In 2016, Atyab Bhatti had a product idea with the slight disadvantage of requiring technology that did not quite exist. He and friends traded emails about an artificial-intelligence travel agent. The proposition was tidy: ask for a trip in ordinary language and let software handle the dreary choreography. The machinery was not tidy. Traditional machine learning was slow and compute-hungry, and it could not produce the quality the idea demanded. The sensible thing happened. The idea waited.
A decade later, the same premise sat inside Slack and Microsoft Teams, making policy-aware reservations in under two minutes. SkyLink, the company Bhatti cofounded with Kush Maheshwari, had won an industry innovation competition, landed enterprise users including McKinsey & Company and Sony, and been acquired by Amadeus. Plenty changed between the email thread and the deal. The stubborn bit was the original irritation: booking business travel required people to act as the connective tissue between too many systems.
Before travel, retail
Bhatti's founder story began before the travel thesis. At the University of Illinois Urbana-Champaign, where he studied mechanical engineering, he met Maheshwari during their freshman year. By 2015 they were working together on SEAK, a mobile marketplace meant to place neighborhood shops, large retailers and independent sellers in one local feed. John Yoon, an economics student, joined as designer. The project had already signed retailers by the summer of 2016. Shops began approaching the team rather than the other way around.
The old project is useful not because it predicted SkyLink in miniature, but because it shows the partnership already in motion. Bhatti and Maheshwari were trying to compress a scattered market into a more approachable interface. SkyLink would eventually perform a cousin of that maneuver for fares, hotels, policy and traveler preference. Mechanical engineering may sound like an eccentric prelude to travel software. In practice, it supplied an apt habit: see the workflow as a system, then look for the friction between its parts.
Bhatti earned both bachelor's and master's degrees in mechanical engineering. His later roles widened the aperture: product management at Malwarebytes, private-equity investing at Stripes, and consulting at McKinsey. Corporate travel arrived as both lived nuisance and business education. A consultant searches a consumer site, messages colleagues, checks a company rule, opens a booking tool and perhaps calls an agency. The trip is straightforward. The coordination resembles amateur air-traffic control.
“You didn't have large language models. Traditional machine learning was very compute-intensive and very slow.”Atyab Bhatti, recalling the 2016 idea
The interface was already open
When large language models improved, Bhatti and Maheshwari returned to the old proposition with a sharper choice. They would not simply build a nicer field of drop-down menus. SkyLink would sit where coworkers were already discussing trips. A traveler could write or say what they needed in Slack or Teams. The system could identify the traveler, consult history and company policy, search connected inventory, recommend options and complete a booking. If a meeting moved, the same conversation could help service the reservation.
The location of the product was part of the product. A new tab demands a new habit; a message behaves like a familiar one. Bhatti described SkyLink as an enterprise-grade system that users could talk to like a person. Behind the casual surface sat the unglamorous requirements of corporate software: travel profiles, preferred suppliers, global distribution systems, direct connections, approval logic, information security and market-specific choices.
Builder's note: adoption gets cheaper when the new behavior borrows the shape of an old habit.
The system also had to serve three parties without making any of them feel like luggage: the employee who wants a tolerable itinerary, the employer that wants control and savings, and the travel-management company responsible for fulfillment and service. SkyLink did not pitch the disappearance of human agents. Bhatti argued that automation could remove repetitive work and leave people with the complicated situations that deserve judgment.
Returning with a vendor badge
McKinsey became the revealing test. Bhatti began knocking on his former employer's door while SkyLink was still young. Being an alumnus was not a magic key. He later laughed that it did not help much at all. The break came through an introduction to Jamie Stewart, McKinsey's director of travel and event technology, who saw what conversational interaction could do when joined to policy, traveler history and real inventory.
The collaboration was unusually granular. SkyLink and McKinsey worked through the fundamentals: how many options to show, which carriers to prioritize, how an experience should vary by market, and how to fit the firm's information-security and AI-governance standards. In Britain, Eurostar choices could appear beside flights. A timeout warning could rescue a fare before it expired. Recurring trips could be booked in bulk. Support remained inside chat, preserving the flow that the product was designed to protect.
Time returned to the traveler
The numbers made the case legible. The average booking took 99 seconds rather than nearly 15 minutes in traditional channels. Ninety-six percent of users gave the Slack-based workflow a positive rating. Eighty-five percent of people who tried it became repeat users, and the same share selected one of the first three options shown. Bhatti also reported stronger hotel attachment and more international round trips rather than costly pairs of one-way tickets. A conversational interface was influencing behavior without presenting itself as a lecture from procurement.
Proof leaves the chat
In October 2024, SkyLink beat 12 other competitors to win the BTN Group's Innovation Faceoff. One judge called it “true conversational commerce.” The award marked Bhatti as the group's Innovator of the Year and gave a specialist product a larger stage. By 2025, Sony and McKinsey were named clients, BCD Travel was offering SkyLink to customers, and Business Travel News placed Bhatti among its 25 Most Influential people in the industry. PhocusWire included SkyLink in its Hot 25 Travel Startups for 2026.
Recognition did not make enterprise sales instant, but it altered the burden of explanation. Bhatti observed that early customers had needed help imagining what was possible. Later customers could see a working precedent. “Going first is hard for clients,” he said, “but going second - it's definitely not as hard.” Enterprise innovation often advances in this slightly comic procession: one brave buyer, followed by a queue of sensible followers who had always been interested.
“We assume we can figure it out. That's served us really well. And we see when it works how quickly it compounds.”Atyab Bhatti on his team's first-principles approach
His public writing reveals the engineer beneath the salesman. Bhatti has explained parameters, hyperparameters, attention windows and hallucinations through travel examples. An ambiguous request for Georgia could mean the American state or the country; a model should not purchase a ticket on the strength of a hunch. The point is less to make AI sound magical than to make its limits operational. In travel, a charming fabrication still misses the flight.
The larger itinerary
On February 25, 2026, Amadeus announced that it had acquired SkyLink. The travel technology group described the company as AI-first, with a proprietary architecture and a multilayer orchestration engine. Tens of thousands of bookings had already moved through the system. The immediate plan was to strengthen conversational capability in corporate travel, particularly in North America. The longer route extended across airlines, airports, hotels and travel sellers.
For Bhatti, the combination joined SkyLink's native AI system to an organization operating in more than 190 markets and processing billions of searches. His stated ambition after the deal was practical: deploy faster and bring new capability to travelers and companies across the industry. The vocabulary matters. After a decade in which AI travel assistance moved from impossible concept to abundant demo, the test had become deployment - reliability, scale and concrete work.
The satisfying shape of Bhatti's story is not that he foresaw everything in 2016. He did not. The first idea was consumer-facing; the eventual opportunity was corporate. Careers at a security company, an investment firm and a consultancy added context the college sketch did not have. A patient idea is not a preserved insect. It changes while the founder learns what the real constraints are.
SkyLink's defining product move was never pretending that travel had become simple. Airlines, hotels, policies, distribution systems and disrupted plans remain gloriously complicated. The product simply relocated some of that complexity behind a sentence a traveler could type. It let a familiar conversation carry unfamiliar machinery, without requiring the traveler to admire the gears. Bhatti waited for the machine to understand the request. Then he put it in the one window everyone already had open.