I Started by Building Websites. The Real Problem Was Much Bigger.

When I first started working on websites for businesses in Nepal, I thought the problem was fairly simple: businesses needed a website. That was it. A business would tell me what it does, give me some photos and information, and my job was to turn those elements into a website that looked professional and worked well on a phone.

However, the more businesses I examined, the more I realized that the website was only the visible part of the problem. A business could have a website and still manage almost everything else manually. Customers would discover the business on Facebook or Instagram, send a message, and the owner would reply. Product details would be copied into another message. Orders would be written somewhere. Payments would happen through a QR code. Customer information might end up in a spreadsheet—or nowhere at all.

The website existed, but the system around it didn't. That realization eventually changed what I wanted to build.

At First, I Thought the Website Was the Product

My entry into this field began with something quite simple. I once overheard a teacher discussing getting a website made and mentioning that an IT company would charge around 400$. At that time, I didn’t know much about web development, but instead of questioning the price, I wondered:

What if I could build the website myself?
So, I started learning HTML, CSS, and JavaScript and began building websites. This experience taught me something crucial: the barrier wasn't always the technology itself. A lot of small businesses simply didn't have someone who could handle the technology for them.

This understanding became one of the reasons I started building HamroLink. At that stage, however, my focus was still centered around the website:

Build the pages.
Make them responsive.
Add the business information.
Add products if necessary.
Connect the domain.
Publish.
Done.

Except it wasn't really done.

Then I Saw What Happened Behind the Website

Imagine a small clothing business. A customer sees a product on social media and asks:

“Price kati ho?”
Then: “M size cha?”
Then: “Delivery Kathmandu ma huncha?”

After that, they send their address, and the business owner asks for payment. Someone has to remember the order. None of these tasks are particularly complicated individually. The problem arises when dozens or hundreds of customers are doing them simultaneously. The business starts depending on people remembering things. The website might showcase the products, but the actual business process is still happening through messages, spreadsheets, phone calls, and manual work.

This realization shifted my perspective. The question wasn't simply:

“How do I help a business build a website?”
It evolved into: “How do I make the digital side of running a business simpler?”
That is a much bigger problem.

The Website Became Just One Layer

Once I viewed the problem this way, the product naturally began to expand. A business might need:

Website
Customers need a place to understand what the business offers.
Commerce
They need to see products, variants, prices, and place orders.
Customer Management
The business needs to track who contacted them, who ordered, and what happened next.
Payments
Customers need local payment methods that make sense for them.
Marketing
The business should follow up with customers instead of starting from scratch every time.
Automation
Repetitive tasks should not require manual intervention every day.
AI
Eventually, people should interact with these systems in a more intuitive way.

This direction led me from merely thinking about a website builder to conceptualizing a connected business system. HamroLink's current website-builder platform reflects this evolution: alongside website creation, it connects online stores, inquiry CRM, WhatsApp communication, local payment gateways, and AI-assisted content tools. The important part isn't the number of features, but the connections between them.

The Hard Part Isn't Generating a Website

AI has made generating a website significantly easier. You can describe a business, generate pages, write copy, create layouts, and produce code rapidly. However, I don't believe the initial version of a website is the hardest problem to tackle.

The more challenging issues arise after the website has been generated:

What data does the business possess?
Where are its products stored?
How are customers represented?
How does an order change the business state?
What happens when a customer submits an inquiry?
Who is authorized to make changes?
How do different parts of the system remain synchronized?
How can an AI agent make modifications without disrupting the underlying application?

These questions are far more intriguing to me than merely generating another landing page. A good AI interface still requires a reliable system behind it. Otherwise, it results in a sophisticated chatbot sitting atop a chaotic structure.

This Is Where AI Agents Become Interesting

I don't believe the most exciting use of AI in business is simply:

“Write me some website text.”

While that is useful, it's just the tip of the iceberg. The more interesting interface resembles:

“I run a clothing store. Add these products, create size and color variants, and make them available for online orders.”

Or:

“Show me which vehicles need maintenance this month.”

Or:

“Send an update to the customer about this order.”

The user shouldn't need to know which screen contains the required setting. They shouldn't have to understand the database structure. They should merely describe the outcome they desire, and the agent can work with the underlying business system to bring it to fruition. This transformation alters the role of AI from content generation to an interface for software, and that is the direction I find more fascinating to explore.

But Building This for Nepal Changes the Problem

Creating software for a local market isn't just a matter of translating an international product into Nepali. The infrastructure matters. A Nepalese business might require eSewa, Khalti, or Fonepay instead of assuming every customer possesses an international card. It might want a .com.np domain and communicate with customers through WhatsApp. Often, a very small team means the owner simultaneously handles sales, operations, and customer support.

These details significantly influence what the product needs to accomplish. For instance, HamroLink's website-builder platform currently supports local payment options including eSewa, Khalti, and Fonepay, along with .com.np domain support. While these may not seem exciting on a feature list, for the person managing the business, they can determine whether the system aligns with their reality.

Thus, I don't want to create another generic website builder and merely label it as “Nepal.” The local context must be embedded within the product.

What I'm Trying to Build With HamroLink

Ultimately, this is what HamroLink has come to represent for me. I began with the idea of helping businesses create websites, only to realize that businesses don't merely want websites. They desire the outcomes that those websites are supposed to facilitate:

Customers finding them.
Customers understanding their products or services.
People placing orders or making inquiries.
Awareness of customer interactions.
Reduction of repetitive tasks.

They want all of this without needing to become software engineers. Therefore, while the website remains important as the public face of the business, it should not exist as an isolated entity.

If you start from the website side, you can see what I'm building toward in the HamroLink website builder. The concept is straightforward: start with the website, then connect the various pieces of the business around it.

Where I Think This Goes Next

I don't believe the future entails every small business acquiring a massive suite of complex software. Quite the opposite may occur. The underlying software could become more sophisticated while the interface simplifies. Today, we often learn software by navigating its interface:

Open this page → click this button → find this setting → enter this information → save.

With AI agents, the interface can increasingly become: Tell the system what you want done.
However, achieving this reliability requires more than a powerful model. It necessitates structured business data, clear permissions, predictable workflows, validation, and a system that understands what has changed after an action is performed.

That is the engineering challenge I find intriguing. Not just: “Can AI build a website?”
But: “Can AI become a useful interface for the systems a business actually depends on?”
That's a much more compelling question. And for me, that is where the journey from building websites starts to evolve into something much larger.