Your CMS can't do that. And that's fine.

I almost missed one of these once.

The agency was OXIK Studio. Spain, mid-2000s. Flash was our secret sauce. We built sites that felt like experiences. Intro animations, immersive multimedia, sound design, motion that followed the cursor. The craft was real. The work was good. We were good at it.

Then the briefs started shifting.

Clients stopped asking for the intro animation. They asked about Google. Why doesn't the site show up on search. They asked about mobile, when mobile barely worked. They wanted content the way you'd want furniture in a house: lots of it, organised, easy to find. They didn't want the experience anymore. They wanted the answer.

I resisted. Not consciously. Emotionally. Flash was what I loved. Flash was what made the work feel like real craft. Static HTML pages with text felt like a downgrade. It felt like I was being asked to give up the secret sauce that made OXIK different from the agency down the street.

Here is the part I want you to remember.

The customers were never as in love with the secret sauce as I was.

They had already moved on, quietly, before any of them had a word for it. They were shopping for content, clear UX, and SEO. They didn't need me to teach them why. They were ahead of me on this one, and I was the one writing the briefs.

By the time I noticed, the graduation had already happened in the market. We caught up. We had to. The agencies that didn't caught up later, or didn't catch up at all.

What stuck with me is the shape of the mistake.

The customer is rarely as attached to your craft as you are. By the time the agency is ready to graduate, the client has been shopping for the new thing for a year.

I see Webflow agencies in that same position right now. The craft is real. The agency is good at it. The clients are quietly asking for things the toolkit cannot do. The agency hears the question. The agency tells itself the client doesn't know what they need yet. The client buys it from someone else.

What is happening

The no-code generation of website tools was right for its era. The era is ending. Most no-code tools are excellent at what they were built for. Agencies just grew past them.

There is nothing wrong with Webflow today that wasn't true of Flash, or WordPress, or Frontpage in their day. They were the right tool for the work the market was buying. The market is now buying different work.

This is not a piece about Webflow being bad. It is a piece about the work changing. The way it always changes.

What no-code was always for

Let me give the no-code generation full credit before I take anything away from it.

Webflow, Framer, Squarespace, Wix, the modern WordPress: these tools did real work. They moved the web forward. A whole generation of designers picked up production responsibility for output they could never have shipped before. A whole generation of agencies expanded their service line without hiring a developer. The output got faster, cleaner, more consistent, more owned.

This was not a small thing. It was a real shift, and the no-code generation deserves the credit.

What it was for:

• Marketing sites and brochure pages. Hero, features, pricing, contact. Light CMS. Done.
• Landing pages with form capture. Fast to spin up. Easy to A/B.
• Light e-commerce that didn't need a checkout pipeline more complex than Stripe and a thank-you page.
• Designers shipping production-grade visual output without a developer in the building.

If that is the work the agency is still doing, the no-code tools are still fine. They were built for it. They are good at it. There is no reason to graduate.

The rest of this piece is not for that agency. It is for the agency whose briefs have started shifting.

What no-code was never for

Here is what no-code was never built to deliver, named plainly.

• Multi-tenant data with per-tenant permissions.
• Real authentication with roles, audit, and per-user data isolation.
• Custom business logic past what a visual rule builder can express.
• Real integrations past the Zapier surface.
• Real-time features.
• Anything past the shape of a CMS or a form.

Somewhere in year three or four of running on a no-code stack, a brief comes in that wants one of these things. It is usually disguised as something smaller. “Can the dashboard show only the data for this client?” “Can the user log in and see their own version of the page?” “Can the form trigger a workflow that updates two systems?”

These are not sites. These are real software.

The agency has three options when this brief lands. None of them are good.

None of these are bad people doing bad work. They are the only moves the toolkit permits.

The structural reason the bridge never came from inside the no-code generation is the kind of thing you can only see in retrospect. A no-code tool that can build real software is no longer a no-code tool. It is an IDE. Every platform that tried to extend in that direction ended up either pushing the agency back toward hiring a developer, or trapping the codebase in a proprietary runtime that fights you every time you want to move it.

The bridge had to come from a different direction.

The direction it came from

The direction the bridge came from is AI-assisted engineering.

Same speed feeling as no-code. Real software underneath.

Real React. Real TypeScript. Real Tailwind. Real Supabase or Postgres or whatever the team picks. The same choices a senior team would make on day one. Real auth. Real data models. Real business logic. Real integrations. The agency owns the codebase. Lift it and run it anywhere. No proprietary runtime trap.

The drag-and-drop work becomes a sentence. The repetitive UI labour goes away. The agency that used to lose two days rebuilding a hero section now describes the change and watches the diff appear. The agency that used to refer a client app out to a contractor takes the brief, builds the first working version on a Tuesday, and ships it the same week.

“But we'd need a developer” becomes “and we shipped it last week.”

I need to be careful here. Not every AI builder is doing this. Most are not. We drew the line between AI-assisted engineering and vibe coding in the previous note, and the summary is short: same model, different posture. The platforms that bake the posture in produce different output than the platforms that don't, even when the underlying model is identical. The platforms that don't are about to teach a generation of agencies an expensive lesson.

This piece assumes the posture from the previous one. The agencies graduating to AI-assisted engineering are not graduating to “any AI builder.” They are graduating to the practice. The tool is downstream of the practice.

The same move, again

I owe you the rest of the list I learned the Flash lesson on.

I have lived through six of these since OXIK. Frontpage to Dreamweaver, when the agencies that didn't move kept charging clients to maintain HTML by hand. Visual Basic desktop apps to the web, when the application layer left the local network and never came back. Custom server-side code to frameworks, when “we wrote our own MVC” stopped being a credential and started being a liability. jQuery to React, when the way the front end was written changed under the same people who had written it for a decade. Serverless to Kubernetes, when the abstraction tax that had felt like a feature started looking like a ceiling.

They all have the same shape.

The customer moves first. The operator catches up six to twelve months later. The operators who never catch up lose the practice.

The agencies that moved from WordPress to Webflow once already know this feeling. The move wasn't “Webflow is better than WordPress.” It was “the work has changed, and the toolkit has to change with it.” This new graduation is structurally identical. Different layer of the stack. Same instinct.

The eighteen-month picture is this. The agencies that make the move become full-service software partners. The agencies that wait become the brand-only shop their clients outgrow. The clients are already shopping for the new thing. That is the OXIK story, told one more time.

It is the same move. On a different layer of the stack. And that's fine.

The pick

The agencies that crossed from Frontpage to Dreamweaver are still around. The agencies that crossed from Flash to JavaScript are still around. The agencies that crossed from WordPress to Webflow are still around.

The agencies that didn't cross any of those bridges are not.

The no-code generation built a bridge for a generation of agencies that did not have one before. That bridge is still there. It is just no longer the bridge to where the work is going.

Webflow taught a generation of agencies that you don't need a developer to build the web. AI-assisted engineering teaches the next generation that you don't need a developer to build software.

Same move. Different layer.

Automo is the platform we built for the agencies making the next move. The team that built it sits with you while you set it up. We've made the mistakes you'd be about to make. Think of us as a partner in the practice, not just the tool.