How I Built a Complex AI Workflow Automation Plugin for WordPress With Zero Coding Experience

Let me introduce myself—this will help the story of our WordPress AI plugin make more sense. 

My name is Navid, and I’m a curious generalist who loves building things. Whether it’s furniture, robots, or businesses, I thrive on creating. Over the years, I’ve built, scaled, and sold businesses in industries like food, 3D printing, and manufacturing (nothing spectacular, but still meaningful).

Currently, I work as a Global Open Innovation Manager at a multinational energy company. My role is to connect startups to corporations in a way that helps them create real value for each other. This means that I get to work with very cool startups with amazing technologies. In recent years, nearly all of them have a connection to AI. 

However, there’s a big challenge: startups are often unprepared to collaborate with large corporations. They lack the infrastructure needed to meet corporate requirements, which makes partnering difficult.

As a result, corporations miss out on the opportunity to quickly and easily experiment with and explore all these cool and exciting technologies. Instead, the process gets bogged down in months of legal, technical, and compliance work before we can even get started, and by that time, everyone has moved on.

The Spark: Addressing a Real Problem

These days, everyone seems interested in AI-powered customer service agents, AI-assisted employees, or AI tools that can handle research and reporting. There are plenty of solutions for building these tools, but as I mentioned earlier, they’re not designed for companies with extensive frameworks and requirements.

So, with my innovation manager hat on, I saw this as a real problem worth solving—at least for myself. I started by exploring all the available platforms that let you create AI-powered automations and applications with flexibility, like Make or N8N. The problem? Most of them aren’t self-hostable. To use them, I’d need to go through the whole lengthy legal, technical, and compliance processes of corporate environments.

On the other hand, the self-hostable options require a level of technical knowledge I simply don’t have—things like setting up databases, dockers (Docker containers), and configuring other infrastructure.

Then it hit me: I already knew how to work with WordPress. I’d built WordPress sites before and experienced firsthand how powerful plugins can be. With the right plugins, WordPress can become more than just a blogging platform—it can power complex web applications. That’s when I thought: Wouldn’t it be great if there was a plugin that brought all the features of an AI workflow automation platform into the WordPress dashboard?

Such a plugin would eliminate the need to set up complicated infrastructure. All I’d need is a WordPress instance, and the automation system could run on my server, with my data. I could leverage WordPress’s built-in features for user management, application building, and testing—no coding required.

But here’s the thing: that WordPress AI plugin didn’t exist. Sure, there are automation plugins for WordPress, but they serve a different purpose. They focus on “if-this-then-that” processes to connect plugins to each other. 

There are also new AI plugins, but they’re limited to offering chatbots, helping with writing, or a mix of both. What I needed was something far more flexible—an AI workflow automation plugin that allowed me to build anything and everything, both front-end and back-end, with any AI model, all through a visual interface, as I didn’t know how to code. 

From Zero to Code: Developing as a Non-Developer

For some reason, I decided this was a cool project to try and see if I could build it myself. With no clue where or how to start, I came up with a prompt and sent it to the three major AI models at the time (Summer 2024, all paid versions): OpenAI’s GPT-4o, Google’s Gemini, and Anthropic’s Claude 3.5 Sonnet. My goal was to see which one could best help me.

This is the prompt I used:

“You are going to help me build a WordPress plugin that enables the user to build AI powered workflows for their wordpress website. These can be anything to automate content generation, or additional features for their users. This system should be able to use different AI models such as OpenAI or Claude. I also want this plugin to be like a visual workflow builder as much as possible. As the first step, I want you to outline what are the key features of this tool that we need to have?”

Google’s Gemini was terrible from the start. GPT-4o came back with some ideas and even generated PHP code that loaded as a WordPress plugin—but it was clunky, non-functional, and completely uninspiring. 

Then came Claude. It created a detailed plan for the project, predicted all the necessary features, and outlined a suggested tech stack for both the front and backend. I was impressed. I got excited and told it to “go ahead and start developing it!” to which it replied: 

“I appreciate your enthusiasm, but I want to clarify that creating a fully-featured WordPress plugin with all the capabilities I mentioned is a substantial project that would require hundreds, if not thousands, of lines of code across multiple files. It’s not feasible to write all of that in a single response.”

Claude was right. Over the next three months, we worked together to develop the first version of the AI Workflow Automation plugin. It took tens of thousands of lines of code to create a beautiful, fully functional AI automation plugin for WordPress, with a visual React-based front end and a well-structured, reliable backend.

To this day, I still can’t sit down and write code. But during this process, I became an excellent product manager. I’ve learned to read and understand code quickly—whether it’s React or PHP—and can usually spot errors and bugs at a glance. I can also clearly explain features and bugs to my developer, Claude.

I believe this is the future of software development: AI handling code generation while humans focus on architecture, requirements, and quality control.

But getting there wasn’t easy. The toughest part of the process was going from an idea to a functional version 0.1 with basic features. My first working feature was deceptively simple: creating a basic node that could call OpenAI’s API.

What looked straightforward quickly turned into a crash course in React state management, API authentication, and WordPress security. Claude walked me through creating the node component, explaining concepts like React hooks and props—both of which were completely new to me.

That first feature took over a month of back-and-forth to get working. But once it did, it became the foundation for every other visual feature that followed.

From there, as the project became more structured, there was less room for error in the model, and things started moving faster. There’s so much I’ve learned, but here are some of the most important takeaways:

1. When in doubt, ask Claude for ideas and rationale. If you’re unsure what to do, ask Claude to come up with an idea and explain its reasoning. Why this technology? Why this structure? Once the explanation makes sense to you, send it back to Claude as guidelines.
2. Break your project into smaller chunks. This is the secret sauce for managing projects with large codebases. Instead of asking Claude to build a full-featured component—say, a front-end with multiple layers of buttons and features and a backend with complex functions—break it into the smallest possible tasks. Start by asking for the most basic version of the front-end component. Once you’re happy with that, move on to the next bit.
3. Always provide context. Don’t just drop in a piece of code and say, “I need you to improve this for XYZ.” Explain what the code does, and include related classes, methods, functions, debug logs, or any other relevant information that speaks Claude’s language: code. Claude works best when it has all the pieces of the puzzle. Its built-in project features are excellent for keeping context consistent in chats—use them. In fact, while developing AI Workflow Automation, I even built a Claude Chat Manager extension to help organize my conversations.
4. Don’t be afraid to start over. If you feel your conversation with Claude isn’t going anywhere, start a new one. This time, give clearer instructions and more context to avoid repeating the confusion of previous sessions. For example, if I request a feature but don’t know how it should be developed, Claude might try a few approaches that fail and spiral into unworkable code. When that happens, I start fresh, explicitly stating that I don’t want it to use methods A, B, or C.

AI Workflow Automation Plugin: What Makes It Unique

There are so many great automation plugins for WordPress that let you connect and automate actions between your plugins. Similarly, many great AI-powered plugins enable chat assistants, AI-generated pages, and content creation.

For me, AI Workflow Automation isn’t another plugin that does either of those things—it’s a new foundational capability for WordPress. It gives WordPress an entirely new dimension, letting you use WordPress in ways it wasn’t originally designed for. Think of it as similar to what WooCommerce does: transforming a WordPress site into a full-fledged e-commerce platform. AI Workflow Automation turns WordPress into a powerful AI-powered automation hub.

With AI Workflow Automation, you can automate actions, build AI assistants, or create pages and posts with AI—but that’s just scratching the surface. The real potential lies in your imagination. You can build entirely new businesses offering AI-driven services, use the plugin as a backend for AI solutions, or even create autonomous agents to perform tasks for you—like conducting market research and generating reports—without requiring your intervention. (Of course, there are still limitations.)

These days, many people are building AI SaaS products, often using one of the countless AI SaaS boilerplates available. But I see the combination of WordPress’s ecosystem and AI Workflow Automation as a real no-code alternative—a true AI boilerplate for creators, entrepreneurs, and innovators.

Real-World Impact: Success Stories from Users

I haven’t had the chance to focus much on marketing the product yet, but I’ve been fortunate to attract a highly enthusiastic group of early users. Both the Lite and Pro (paid) versions have been well received, and several large agencies and consulting firms are already using the plugin for various purposes—just as I envisioned.

Most users are leveraging it for autonomous content generation. They’ve built complex agentic workflows that conduct in-depth research, draft, write, review, optimize content for SEO, and even publish it. But there’s more to it than just content creation.

For example, one consulting firm is using AI Workflow Automation to quickly create MVPs and demos for complex AI tools and services. They’ve built agents that transform complex input data into properly structured, formatted outputs, which are then used in business simulation apps.

Addressing Some Elephants In The Room: I Hear You, My Developer Friends

I know some of you have some questions, so I will try to answer some of the most commonly asked ones:

How do you know the AI isn’t generating bloatware or malware?

Reading and understanding code is surprisingly straightforward when you grasp the logic behind it. It’s easy to spot irrelevant or malicious code, and I personally review every single line to understand its function before implementing it.

Also, there’s no inherent reason an AI model would autonomously inject malware into your code—it’s not in its design. You could ask the same question about hiring freelancers, and the answer would be the same. In fact, AI is arguably more predictable than a human coder.

You don’t know coding. How do you ensure the code is secure for users?

That’s the beauty of the agentic approach to AI models rather than zero-shot. I use dedicated chats and projects for specific roles: some agents focus solely on writing code, while others review and optimize it or check it against WordPress best practices.

Thanks to this process, I passed the WordPress repository’s publishing checks and regularly re-check and re-optimize the code. For example, I ensure all input fields are sanitized, sensitive data is encrypted and decrypted properly, and security best practices are followed. Is it perfect? Probably not—just like most other plugins. But it’s continuously improving.

How do you test and debug the system?

I’ve created a standardized testing approach tailored to the plugin. That said, like with any software, some bugs only appear with real-world use. I’ve been fortunate to have customers with strong IT teams who’ve stress-tested the plugin and reported bugs or broken features. Their feedback has been instrumental in making the plugin more stable and purposeful.

You might be able to develop something, but how will you handle maintenance and improvements?

So far, I’ve released around 20 versions of the plugin. The codebase has doubled in size since the first version, with twice the features and a much-improved UI. It all comes down to methodology on how you manage your AI agents so you don’t lose track of things.

That said, I anticipate that as the plugin grows, it will become increasingly valuable to bring in expert programmers to review the code and help build more advanced features as we move forward.

Future Vision: WordPress for the AI-Era

WordPress, as one of the most powerful web development platforms today, will inevitably become deeply integrated with AI technologies and tools.

I see AI Workflow Automation as a foundational building block for this future, empowering developers and creators to build complex, locally hosted agents on WordPress that can perform complex operations fully autonomously.

I also envision the core technology behind AI Workflow Automation—and its visual approach to designing agent workflows—becoming a feature that could integrate with other plugins. This would unlock an entirely new dimension of functionality and services for their users, pushing WordPress further into the AI era.

A Final Message to The WordPress Community

I love WordPress, and I love the community it has around it with all its (recent) drama. Just as an example, the WP Mayor team has been such a delight to talk to and share ideas with from the first moment, accepting me as an outsider with open arms to even write for them. 

To me, WordPress is the perfect platform for building digital solutions. I encourage all of you to give it a shot! With all the AI-powered tools and assistance available today, development is no longer off-limits for us non-technical enthusiasts—and it’s only going to get better and easier.

And the more builders and contributors there are, the better. 

Keep on building, keep improving WordPress, and let us know if you’d like to know more details about the development process, making agentic AI workflows, or anything else WordPress AI plugin-related!

[Editor’s Note: Read our unbiased review of the WP AI Workflow Automation plugin.]