Inspiration:

Small-business owners are experts at running their businesses — they shouldn't also have to become website designers.

Think about a local salon owner, plumber, caterer, tutor or restaurant owner. To build a website today, they usually have to hire someone, learn a website builder or learn how to communicate with an AI coding tool.

We asked a much simpler question:

What if building a website were as easy as making a phone call?

That became TikkaBurrito

What does TikkaBurrito do?

A business owner calls TikkaBurrito and simply talks about their business.

The AI asks the kinds of questions a web designer would ask: What does the business do? What services do you offer? What style do you want? What information should customers see?

TikkaBurrito then turns that conversation into a working website, checks the website, publishes it online and sends the live link back to the owner.

Customers can also communicate through WhatsApp, voice notes or text, and the platform supports roughly 30 languages.

Later, the owner can call or message again and say something like, “Change the colors,” “Add this service,” or “Use this photo,” and the AI updates the existing website.

How does AI transform the business?

In most website businesses, humans do the interview, create the requirements, design the site, write the code, test it, publish it and make future changes.

At TikkaBurrito, AI performs most of those jobs.

AI talks with the customer and understands what they want. AI plans the website. AI creates it. AI checks whether it works. AI reviews the finished site. AI publishes it. If something goes wrong, another AI process attempts to diagnose and repair the problem.

The customer can then come back later and ask the AI to make changes.

AI isn't an extra feature inside TikkaBurrito. AI is what allows TikkaBurrito to operate like a web agency without needing a traditional web-agency workforce for every customer.

How we built it

We built TikkaBurrito as a cloud-based AI business.

Gemini provides intelligence for understanding information, generating structured outputs and reviewing work.

Google Firebase and Firestore remember users, projects, conversations and build jobs so the system does not forget what a customer asked for.

Google Cloud Storage stores generated website files and customer photos.

Vapi handles phone conversations, WhatsApp handles messages and voice notes, and Twilio sends SMS messages.

Railway and Vercel run and publish the websites.

All of these pieces are connected by our own AI orchestration system, which takes a customer from conversation → website → quality check → live deployment.

Challenges we ran into

The hardest challenge was discovering that talking to AI is very different from filling out a form.

People pause. They change their minds. Calls disconnect. They may speak in noisy environments. They might call back three days later and simply say, “Make it darker,” expecting the AI to remember what they mean.

We had to teach the system to preserve the conversation, remember the customer's project, handle interruptions and continue working even after the phone call ends.

The second major challenge was reliability. Generating code is easy; reliably giving a nontechnical customer a working website is much harder. We therefore added several layers of checking, including automated testing, AI review and a final visual review of the live site.

Accomplishments that we're proud of

We are proud that TikkaBurrito is not just a demo where AI generates some code.

A customer can start with a normal conversation and end with a real website online.

The same system works across six ways of communicating — phone calls, WhatsApp text, WhatsApp voice notes, WhatsApp calls, browser voice and regular text chat.

It can work in roughly 30 languages, remember previous conversations and recover from many build failures without asking the customer to understand a technical error.

The goal is simple: the customer should never need to know how complicated the technology behind the experience is.

What we learned

Our biggest learning was that removing an interface is actually harder than creating one.

A normal website builder has buttons, forms, menus, confirmation screens and undo buttons.

When the interface is simply a conversation, the AI has to understand uncertainty, remember context and know when to ask another question.

We also learned that customers naturally describe what they want in messy human language. Preserving what they actually said and letting AI structure it afterward works better than forcing them to speak like a software engineer.

What's next for TikkaBurrito

Our next goal is to make TikkaBurrito accessible to millions of small businesses that are primarily phone-first rather than computer-first.

We are expanding phone access in India, improving the kinds of applications businesses can create and reducing the cost of every build by assigning different AI tasks to the most efficient model.

Long term, we believe starting and operating a digital business should be possible simply by talking.