers liked to build websites but also wanted mobile apps at the same time, and the content was the same.
We were using WordPress at the time, but it was no longer a good fit for those products and projects. Consequently, we created our own product, Strapi. It had another name initially; I think the first name was Wisti or something like that. I have never shared that before; it is an exclusive news item. We built it for our own customers and published it on GitHub. We hoped to get some traction, but it was nothing serious, just a few hundred stars on GitHub.
At the end of our studies, we decided to create a company and give it a try. The three of us were entrepreneurs and engineers. I started to code at thirteen, so I have mostly an engineering background, but we still have different skills. Pierre, the CEO, was clearly the best at sales and managing the company. I was better at managing projects and building a product. This is how I ended up being the CPO. Now we are a company of 70 people, fully remote. We launched Strapi ten years ago and launched a new product called FIMO two weeks ago.
Why Launch FIMO Now? The Cargo Ship Problem
It was a tough call to launch FIMO now while I am already a CPO at a successful company. Strapi is working well; we are doing millions in revenue and will be profitable in less than six months. We raised $50 million with the best US VCs, and we have millions of users. Most of the big corporations in the world use it. If you go to the websites built by Apple, Adidas, Victoria's Secret, or IKEA, they all use Strapi for their main websites. It is huge.
However, AI is redefining pretty much everything in our industry. We had two options: either we watch an AI native CMS competitor be created in front of us, or we create our own competitor that could become our best ally. It is not easy to change direction when you have a ten-year-old product; it is like a big cargo ship. We have very long-term contracts with companies and millions of users. If we want to make Strapi AI native, it is going to take some time.
The market window is right now. It is not tomorrow or in two or three years. We decided it was faster, better, and certainly a better decision to start a new product from scratch. We can learn from it, apply those learnings directly into Strapi, and avoid failing with our major product. It is better to fail on a small product, apply those learnings to a mature product, and see what happens with FIMO.
The Agentic Web: Who Builds the Web Next
When I pitch FIMO, people get it. I spent the last three months of 2024 in San Francisco to get into the right mindset to build an AI native product. The pitch for FIMO is pretty simple. We think the web is going to be different in the sense that the web was mostly built by humans for humans. Digital agencies and freelancers were coding manually on their tools, and the web was mostly browsed by humans clicking and typing in forms.
In the last two years, we have seen more AI coding tools and agents like Claude, Bolt, and Cursor developing websites during the entire night while we are sleeping. They develop the website and apps for us, and when you wake up the next morning, it is done and working. The web is going to be built by agents, mostly for agents and humans. That is the big shift, which we can also call the agency shift. It means the decision-maker is changing, and most decisions will be made by agents.
Agents are the ones coding, and we think this will apply to websites too. They are going to be the ones maintaining those websites. They will add pages, update content, and analyze drops in analytics to suggest trying something else for the homepage. They will see when a competitor releases a new feature or changes their pricing and suggest a defensive approach. With FIMO, we are helping people build the first generation of autonomous, proactive, and intelligent websites. Right now, we have built the first brick, which is the website builder, but we are working on the agentic part.
When Content Management Breaks at Scale
Content management is like any company; when you start to grow and have only ten people, it is easy to share information in one room. When you are in a big group of 10,000 people, there will always be someone missing or someone who cannot attend a meeting. The bigger you are, the more communication debt starts happening. It is the same for content management. If you have a website with ten pages, it is super easy to maintain.
If you have a giant website like Adidas, they have between 10,000 and 100,000 pages. It is super hard to maintain those pages to ensure you don't have contradictory information or to make sure the new tone of voice applies everywhere. Even if you have an army of content writers and editors, it is not easy. This is where AI will fill the gap. AI is excellent at manipulating text and browsing your website to identify and fix those gaps.
Maintenance is what we often forget when we build a website. It is easy to build because you have a clear idea of what needs to be done, but problems happen when the website has been live for three months and you change your pricing or positioning. If the designer decides to change the images, software and traditional CMS designs don't truly help solve that problem. Traditional tools make it easier to do bulk actions, but AI changes the game because you need to do human-level work at a scale a human cannot manage.
Where is the Human Limit? The Apple Christmas Thought Experiment
I was writing an essay about this. Imagine that in three years, during the Christmas season on Apple.com, an AI agent decides to remove all products except the AirPods. It does this because the AI knows that during that season, those products sell the most and the margins are highest. The agent has one goal: maximize revenue. Is that a bad decision?
When you look at chess or Go players losing against AI, there is often a move—like move 37—that everyone thinks is a huge mistake. Actually, that turns out to be the smartest move ever; we just didn't know it at the time because it was new. If an agent makes a decision based on data points that ends up meeting company goals, I don't know if that is bad. It might be a bit sad to see, but I don't know where the limit is.
Some companies would never let an agent manage their homepage without permission, but for most pages, you might not care. If the tone of voice is great and the AI has the right constraints, we can let agents do most of the work so we can focus on something else. Regarding branding and the general feel of a website, I haven't seen anything super successful yet. There are emotions and identities that AI doesn't get yet. In industries like perfume, where creativity is unique, I don't think AI will replace humans soon. For a SaaS company, where the tone is usually friendly and concise, it can be reproduced well.
Workflows Nobody Actually Uses
I learned a lot at Strapi about how organizations actually ship content. Take Adidas as an example: when you browse their website, you are seeing content produced in a giant Strapi application. They have fifteen people working on it, and they produce that content eighteen months before publishing it. That is a unique publication workflow, but it is an exception. Most people think they need reviews and workflows, but they actually don't.
They often want those features just to reassure themselves or to check a box during the procurement process. In reality, they don't implement them, or they just need simple boundaries to ensure they don't publish to production without a quick review. Most people manage small to medium websites with only one or two people. You don't need complex workflows for that. Even among our enterprise customers who buy those features, only about 15% use them. In the entire user base, it is less than 1%.
The Signal That Started It All
I remember a board meeting at the end of 2022 where AI was a main topic. We discussed if we should go all-in on AI then, but it was before ChatGPT, and we decided the tech wasn't ready. Then came the revolution of 2024 and 2025 with coding agents getting better. The real signal was when customers told us during negotiations that they would just recode a feature themselves instead of buying the enterprise edition.
Coding is one thing, but maintenance is also cheap now. If a customer develops a small CMS feature and the developer leaves, it doesn't matter as much anymore because an AI coding agent can maintain it. Previously, knowledge would leave with the person, but now we don't care as much. If people can recode the features we sell, those features are worth less. We decided to create an AI native CMS that embraces AI at its core from the beginning.
The website building market is giant, worth billions, and still growing. You have categories like IDEs like Cursor, traditional CMS like WordPress, and visual platforms like Webflow or Framer. All these categories are being hit by the vibe coding category. Do you really need a visual editor if you can constrain the AI to use your brand? You might not need a complex CMS or an engineer if a prompting agent can do it for you for $50. It is faster, cheaper, and gives you more control.
FIMO vs. Lovable: Building Apps vs. Building Websites
There is a lot of noise in the AI builder space, but FIMO is different. Tools like Lovable are amazing, but they are made to build apps, which are different from websites. A website is about promoting content and your brand. If you use Lovable, the content is often generated during code generation and stored as static text in the code unless you specifically ask for a database.
Furthermore, those websites are not optimized for SEO or AEO (Answer Engine Optimization). If you store content in a database with Lovable, you just get a raw interface made for developers. You don't have a rich text editor, a media library, or easy formatting. You end up burning a lot of credits just to replace an image. With FIMO, we have three different agents working in the background: one for design, one for code, and one for content.
Our content fits your brand better because we always store it in a structured CMS with a real interface. Because the content is structured content, we can manipulate it. We can compare your content to a competitor's pricing update because the agent has that context. We can translate it, change the tone of voice across all pages at once, and even convert a Slack thread into a blog post automatically. We are the content experts maintaining the prompts to optimize your visibility based on your specific goals, like maximizing signups or demos.
What Does AI Native Really Mean?
Being AI native is about a mindset and a specific stack. At FIMO, we have principles like never starting from a blank page. AI avoids the blank page by generating a first version for you to iterate on. If you want to add a new page, you tell the AI what you want, and it structures your ideas into something tangible. Our process is also different; 90% of our code is AI-coded.
Our engineers and designers don't open tickets in Linear for QA; we all "vibe code" and fix issues as we see them. If a button is misaligned by two pixels, I just code it and ship it. This takes two minutes instead of two weeks of backlogs and discussions. We were able to redo many Strapi features with a team of three engineers and one designer in six months. It is not just about the product; it is about the economics and the lean processes.
The roles of PM, designer, and engineer aren't merging, but their scopes are changing. PMs and designers can now code small fixes, which removes low-value tasks from the engineer's backlog. Engineers then have more time to think about the product and challenge the UI. Many roles are being automated; we have no one managing documentation because it is AI-generated. PMs aren't writing transcripts or user stories anymore, which gives them more time to develop their product sense and talk to users.
The Skills That Matter Now
PMs need to go back to the roots of the role. You need to set a clear vision without controlling every detail of the team. It is essential to know your market and competitors. I believe every PM should have a "suitcase of agents" to automate their job. If you automate the boring tasks, you can spend more time understanding the market and keeping pace with the AI field.
You must ensure engineers understand the audience so they don't ship in the wrong direction. For example, at FIMO, our engineers initially built an amazing code editor tab because they love code. However, our customers are not coders and don't care about the code; they just want a website. We had to remove those features the day after release because they didn't match the customer profile. Your job as a PM is to be the director.
I actually regret not launching FIMO even earlier. Even when the product was unstable in November, we should have shipped it. You always worry about the first impression, especially when you aren't the first in the market, but you can't wait. You just have to try and iterate. In the AI native space, you shouldn't optimize for economics or margins yet. Optimize for usage and traction first, then use different models later to fix the margins.
Advice for Founders and PMs
If you are a Series B company like Strapi, you cannot innovate with 100 people involved. Innovation comes from a compact, lean team. We pulled three engineers, our CTO, a designer, and myself into a private channel to work in isolation for six months. This protected the team from distractions like setting up support workflows or complex documentation before we even had a product.
We used a different billing system and different support software to avoid constraints. If you have an existing non-AI native product, create a side team and let them work outside your standard processes. As a PM, if you feel the market is moving, you need to ask your leadership if you should wait or start from scratch with a specific target. I saw many US companies making this move, though fewer in Europe where we might be more conservative.
I believe the web will be completely different in five years. We might even have browsers that specifically highlight human-created content because the web will be full of agentic content. We won't be able to trust what we see on a screen anymore. This means creative jobs and real-life experiences will become even more important. Perhaps we will even spend less time watching screens, which would be a good thing.
Rapid Fire Questions
What's the best product decision you made in the last six months? To fully delegate my job at Strapi to our product team so I could focus on FIMO.
If you could have one feature in FIMO, what would it be? Full autonomy for websites.
What's one thing you believe about AI that most people would disagree with? It is not going to replace jobs, but it is definitely going to redistribute them. The jobs will still exist, but we will need different skills to do them.
Five years from now, what do you want to be known for? Having created a new generation of autonomous websites.
Closing Thoughts: Aurélien Georget has a grounded take on AI. No hype, just the perspective of someone who has been in the trenches for ten years. If you want to see the future of the web, check out FIMO.ai. I'm Cedric, and this is The French Product. See you in the next one.