How Veeza AI turned WhatsApp into its front door
Veeza AI runs its visa agent on Keiki, so customers across the Global South can apply, pay and get updates without leaving WhatsApp. Since the move, Veeza AI has doubled revenue and tripled usage.
Short answer: Veeza AI helps people with weaker passports apply for tourist visas, and its customers do everything on WhatsApp. When support demand doubled every week, founder Aly Moursy moved its WhatsApp agent to Keiki, which now answers customers in seconds, resolves most conversations without a person, and starts visa applications inside the chat.
Veeza AI helps people with weaker passports apply for tourist visas across the Global South, with the Middle East and North Africa as its launch pad. Its customers don't browse websites. They message on WhatsApp. When demand doubled every week and its founder, Aly Moursy, was answering all of it himself, Veeza AI moved its WhatsApp agent to Keiki. Support stopped being the bottleneck and became Veeza AI's best sales channel.
- revenue since moving to Keiki
- 2×
- usage across Veeza AI
- 3×
- of messaging-infrastructure upkeep saved every week
- 20+ hrs
The company
Veeza AI is a San Francisco company in Y Combinator's Fall 2026 batch. It opened in closed beta in April, moved to public beta in June, and serves travelers in Egypt first.
Five billion people need a visa for almost every trip they take. Veeza AI's goal is to become the operating system for their legal global mobility. Today, it is focused on tourism: the visas people need just to take a trip.
The product takes the hardest part of travel and makes it simple. A UK visitor visa takes 68 steps on the government website and 9 on Veeza AI. Documents are read automatically, questions are merged, and a work-history section becomes "send us your CV." An AI reviewer checks every application for errors and missing documents, and a human visa expert reviews it before submission. Pricing is just as simple: free to start, and you pay only when your application is ready, a one-time service fee of $40 to $80 per traveler, depending on the country, with up to 25% off for groups. Veeza AI also provides certified translations, which embassies often ask for.
The strategy depends on timing. A visa is the earliest signal that someone is about to travel. General travel companies pay heavily to find the small share of people who will take an international trip. Veeza AI meets them at the very first step.
That makes one surface more valuable than any other: the conversation. Every question, update and payment happens there.
WhatsApp is the internet
Aly studied engineering at Carleton University in Canada, worked as a product manager and UX lead at Bell Canada, co-founded the same-day delivery app Dlvvr, and later led the launch of Efreshli AI. He planned Veeza AI the way he had built products before: a polished web experience with conventional support tooling behind it.
"For a regular person in the Global South, WhatsApp is the gateway to the internet. You can't force your customers to follow things you believe in. You need to meet them where they are."
Aly Moursy, founder of Veeza AI
The turning point was small. Customers were messaging Veeza AI's WhatsApp line to ask for the link to Veeza AI's own dashboard. WhatsApp wasn't a support channel for the product. It was how people got into the product. Even after paying, most customers didn't log in again. They asked for status updates, appointment dates and payment links in the same thread, and when the visa came through, they asked the same thread for a flight and a hotel.
Aly traces the habit back to the region's internet history. Broadband was expensive, chat used very little of it, and messaging became the default way to get anything done. When connections improved, the habit stayed. People with a problem don't want another login. They want an answer.
Before Keiki
Growth arrived faster than one person could answer
Veeza AI began running click-to-WhatsApp ads: "Get a visa. Start on WhatsApp." They worked. Word spread across Egypt and well beyond it, and support conversations started doubling every week.
All of them reached Aly.
Every conversation was a sale, not a ticket
Most of the people writing in hadn't bought anything yet. They were asking whether Veeza AI could help, what it cost and how long it would take. A slow or confusing reply wasn't a support failure. It was a lost customer.
Trust matters too. People are sending passports and paying in dollars to see family they haven't seen in years. Aly had been winning them over personally, by answering quickly, showing his face and jumping on calls. That works for a few hundred people a week. It doesn't work for thousands.
The bots couldn't carry the conversation
Veeza AI had tried to automate. Its first WhatsApp bot was deterministic: it copied the website's step-by-step flow into chat. Customers don't talk in steps, and the bot broke whenever someone asked something out of order.
The second version ran on one long system prompt with a single tool for creating payment links and a handoff to a human. It kept the line open, but it couldn't look up an application, check a country's requirements or start a new one. Anything beyond a payment link came back to Aly, and he had to jump between the bot, the dashboard and the customer to answer it.
It also handled only text. Customers write in Arabic, in English and in Franco-Arabic (Arabic in Latin letters), and many send voice notes. The bot couldn't understand voice notes at all.
The founder couldn't see what was happening
With thousands of conversations a week, Aly had no clear view of which ones were going well, which customers were frustrated and which were about to buy. Finding the conversations that needed him meant reading them.
Building it in-house wasn't an option
Veeza AI could have built a proper agent: model hosting, tool calling, memory, voice, human handoff, monitoring. But that meant months of infrastructure work for a solo founder, at exactly the moment the top of the funnel was taking off. Experimenting on live customers was too risky.
So Veeza AI did the only thing it could. It cut its ad spend. Its fastest-growing channel had become the thing holding the business back.
After Keiki
Veeza AI deployed its WhatsApp agent on Keiki and connected it to its own service over API and MCP. Keiki hosts the agent, runs the model, delivers messages, keeps context and records everything it does. Veeza AI owns the personality, the instructions and the business logic.
An agent that understands people, from a simpler prompt
Keiki's agent runtime handles tool use, memory and conversation flow, so the instructions only have to describe Veeza AI. Aly replaced the long, rigid prompt with a short one, and the agent got noticeably better.
"It just understands user intent so much better. With Keiki I was able to write a much simpler system prompt, and it's smart, and it's funny."
Aly Moursy, founder of Veeza AI
Aly modeled the agent on how he sells: quick, warm and clear about what happens next, so a nervous customer feels safe handing over a passport.
It speaks the way customers speak
Arabic, English, Franco-Arabic and WhatsApp voice notes all arrive in the same thread, and the agent answers in kind. A customer can send a voice note in Egyptian Arabic and get a payment link back.
It does the work, not just the talking
The agent works inside Veeza AI's systems. In one conversation it can:
- look up the customer's account and the status of their application
- check what a traveler with that passport needs for that country
- quote the price
- save the customer's answers as they go
- start a new application
- send a payment link the customer can claim straight from the chat
A question becomes a paid application without the customer opening a browser.
Aly steps in only where he's needed
Each conversation carries a live mood signal, such as about to buy, frustrated or interest lost. Conversations that need a person are flagged. Aly can take over a conversation, or steer the agent while it keeps talking, and hand it back afterwards. Instead of reading thousands of threads, he goes straight to the few that need him.
Every conversation is visible
Traces show each step the agent took, the tools it called and why, and transcripts show exactly what customers said.
It improves every day
Veeza AI reviews the previous 24 hours of conversations each morning: what converted, what confused people, where the agent should have handed off. Fixes go straight back into the agent the same day. Because Keiki routes between models, Veeza AI can move to a newer model without rebuilding anything.
Aly can build again
Aly could have built all of this himself. He chose not to. With Keiki he can see every conversation, test every change and ship improvements the same day, and he's building with confidence again instead of maintaining infrastructure.
Why it matters
Veeza AI's problem wasn't really support volume. Its most valuable channel had become its biggest constraint. Every conversation was a potential sale and a chance to build trust, and Veeza AI had to turn them away.
Keiki changed what one founder could run:
- Support became a sales channel. The agent answers prospects in seconds, any time of day, and moves them to a paid application in the same thread.
- Aly's time went back to the business. He handles the conversations that need judgment, not the ones that need a lookup.
- Growth could switch back on. With the agent absorbing the volume, the ads that had been paused could run again.
The impact
Keiki took over the conversations that had forced Veeza AI to slow down, at a volume no single founder could answer. Since moving to Keiki, Veeza AI has doubled its revenue and tripled its usage. And because Keiki runs the messaging infrastructure, Aly no longer maintains it himself: that's more than 20 hours a week back, spent on the product and the customers instead.
What's next
Next for the agent are proactive application updates and the same experience on more messaging channels.
"Messaging in the Global South is the internet. Services will happen naturally, now that personal assistants are about to work, on WhatsApp or any messaging platform."
Aly Moursy, founder of Veeza AI
Put your product on WhatsApp
If your customers already message you, your agent should live there too. Connect your API or MCP server, add a WhatsApp line, and let customers reach your product where they already are.
Start building on Keiki or read how to turn your API into a WhatsApp agent.