Everyone Says Coding Is Dead. My Clients Prove Otherwise.

Yet again, it happened. Just last week. Another client was arguing with me that coding is dead. That people like me do not have jobs anymore. If only I get a dollar every time I hear this.

Usually, this happens after they discover that some new AI coding tool that can generate a login page, a contact form, or a nice-looking dashboard in 23 seconds. Wow!

Then suddenly everybody becomes a software architect.

Very powerful.

Next week maybe the same person will discover Canva and declare graphic designers dead also.

But let me say this as someone who has spent about 30 years building websites, web apps, mobile apps, CRM systems and custom software for Singapore businesses:

Coding is not dead! I repeat. Coding is NOT dead.

What is dying is the old idea that coding is only about typing lines of code like a hardworking keyboard uncle in a dark room.

That part is changing. AI coding tools, vibe coding, GitHub Copilot, ChatGPT, Cursor and all these new toys can definitely help developers move faster.

But my clients do not come to me because they are missing "some code".

They come to me because their business process is messy. Their staff are doing the same work three times. Their Excel file has become a national security threat. Their CRM does not match how they actually work. Their website brings enquiries, but nobody follows up properly. Their custom system was built by someone who disappeared into another dimension.

So yes, AI can write code.

But can it sit with a business owner and understand why the quotation process is stuck?

Can it explain why the sales report is wrong because three departments define "confirmed customer" differently?

Can it tell your admin staff, "Actually, the problem is not the button. The whole workflow before the button is confused"?

That one, my friend, is where real software work begins.

AI coding is real. Let's not pretend otherwise.

First, let us be fair.

AI coding is not a joke.

GitHub Copilot research found that developers using Copilot completed a coding task much faster than developers without it. That is not small. That is not "wah maybe got slight improvement only". That is serious.

AI tools can help with generating simple code, explaining unfamiliar code, writing boilerplate, creating test cases, suggesting bug fixes, drafting documentation and speeding up repetitive coding tasks.

I use AI too.
(Well, you won't believe me even if I say otherwise, right?)

Why would I not?

If there is a tool that helps me move faster, I will use it. I am not going to sit there and proudly type everything manually like it is 1998 and I am trying to win a typing competition.

But there is a big difference between:

AI helps developers work faster

and

AI means nobody needs developers anymore.

That jump is where many people fall down, roll three rounds, and still insist they meant to do that.

Vibe coding looks easy because the demo is always easy.

Vibe coding is very seductive.

You type something like:

Build me a CRM for my sales team with dashboard, invoice tracking, WhatsApp follow-up, user roles, reports, and make it modern.

Then the AI generates some screens.

Wah. Got sidebar. Got chart. Got button. Got login page.

Business owner sees it and thinks: "Finished already what."

No, boss. That is not finished.

That is the software equivalent of looking at a showroom kitchen and assuming dinner is cooked.

A serious business system is not just the screen.

It is the logic behind the screen.

For example:

• Who can see which customer?
• What happens when a lead becomes a client?
• Can two staff edit the same record at the same time?
• What if the customer has two branches?
• What if the invoice is partially paid?
• What if the sales person leaves the company?
• What if the boss wants monthly report by service type, but finance tracks by invoice category?
• What if somebody clicks delete and suddenly the office becomes very quiet?

These are not "coding problems" only.

These are business thinking problems.

AI can generate code. But someone still has to decide what the code is supposed to do.

My clients don't ask for code. They ask for relief.

Most clients do not come to me and say:

Anees, please write 3,000 lines of PHP, some JavaScript, and a database schema.

Nobody talks like that, unless they need sleep.

They say things like:

• My staff are wasting too much time.
• Our current system cannot handle this.
• We are using too many Excel files.
• The website enquiry comes in, but we don't know who followed up.
• Can we automate this?
• Can I see everything in one dashboard?
• Can the system remind my staff?
• Can customers submit this online instead of WhatsApping ten screenshots?

See the difference?

The client is not buying code.

The client is buying clarity, speed, control, fewer mistakes, better follow-up, and less daily headache.

Code is just the material.

Like cement.

Nobody buys a house because they love cement. They buy the house because they need a place to live.

Same with custom software development.

The real value is not the code itself. The value is the business result the code creates.

AI can build a form. But can it understand the follow-up?

Let us use a simple example: a contact form.

AI can build a contact form very quickly.

Name. Email. Phone. Message. Submit button. Done.

Clap clap.

But for a real Singapore SME, the important questions come after the form:

• Where does the enquiry go?
• Who gets notified?
• Should it go to email, CRM, WhatsApp, or all three?
• What if the phone number is fake?
• What if the customer selects urgent?
• What if nobody follows up within 24 hours?
• Should the lead be tagged by service type?
• Should the website record which page the enquiry came from?
• Should the boss see a weekly enquiry report?
• Should repeat enquiries from the same email be grouped?

This is why "AI can code" does not automatically mean "AI can build your business system".

The coding part is only one slice of the cake.

And sometimes not even the most important slice.

The problem is usually not the code. It is the workflow.

After many years of building systems, I can tell you something very confidently:

Many software problems are actually workflow problems wearing a software costume.

A client may say:

"The system is slow."

But after checking, the real issue is that five people are entering the same information in different places.

Another client may say:

"We need a dashboard."

But the real issue is that nobody agrees which numbers are correct.

Another client may say:

"We need AI."

But the real issue is that their current process is not even clear enough for a human to follow.

This is why I always say:

Do not automate confusion. You will only get faster confusion.

Very advanced confusion, maybe with charts.

Before coding, before AI, before automation, before the exciting dashboard with blue gradient and the word "insights" everywhere, the business process must make sense.

That is where experienced developers still matter.

Not because we type magic code.

But because we know where business logic usually breaks.

AI is good at generating. Humans are still needed for judging.

AI coding tools are impressive because they generate things fast.

But speed is not the same as correctness.

Many developers are now using AI tools, but trust remains a major issue. Developers like the speed, but they are still careful about accuracy.

And this makes sense.

AI can confidently produce code that looks correct, sounds correct, and fails quietly in production like nothing happened.

Very polite disaster.

A serious developer must still check security, database design, error handling, user permissions, performance, edge cases, mobile responsiveness, browser behaviour, integration logic, data validation and long-term maintainability.

For a toy project, maybe you can vibe your way through.

For a business system handling customers, invoices, leads, payments, reports, inventory, or staff operations, "looks okay lah" is not a testing strategy.

Real example: the Excel file that became too powerful.

Many Singapore SMEs have one Excel file that secretly runs the whole company.

You know the one.

It is usually called something like:

Final_Updated_NEW_latest_v7_REAL_USE_THIS_ONE.xlsx

Inside this file, there are customer records, payment status, job notes, quotations, delivery info, staff remarks, and possibly the meaning of life.

One staff member understands it.

If that person takes leave, the whole company prays.

Now imagine someone says:

"No need developer. Just ask AI to build a system."

Okay. Build what?

What are the columns? Which columns are still used? Which are old? Which data is wrong? Which process should stay? Which process should be redesigned? Who can edit? Who can approve? What report does the boss need? What happens to old records?

AI can help generate parts of the system.

But someone still has to untangle the business mess.

That is not dead coding.

That is CRM and business system development.

The new developer is not just a coder.

I think the future developer is not someone who only types code.

That type of developer will struggle.

The future developer must be part programmer, part business analyst, part workflow cleaner, part debugger, part system designer, part client translator, part security checker and part reality inspector.

Yes, reality inspector.

Because sometimes the client says, "Very simple only."

Then you ask three questions and discover there are seven user roles, two approval flows, three exception cases, and one staff member called Jenny who has been manually fixing everything since 2016.

"Very simple" is usually where the adventure begins.

This is why clients still need experienced developers.

Not because AI cannot write code.

But because AI does not know Jenny.

Vibe coding is fine for learning. Be careful with serious systems.

Let me be clear.

I am not against vibe coding.

For learning, experimenting, prototypes, mockups, personal tools, landing pages, small scripts, and quick ideas, it can be amazing.

If you want to test an idea quickly, go ahead.

If you want to understand how a page might look, go ahead.

If you want to build a small internal helper, go ahead.

But if you are building something serious, be careful.

Serious systems need proper requirements, data structure, user roles, secure login, audit trails, backups, error handling, testing, documentation, maintenance and a plan for future changes.

This is where "just vibe it" becomes "why is the system broken after three months?"

And then suddenly coding is not dead anymore.

Coding is urgently needed, preferably before lunch.

What Singapore SMEs should do before building with AI.

If you are a Singapore business owner thinking about AI coding or custom software, here is my practical advice.

1. Start with the business problem.

Do not start with: "I want an AI system."

Start with: "What problem are we trying to solve?"

Examples include losing enquiries, repeated data entry, slow reports, customers not getting updates, job status being hard to track, a CRM that does not match your workflow, or too much dependence on Excel.

A clear problem leads to better software.

A vague problem leads to expensive decoration.

2. Map the current workflow.

Before building anything, write down what happens now.

Who receives the enquiry? Who follows up? Where is the data stored? Who approves? What happens if something is missing? Where does the customer get updated?

You do not need a fancy diagram. Even a simple list is enough.

If you cannot explain the workflow, AI cannot magically understand it.

3. Decide what must be custom.

Not everything needs custom software.

Some things can use ready-made tools.

Some things can use automation.

Some things can use AI.

Some things need a proper custom system.

The trick is knowing which is which.

A good developer should not force custom software for everything. Sometimes the best answer is a simpler tool, a better form, a cleaner CRM setup, or a small automation.

Yes, I know. Very shocking. A developer saying not everything needs custom development. Please remain calm.

4. Use AI to speed up, not to skip thinking.

AI can help with drafts, code suggestions, test cases, documentation and prototypes.

But do not use AI as an excuse to skip planning.

The better your thinking, the better AI can help.

The worse your thinking, the faster AI helps you produce nonsense.

5. Budget for maintenance.

Software is not a one-time event.

Your business changes. Your staff change. Your workflow changes. Your customers change. Your reporting needs change.

So your system must be maintainable.

This is one of the biggest problems with AI-generated code used blindly: it may work today, but nobody understands it tomorrow.

And tomorrow always comes.

Usually after the invoice is paid.

Is coding dead? No. The job description is changing.

Coding is not dead.

But the meaning of coding is changing.

The value is moving away from simply typing syntax.

The value is now in understanding the business, asking better questions, designing the right workflow, choosing the right tools, building maintainable systems, checking AI-generated work and protecting the client from expensive mistakes.

AI will make good developers faster.

It may make careless developers more dangerous.

And it will make business owners think more carefully about what they are actually buying.

Because if all you need is a simple demo, AI can help.

But if you need a system that your staff depend on every day, that handles real customer data, that supports your sales process, that produces reports you can trust, and that can grow with your business, then you still need proper software thinking.

My clients prove coding is not dead.

Every time a client comes to me with a messy workflow, coding is not dead.

Every time a business wants to replace repeated manual work, coding is not dead.

Every time a boss wants one dashboard instead of five Excel files, coding is not dead.

Every time a website enquiry needs to flow into a CRM, trigger a WhatsApp follow-up, and appear in a report, coding is not dead.

Every time an AI-generated system breaks because nobody understood the database, coding is definitely not dead.

It is alive.

Maybe drinking kopi quietly in the corner, waiting for the hype to calm down.

Final thought.

AI is powerful. Vibe coding is interesting. Coding tools are improving very fast.

But business software is not just about producing code.

It is about solving real problems for real people in real companies with real constraints, real staff, real customers, real mistakes, and real consequences.

So the next time someone says "coding is dead", ask them:

Dead for what?

Dead for a demo page? Maybe.

Dead for serious business systems? Not even close.

If your business is stuck with messy spreadsheets, repeated admin work, broken follow-up, confusing reports, or a half-working system, the answer is probably not "just ask AI".

The better answer is:

Understand the workflow. Fix the logic. Use AI where it helps. Build properly where it matters.

That is not dead coding.

That is coding growing up.

And honestly, about time.