Digitizing Vrajatiyans: replacing the brochure, not the agent

Vrajatiyans is a travel agency that works. Customers reach it over WhatsApp and phone calls; brochures go out by hand; packages are confirmed in conversation; documents are collected manually; payments are coordinated person to person; hotels, drivers, and tickets flow through a network of vendors the agency has trusted for years. Nothing about it is broken. That's the first thing to understand before writing any code — and it took longer to understand than any code I've written since.

The business as I found it

Every step of the workflow was a human being doing something a computer could do — and, crucially, a few steps where a human was doing something a computer couldn't: judging which vendor to trust for a last-minute change, reading a customer's hesitation over the phone, knowing which hotel quietly renovated last season. A naive "digitize everything" plan would have automated away the parts that make the agency worth choosing.

So the design constraint became the principle for the whole project: replace the brochure, not the agent.

Three stages, on purpose

Stage one — discovery. An online platform where customers can browse travel packages instead of waiting for a PDF over WhatsApp. No accounts, no payments, no workflow change for the agents. The narrowest possible slice that delivers value, and a way to earn the business's trust in the process itself.

Stage two — transactions. Authenticated accounts, online booking, digital payments — with the human agents deliberately kept in the loop for confirmation and judgment. The system proposes; the agent disposes.

Stage three — operations. Secure document collection, vendor coordination, hotel confirmations, driver management, administrative approval, automated ticket generation. The invisible machinery — and the hardest part, because this is where the vendors' trust network meets a database schema.

What stage one taught me

The difficult engineering wasn't CRUD. It was requirements discovery: sitting with how the business actually operates, finding the boundary between what should be software and what must stay human, and staging the transformation so that no single step risks the operation that pays everyone's salaries. Generating code is cheap now. Understanding a business deeply enough that the code improves it — that's the work.

Stage two is in progress; this study grows as the stages ship.