AI-Assisted Development Without Losing Engineering Judgment

Modern software development gives us access to increasingly powerful AI tools. But using AI effectively is not the same as letting AI make every technical decision.

For me, AI is most valuable when it accelerates the development process while the developer remains responsible for architecture, requirements, user experience, integrations and the final result.

This approach became particularly relevant in my work at CLYRA, where I worked across frontend development, backend development, UI/UX, automation and external integrations.

AI as Part of the Development Workflow

At CLYRA, Lovable was an important part of my development workflow.

Rather than treating it as a replacement for development, I used AI-assisted tooling to accelerate activities such as prototyping, exploring interfaces and moving ideas towards implementation.

The important part was not simply generating code.

The important part was deciding:

• What should be built?
• What should the user experience look like?
• How should different services communicate?
• What belongs in the frontend?
• What belongs in the backend?
• How should an external API be integrated?
• How should the resulting functionality behave in a real product?

AI could accelerate the path from an idea to an implementation, but engineering judgment remained essential.

React, Node.js and Tailwind CSS

My work also involved technologies that sit at the core of modern web development.

I worked with:

• React.js
• Node.js
• Tailwind CSS
• JavaScript
• REST APIs
• UI/UX
• n8n
• External service integrations

This allowed me to work across different parts of a product rather than focusing exclusively on one layer.

On the frontend, React and Tailwind CSS supported the development of modern interfaces.

On the backend, Node.js enabled application logic and integrations.

And between systems, APIs connected the different services required by the product.

Automation with n8n

Another important part of the workflow was automation.

I worked with n8n to create and manage workflows, connecting different services and reducing repetitive manual processes.

This is one of the areas where I find the combination of AI and automation particularly interesting.

AI can help accelerate the creation and exploration of solutions.

Automation can make those solutions repeatable.

But someone still needs to understand the business process being automated.

That distinction matters.

A technically impressive automation that solves the wrong problem is still the wrong solution.

Integrating Real-World Services

The work at CLYRA also involved integrations with external services, including:

WhatsApp API

Used as part of communication and platform workflows.

QR Codes

Used as part of the product's digital interaction flows.

Stripe

Used for payment functionality and plan management.

These integrations reinforced something that is easy to overlook when learning software development:

«A product is rarely just the interface.»

A real application needs its frontend, backend, APIs, authentication, automation, payments and user experience to work together.

Customer Support Changed How I Think About Software

For two months at CLYRA, I also worked in customer support.

(I am highlighting my corporate experience at Clyra because my prior experience involved more autonomous roles with fewer administrative constraints)

Customers contacted the team with questions and problems related to the platform, and we helped them understand and resolve those issues.

That experience was valuable because it put me on the other side of the product.

A feature can work technically and still be confusing to the person using it.

A workflow can be logically correct and still fail to solve the customer's actual problem.

This is one reason I consider customer understanding an important part of software development.

Building Without AI

AI-assisted development is powerful, but I don't consider it a requirement for building software.

I am also _comfortable developing projects _using conventional engineering practices and technologies without relying on AI to generate the solution.

One example is Kwendi, a project focused on technology, education and the learning of Umbundu.

The project involves product thinking, application architecture, UI/UX and the design of a learning experience.

The important lesson for me is simple:

AI is a tool. It is not the developer.

A strong developer should be able to work with AI when it creates value, but should also be able to reason about a problem, design a solution and build software independently.

What I Take From This

My approach to modern software development can be summarized in four ideas:

Understand the problem → Design the solution → Use the right tools → Validate the result.

Sometimes the right tool is AI.

Sometimes it is automation.

Sometimes it is simply writing the code yourself.

The technology should serve the problem — not the other way around.

And that is where I believe AI-assisted development becomes genuinely powerful: when it increases the developer's ability to experiment, learn and deliver without removing responsibility for the final solution.