There’s a difference between being invested in a tool and being tribal about it. One means you’ve spent years learning something, building a business around it and trusting it with your clients’ websites. The other means defending it regardless of whether it still does the job you need.
That distinction matters more than ever. Elementor is changing, Bricks is expanding, and AI is opening up workflows that would have felt quite unrealistic not long ago. At the same time, more designers are asking whether every website they build actually needs WordPress.
I’ve been exploring those questions with Jimmy and Jeffrey, and the interesting thing is that we don’t all want the same workflow. We agree on quite a lot, but we approach the work differently. That’s healthy. The goal isn’t to crown one platform the winner. It’s to find better ways to build, maintain and enjoy the work we do.
Key Takeaways
- Evaluate platforms by their fit for your projects, not loyalty to a particular ecosystem.
- Give AI clear frameworks, persistent instructions and small tasks to improve control and manage usage.
- Visual builders preserve direct editing options even when AI handles much of the implementation.
- Separate content management from design only when it simplifies publishing and ongoing maintenance.
Table of Contents
- Leaving Elementor Isn’t Always “Just Changing Tools”
- Bricks 2.5 Beta: Improving the Work Around Building Websites
- AI Works Better When You Give It a Structure to Follow
- Gutenberg, Native WordPress and the Question of Visible Control
- Do We Still Need Visual Builders When AI Can Write the Code?
- Direct AI-to-Code Can Work Too, With the Right Preparation
- How to Stop AI Usage Limits Taking Over Your Day
- When an AI Subscription Becomes a Business Expense
- AI Makes Rebuilding and Migrating Websites Less Daunting
- Does the Website’s Front End Need to Live With Its Content?
- The Most Exciting Part: Building Things That Solve Your Own Problems
- Stay Curious Without Turning Every Tool Into a Team Sport
Leaving Elementor Isn’t Always “Just Changing Tools”
Jeffrey’s decision to stop using and teaching Elementor wasn’t something he arrived at overnight. He had already recorded two attempts at explaining it without publishing either. The version he finally released followed another month of thinking, outlining and trying to find the right words.
From the outside, it’s easy to say, “It’s just a tool. Use something else.” But that misses the investment behind it.
When you’ve spent years mastering a builder, you’re not simply learning where its buttons are. You’re building:
- A workflow you can repeat reliably.
- A stack of compatible tools and extensions.
- Client websites you’re responsible for supporting.
- Training, documentation and possibly an entire education business.
- Confidence that the company behind the product is moving in a direction you can support.
If that confidence disappears, moving on can be difficult. That doesn’t automatically make the reaction irrational or tribal.
Jeffrey’s frustration centred on Elementor’s decisions around version 4. In his view, the way it was introduced left users, educators and add-on developers in an awkward position. People wanted to move forward and learn the new system, but he didn’t feel comfortable teaching or recommending it in its current direction.
That’s his assessment, not proof that every Elementor user should leave. But it is a legitimate professional concern, and dismissing it as “hate” doesn’t help anyone.
Constructive criticism and product loyalty should be able to coexist. If we expect creators to listen to reasonable criticism of their work, it’s fair to expect software companies to do the same.
Elementor’s strengths were real, and so were its limitations
It’s worth remembering why so many of us used Elementor in the first place.
I came from Visual Composer and WPBakery. Elementor made building websites feel different. Its early development, followed by dynamic data and templating in version 2, opened up possibilities that other builders either didn’t offer or didn’t make nearly as approachable.
Jimmy described that period as Elementor’s golden age, which feels about right. He started with Divi, moved to Elementor and used it for several years. His original reason for switching to Bricks was more practical than emotional: he wanted to move away from an ongoing builder subscription and bought an early lifetime licence.
Once he became comfortable with class-based styling and the control available in Bricks, returning to Elementor felt restrictive.
That reflects a broader progression. Earlier builders made visual website creation accessible. Later tools pushed towards reusable classes, design systems and more flexible component structures. An accordion, for example, becomes more adaptable when you can work with its underlying elements rather than only the settings exposed by a fixed widget.
The difficult part for Elementor is serving different audiences. Some people want a visual design tool without thinking about the underlying code. Others want a more developer-oriented environment.
My feeling is that a more clearly separated professional or developer product might have made that transition easier to understand. Trying to keep one foot in version 3 and another in version 4 risks making the direction less clear for both groups.
Bricks 2.5 Beta: Improving the Work Around Building Websites
What interests me about the Bricks 2.5 beta isn’t simply another collection of design controls. It’s the attention being paid to the work surrounding the build.
Once a builder gives you the foundations to create complex websites, the next question is where friction still exists. Client feedback, alternative designs, media management and naming conventions are all obvious candidates.
Drafts and alternative page versions
The draft functionality stood out to Jeffrey straight away. Being able to prepare different versions of a page, let a client review them and then publish the chosen version without disrupting the live page is genuinely useful.
It has a little of the spirit of version control, but in a more approachable visual workflow. You can explore alternatives without treating every design experiment as a change to the public website.
That’s the sort of improvement that might sound less exciting than a flashy new element but can make client work considerably easier.
Native BEM naming and fewer reasons to leave the editor
Another useful addition discussed in the beta is native support for BEM naming. If you prefer to start building and organise your classes afterwards, bringing that process into the editor removes the need for a separate tool.
For anyone unfamiliar with it, BEM is a naming convention built around blocks, elements and modifiers. The value is consistency, especially as a project grows.
There’s a broader pattern here. Recent Bricks development has brought more tasks into the building environment, including media-library access, image optimisation and closer integration with framework values such as colours.
There’s still room to improve. I’d like image optimisation to extend further into format conversion, for example. But keeping related tasks in one consistent interface makes sense.
Jumping between a builder, the WordPress dashboard, media tools and other settings introduces friction. Reducing those jumps makes the environment easier to understand and more pleasant to use.
Some native additions overlap with features previously supplied by external plugins. That’s a wider discussion, but from a workflow perspective, it’s understandable that a mature builder would absorb more of the everyday tasks its users need.
AI Works Better When You Give It a Structure to Follow
Connecting an AI agent to a builder doesn’t remove the need to understand the builder. If anything, it makes that understanding more important.
Jeffrey described a recent Bricks project where he established the foundation himself, then used Claude Code through an MCP connection to help complete the work. MCP, short for Model Context Protocol, provides a way for AI applications to connect with tools and data.
The important part wasn’t merely the connection. It was what existed before the AI started:
- A starter framework he understood.
- Established variables and reusable classes.
- A clear foundation for the design.
- Instructions about how the builder should be used.
He approached it much as he would bringing another person into the project. Set the direction, explain the system and then let them help within it.
That’s very different from connecting an agent to an unfamiliar tool and asking it to build an entire website.
Why a giant HTML block can defeat the point
One of Jeffrey’s rules for Claude is not to take a design and simply drop it into an HTML block.
That approach can produce something that looks right while bypassing the structure you chose the builder to provide. You may have a visual editor, but the actual page is no longer meaningfully editable through its normal controls.
If the goal is a maintainable builder-based website, the AI should work with its classes, variables and elements rather than route around them.
This isn’t a universal ban on HTML blocks. It’s about matching the output to the workflow you intend to maintain.
Tell the AI what to use, what not to use and where its work belongs. Without those boundaries, it can produce something impressive that becomes awkward later.
Gutenberg, Native WordPress and the Question of Visible Control
I still don’t particularly enjoy building inside native WordPress. The movement between the dashboard, block editor and site-editing interfaces can feel disjointed, and some straightforward tasks involve more interface hunting than I’d like.
Jeffrey and Jimmy have similar frustrations, although that doesn’t mean Gutenberg is the wrong choice for everyone.
There is a sensible argument for using core WordPress with AI: fewer additional layers. A page builder adds an abstraction on top of WordPress, and connecting AI to that builder introduces another part of the process.
In my own experiments, more direct workflows have sometimes required surprisingly little time and token usage. But that’s an observation about particular projects, not a guarantee that one stack will always be cheaper.
The trade-off is control. With a tool such as Bricks, I can see and manipulate the structure in a way I find more comfortable. Native WordPress can be extended through code, including AI-generated code, but I don’t necessarily get the same visible access to what’s happening underneath.
If you’re efficient and comfortable in Gutenberg, use it. If Bricks gives you a clearer, faster working environment, that matters too. There’s no prize for choosing the workflow someone else considers more pure.
Do We Still Need Visual Builders When AI Can Write the Code?
For me, the strongest argument for a visual builder is not that AI can’t create a website. It’s that I don’t want every future change to depend on asking AI to make it.
A visual layer gives me a bridge between the generated code and the thing I’m trying to change.
Six months after launch, I might want to adjust a heading, update a colour variable or change some spacing. If I can make that edit directly in a few seconds, why turn it into a conversation?
We’ve all had the experience of asking AI for one small adjustment and discovering it has enthusiastically changed something else as well. Sometimes its initiative is helpful. Sometimes it creates work.
A hybrid workflow gives you options. Use AI when it’s faster. Make the edit yourself when that’s simpler.
Where a visual environment earns its place
I’ve used direct AI-to-code workflows for simple landing pages. For a small, self-contained page, that can be perfectly reasonable.
For a site with content collections and ongoing publishing, I prefer having a more structured editing environment. When working on the Essential Web Designer’s Documents Pack site, for example, I wanted access to the content collections as well as the design.
I wanted to add or edit content manually, with AI, or with a mixture of both. I also wanted to make a visual refinement and push the change to GitHub without needing an agent for every step.
Jeffrey comes at it from a similar perspective: he’s a web designer and enjoys working visually. His ideal is a standalone, AI-friendly platform with approachable design controls and the practical features client websites need, such as forms and email delivery.
He sees tools moving in that direction, but that’s an expectation about where they may go, not a claim that every part is already solved.
Direct AI-to-Code Can Work Too, With the Right Preparation
Jimmy’s recent experience offers a useful counterpoint. He has been experimenting with turning a WordPress website into a static site using AI directly, without a visual builder in the middle.
What made that feel more controlled than his earlier experiments was the preparation.
He used two project files:
AGENTS.md: Project instructions and working guidance for the coding agent.design.md: The design rules and decisions the project should follow.
He took the CSS framework from his existing WordPress setup, supplied it to the AI and used it as the basis for the design documentation. He then asked the AI to reproduce the existing live page.
He was impressed by how closely it matched. More importantly, he felt able to request changes without the whole thing unravelling.
His habit of asking the agent to update the project instructions as decisions change is worth noting. Those files aren’t magic, and they don’t guarantee perfect behaviour. They provide persistent guidance so the workflow relies less on whatever happens to be in the current conversation.
Jimmy started his professional web design work visually, rather than as a hand-coding developer. Yet he now finds that being able to describe an outcome clearly can sometimes replace the need to manipulate it visually.
That rather neatly challenges the assumption that your starting point determines how you’ll always want to work. It doesn’t.
How to Stop AI Usage Limits Taking Over Your Day
Our experiences with AI subscriptions differ quite a bit. Jimmy has been experimenting within a roughly €20 monthly budget. Jeffrey uses separate subscriptions for coding and research, with a combined cost of around $120 a month at the time of our discussion.
Neither figure tells you what your own usage will look like. The model, task, reasoning effort, conversation length and connected tools all affect the experience.
There were, however, several practical habits we agreed on.
1. Match the model to the task
The latest, most capable model isn’t automatically necessary for every job.
In our experience, a stronger model can be valuable when developing a visual direction. Once that direction is established, a less demanding model may handle routine implementation and iteration perfectly well.
Learn where your preferred models are useful rather than assuming the newest one should do everything.
2. Keep reasoning effort proportionate
Jeffrey found that high-effort settings could consume his allowance while exploring far more than the task required.
One experiment involved an add-on for Fluent Community’s leaderboard. The agent went into an extensive problem-solving exercise, using a substantial amount of his available usage on work he hadn’t intended.
His description was rather good: it felt like hiring 20 developers to solve problems that weren’t there.
Higher effort has its place. It simply shouldn’t be the automatic setting for a small, well-defined change.
3. Break work into small, explicit tasks
Large, open-ended requests give an agent plenty of room to wander. Smaller tasks make it easier to inspect the result, catch mistakes and control scope.
Be specific about the relevant files, components and expected outcome. An agent working through a WordPress project without clear direction can spend time exploring parts of the system that have nothing to do with your request.
4. Keep conversations focused and preserve decisions
Long conversations can carry a lot of accumulated context. A tiny styling change doesn’t always need the entire history of a project attached to it.
Keep chats focused on a task, then have the agent record the important decisions in the project documentation before moving on. Use project organisation where it’s available, rather than relying on one endlessly growing conversation.
The aim is continuity without unnecessary baggage.
When an AI Subscription Becomes a Business Expense
There’s a difference between paying for experimentation and paying for a tool that helps deliver profitable client work.
Jimmy is deliberately keeping his spending low while he learns. He isn’t yet offering those experimental AI workflows to clients, because he wants to understand what he’s building first.
Once the tools contribute to paid work, the calculation changes.
We’re already used to paying for builders, plugins and services. An AI subscription belongs in the same business discussion: what does it cost, what does it replace, and how much useful work does it help you complete?
A more expensive plan may make sense if it reliably saves delivery time. It may not make sense if you’re paying mainly to avoid thinking about scope or model choice.
I also don’t want efficiency to remove everything enjoyable from the process. Part of the appeal of a hybrid workflow is letting AI handle repetitive work while I spend time on the design decisions I actually like making.
AI Makes Rebuilding and Migrating Websites Less Daunting
Moving from Elementor to Bricks used to mean accepting a substantial manual rebuild. The same applied when replacing an abandoned theme or removing an unreliable collection of add-ons.
AI can now help with recreating layouts and transferring a design into a different environment. You still need to set the direction and check the result, but it can reduce some of the repetitive work that made migration difficult to justify.
That matters when a dependency becomes a problem.
I’ve had situations where a theme disappeared or stopped supporting the environment a site needed. If the website was relatively recent, going straight back to the client with a large rebuild bill wasn’t necessarily reasonable. Sometimes I absorbed work behind the scenes because it reduced future support problems.
Jeffrey’s position on a full rebuild is generally to have the client conversation and charge appropriately, often using it as an opportunity for a redesign. But he also recalled removing troublesome Elementor add-ons and rebuilding those sections with core functionality at his own expense.
Those are different situations, and they need different commercial decisions.
The useful change is that AI may lower the effort involved. It doesn’t remove your responsibility to inspect the rebuilt site or decide who should pay for the work.
Does the Website’s Front End Need to Live With Its Content?
Another option worth exploring is separating the website’s design from the place where content gets managed.
In a typical client WordPress setup, we might already restrict access to the design tools. The client needs to publish an article, not rebuild the header.
A separate front end and content system takes that distinction further:
- The design and public-facing pages live in one environment.
- The client manages content through a separate dashboard.
- The content is used to populate the public website.
I’ve been exploring an Astro-based front end alongside a content-management option such as mDash. WordPress can also remain the content system if that suits the project.
This is an option, not a declaration that everyone should leave WordPress.
Maintenance and security are part of the decision
A statically generated front end can reduce the public-facing runtime components involved in serving a site. For appropriate projects, that can mean fewer of the PHP, database and plugin maintenance concerns associated with a conventional WordPress installation.
That does not make the entire system maintenance-free or automatically secure. The CMS, hosting, dependencies and any connected services still matter.
But reducing routine update and monitoring work is a legitimate business consideration, particularly when you look after many websites. Being able to charge for maintenance doesn’t mean you necessarily want more of it.
The Astro documentation on rendering modes provides useful context for understanding how pre-rendered pages differ from pages generated on demand.
Find out whether clients really want to manage content
Jimmy raised another practical point: clients often request content access and then never use it.
Even when training is available, some still ask him to publish their articles. That’s fine, but it changes the requirement. If the designer will handle every update, maintaining a separate client-friendly publishing environment may not provide enough value to justify the extra moving parts.
On the other hand, a client who genuinely publishes regularly may benefit enormously from a clean, focused dashboard without access to the design system.
The questions are straightforward:
- Who will actually add and edit content?
- How often will they do it?
- What functionality does the site genuinely require?
- Which parts of the existing WordPress stack remain valuable?
- Will separating the systems simplify the work or complicate it?
Jimmy still values WordPress products such as Fluent Community. Jeffrey remains open to platforms outside WordPress. Neither position requires rejecting the entire ecosystem.
Choose the architecture for the project, not to prove loyalty to a platform.
The Most Exciting Part: Building Things That Solve Your Own Problems
For all the discussion about builders and platforms, the biggest source of enthusiasm was something simpler: making useful things we couldn’t previously have built ourselves.
Jeffrey has been experimenting with products, desktop apps and other small tools. The satisfying shift has been moving from projects that look interesting but sit unused to things he actually uses every day.
Jimmy offered a particularly good example. Finding relevant subjects for his community had become difficult because so much was happening across WordPress, AI and web development.
So he built himself a web app that:
- Collects updates from roughly 500 to 600 website sources.
- Checks those sources hourly for new articles.
- Checks selected YouTube sources five times daily to help stay within the free API allowance.
- Presents new items as cards on a Kanban-style board.
- Lets him record whether he agrees, disagrees or has mixed feelings about an item.
- Generates draft posts for different platforms using that input.
It doesn’t automatically publish those posts. He copies them into the relevant platforms himself.
That detail matters. The tool supports his work and incorporates his opinion rather than simply automating a stream of generic content.
He’s treating it as a learning project, not rushing to sell it before he understands it. The excitement comes from solving a real problem and developing skills along the way.
I recognise that feeling. AI can help take an idea that sits outside your current abilities and turn it into something you can explore, test and improve. It doesn’t remove the learning curve. It gives you a way into it.
Stay Curious Without Turning Every Tool Into a Team Sport
You can appreciate what Elementor made possible and decide not to use it anymore. You can enjoy Bricks without insisting that every website belongs in Bricks. You can build with WordPress while exploring static sites, separate content systems and direct AI-to-code workflows.
Those aren’t contradictions. They’re sensible responses to different needs.
The important things remain familiar: understand the tools, build securely, create something maintainable and deliver what the client actually needs.
Beyond that, leave room to experiment. There’s genuine enjoyment in learning a different way to work, especially when it helps you solve a problem that has been annoying you for years.
Use the tools where they make sense. Question them when they don’t. And don’t let defending yesterday’s workflow stop you discovering tomorrow’s.


