I Don't Know How to Code. But AI Helped Me Build Working Systems. Here's What Nobody Tells You.
I Don't Know How to Code. But AI Helped Me Build Working Systems. Here's What Nobody Tells You.
From Git errors and sleeping servers to real email automation — a non-developer's journey into building software with AI.
Tags: #ai #beginners #webdev #discuss
I'm Not a Developer. Let's Start There.
I don't come from a software engineering background.
I don't wake up thinking about JavaScript frameworks, REST APIs, environment variables, or deployment pipelines.
I don't have years of programming experience.
But I do understand something important:
I know what problem I want to solve.
And recently, I discovered that knowing how to describe a problem, ask the right questions, test the results, and keep improving can take you surprisingly far.
Not without frustration.
Not without mistakes.
And definitely not without a few cups of tea. ☕
This is the story of how I worked with AI to turn ideas into functioning websites, inquiry systems, and email automation.
It isn't a story about AI magically replacing programmers.
It's about something more interesting.
What happens when someone who understands a problem starts collaborating with technology they don't fully understand yet?
It Started With a Simple Requirement
I was working on two projects:
DeTLeng Automation, a website where people can describe repetitive manual work they want to automate.
And AL-NOOR Asian Educational Centre, an educational website where students can explore courses and request information.
The websites existed. But I wanted them to do more than display information.
For example, the educational website had a consultation button that opened the visitor's own email application with a prewritten message.
That worked, technically.
But it wasn't the experience I wanted.
I wanted students to fill out a proper form directly on the website.
They should be able to choose a course, enter their email address and WhatsApp number, ask questions, and submit everything without opening Outlook or Gmail.
Our team should receive the enquiry.
The student should receive a professional confirmation email.
And a clear success popup should appear.
Simple, right?
Well, welcome to software development.
Suddenly, I Was Learning Words I Had Never Needed Before
To accomplish what sounded like a small task, I found myself dealing with:
- Git and GitHub: commits, branches, rejected pushes, remote changes, backups and version history.
- Render: deploying a Node.js backend, configuring environment variables, understanding free-instance cold starts.
- Resend: API keys, verified sending domains, email templates and delivery logs.
- Cloudflare and Namecheap: DNS records, DKIM, SPF-related configuration and domain verification.
- Frontend and backend integration: forms, validation, CORS, sending two emails, and showing the correct result to the user.
I didn't begin by knowing all these things.
I learned about them because I kept asking AI what we needed to do next — and why.
Sometimes I asked the same question more than once.
Sometimes I shared screenshots instead of technical explanations.
Sometimes I simply said:
"Please don't disturb the existing website. The chatbot is already working!"
That was a real concern.
For an experienced developer, changing a few files may seem routine.
For me, those files represented months of work that I couldn't afford to lose.
So we needed to be careful.
And that became one of my most valuable lessons:
Building something new means very little if you accidentally break what already works.
When GitHub Said No
At one point, I had the changes ready to publish.
The files were modified.
The new inquiry form was prepared.
The code had passed local tests.
I ran the Git push command.
And GitHub rejected it.
The remote repository contained changes that my local copy didn't have.
To a developer, that's a familiar Git synchronization problem.
To me, it initially looked like the entire project had gone wrong.
But instead of treating the error as the end, I brought the terminal output back to AI.
We examined what happened, fetched the remote changes, and discussed how to protect the existing work.
I also had a separate local backup.
This taught me something beyond memorizing commands:
A good AI workflow isn't just "give me code." It's "explain the risk, show me the options, and help me understand what I'm about to change."
AI can suggest a command.
But the person operating the system still has to understand the consequences.
Especially when real websites and real users are involved.
Then Came the Email Adventure
I wanted a professional email system.
But I didn't want to buy a separate paid mailbox just to send automated notifications.
That led us to Resend.
We configured sending domains and verified DNS records through Cloudflare.
Our websites could then send transactional email using branded addresses.
For example, DeTLeng Automation sends from a verified address on its mail.detleng.com subdomain.
AL-NOOR uses its own verified sending subdomain.
Incoming business emails, meanwhile, can be forwarded to an existing mailbox through email routing.
It sounds straightforward when described in one paragraph.
But the actual experience involved checking DNS records, adding API keys, configuring Render, testing email delivery, and investigating problems.
And then something interesting happened.
The email was being sent. But the recipient couldn't see it in the inbox.
It had landed in Spam.
At first, this felt like another failure.
But it was an important distinction: sending an email successfully and getting it into someone's primary inbox are not the same thing.
We tested Gmail and Outlook, inspected the results, and confirmed that the email content was arriving.
This wasn't just coding.
It was learning how real systems behave.
The Server Went for a Tea Break ☕
Here is my favourite part.
We built an inquiry form.
I filled it out.
I clicked Submit.
And waited.
And waited.
Sometimes it took around 40 to 60 seconds for the success response to appear.
I jokingly asked AI:
"Did the server stop at its uncle's house for a tea party?"
The real explanation was less entertaining.
We were using a free Render instance, which can go to sleep after inactivity.
When a new request arrives, the service may need time to wake up.
That delay can affect how quickly the form receives its response.
For someone learning software development, this was a powerful discovery.
The problem wasn't necessarily my form, the API code, or the email service. It could be the hosting environment itself.
I had been thinking about a website as a collection of pages.
Now I was beginning to understand it as a collection of connected systems.
Every part matters.
The Moment It Actually Worked
Eventually, we had what I originally wanted.
A visitor could fill in a form.
The backend could process the request.
The business could receive an enquiry notification.
The visitor could receive an acknowledgement email.
And the website could display a professional success popup with a reference number.
We tested these behaviours rather than relying only on AI saying the code looked correct.
On the AL-NOOR website, we also improved the experience by replacing the old email-based consultation workflow with a dedicated course enquiry form.
We added required contact information, including a WhatsApp number with a country code, and a proper submission experience.
The important part wasn't the amount of code written.
It was seeing the whole journey work.
Someone enters information. A real backend processes it. Real emails are sent. The user receives confirmation.
That was the moment the technical vocabulary started making sense.
Not because I had memorized every concept.
Because I could connect each concept to an actual result.
So, Did AI Turn Me Into a Developer?
Not overnight.
And I wouldn't claim to have the same knowledge as someone who has spent years studying software engineering.
Experienced developers understand architecture, maintainability, security, debugging, and failure scenarios at a depth I'm still learning.
Their knowledge matters enormously.
But this experience changed how I think about my own abilities.
Previously, I might have looked at a project involving APIs, backend servers, deployment, and DNS configuration and thought:
"This is not for someone like me."
Now my reaction is different:
"I don't know how to build this yet. But can I break it into smaller problems and work through them?"
That single change in thinking is powerful.
AI didn't eliminate the need to learn.
AI changed how I could begin learning.
I could describe a requirement in ordinary language, ask for an explanation, examine the proposed solution, try it, report an error, and improve it.
I didn't have to understand everything before starting.
But I did have to stay involved.
I had to test things.
I had to recognize when something wasn't working.
I had to ask better questions.
And occasionally, I had to tell AI to stop overcomplicating things and help me finish the task!
Maybe We're Asking the Wrong Question About AI
A lot of discussions focus on whether AI will replace developers.
I think there's another question worth exploring:
How many people have useful ideas but never attempt to build them because programming looks inaccessible?
Think about a teacher who wants to automate student enquiries.
A small business owner who spends hours preparing the same reports.
An administrator who manually copies spreadsheet data every morning.
A professional who knows exactly where time is being wasted but doesn't know what an API is.
These people may not be programmers.
But they understand their problems.
And that understanding has value.
With AI assistance, documentation, patience, testing, and responsible technical support, some of these ideas can become working tools.
Not every project will be simple.
Not every AI-generated solution will be correct.
And serious systems still require proper security, maintenance, and engineering review.
But the starting line has moved.
You don't need to know everything about software to begin exploring what software could do for you.
I Built Something. Now I'd Love You to Try It.
This journey led me to work on DeTLeng Automation.
The idea behind it is simple:
Start with the work you're doing manually, not with a technology you think you need.
You don't have to tell us whether the answer is Python, JavaScript, an AI agent, or an API.
Just explain your problem.
I'd love to invite the DEV Community to try our Automation Desk:
👉 DeTLeng Automation — From Repetitive Work to Working Systems
Open the page and fill out the form.
Tell us:
What are you doing manually?
Perhaps you're copying data, preparing reports, answering repetitive enquiries, or performing the same administrative task every day.
What would you like to happen instead?
Describe the result you wish you could achieve automatically.
How often does this happen?
Every day? Every week? Every customer request?
Submit your brief.
The system is designed to receive your request, send an acknowledgement, and bring the enquiry to our team for review. It isn't a promise that AI will instantly build a complete solution, but it is a practical first step toward exploring one.
And if you're a developer, I'd particularly welcome your feedback on the experience.
Because this project isn't about proving that developers are no longer needed.
It's about making communication between people with problems and people who build solutions easier.
My Message to Non-Developers
If you're reading this and thinking:
"I have an idea, but I don't know programming."
I understand that feeling.
I was there.
I'm still learning.
But don't underestimate the value of understanding your own work.
Start with one problem.
Explain it clearly.
Use AI to explore possible solutions.
Ask questions until you understand the important parts.
Test what gets built.
Learn from what breaks.
And respect the engineering that makes reliable systems possible.
You may discover that you can contribute much more to a technical project than you previously imagined.
And My Message to Developers
Your experience is not less valuable because more people can generate code.
If anything, making code easier to produce makes good engineering judgment even more important.
People like me need developers who understand what makes systems reliable, secure, and maintainable.
But perhaps developers can also learn something from non-technical people who understand particular industries, workflows, customers, and everyday frustrations.
The most interesting future isn't necessarily developers versus non-developers.
It may be developers, domain experts, and AI collaborating to build things none of them would have built in quite the same way alone.
I started with a simple request.
I wanted forms to work, emails to arrive, and visitors to receive a proper confirmation.
Along the way, I encountered Git conflicts, DNS records, deployment settings, spam filters, backend tests, and a server that occasionally needed waking up.
I'm still not a professional programmer.
But I'm no longer someone who assumes that building software is completely beyond my reach.
And that, for me, is real progress.
I didn't start by learning how to write code. I started by learning how to communicate a problem.
Maybe that is where more people should begin.
Over to you, DEV Community ❤️
Have you ever built something outside your technical comfort zone with AI?
If you're an experienced developer, what advice would you give to non-developers who want to build responsibly?
And if you have a repetitive task you'd love to simplify, I'd love you to try the DeTLeng Automation Desk and share your thoughts.
You bring the problem. AI can help you explore the possibilities. Human judgment turns the possibilities into something useful.
Let's build more, learn more, and make technology accessible to more people.
— A non-developer who is still learning, still asking questions, and still building.
Comments
Post a Comment