Students Develop AI-Run Tool For Airlines

The tool, called Onboardly, was built by Sneha Khowala along with help from Jash Lodhavia as part of the “AI in Business” course led by Dr Daniel Ringel, assistant professor of Marketing for Data Science and AI at UNC Kenan-Flagler and Visiting Faculty at BITSoM.

Sneha said the idea emerged from her experience working in airline operations, where basic information needed for day-to-day functioning was often difficult for new hires to access quickly.

“Airline onboarding is highly fragmented and time intensive. New employees struggle to find the right information at the right time, while managers and human resources teams repeatedly answer the same operational and policy related questions. This affects productivity on both sides,” she told Deccan Chronicle.

She said that even routine coordination could become inefficient. “I often had to make multiple phone calls just to figure out who was on shift or who to contact for last minute operational changes like cargo loading,” she said. “When this is happening across teams and locations, it leads to delays, confusion, and unnecessary operational friction.”

In just two weeks, the team consolidated Air India’s scattered onboarding resources, including standard operating procedures, safety manuals, human resources policies, and internal documentation, into a single conversational system.

Onboardly functions as a digital onboarding guide that employees can use from their first day, allowing new hires to ask role specific questions such as where to find policies, which procedures apply to their role, or whom to contact for specific issues. Instead of navigating large manuals or waiting for responses from supervisors, users receive structured answers intended to reduce confusion during the early stages of employment.

“The idea is to help new employees focus on understanding their role rather than spending time figuring out systems,” Sneha said. “At the same time, it reduces the load on managers, who are otherwise pulled away from core responsibilities.”

She explained that the tool uses retrieval augmented generation, role-based access controls, and a simulated enterprise knowledge base, with features such as login-based access, chat history, workflow explanations, and automated safety alerts.

According to the team, this has reduced repetitive queries to human resources teams and managers by 35 to 60 per cent, accelerated new hire productivity by nearly three weeks, and improved operational readiness by standardising information access across teams.

When asked about the safety-sensitive nature of aviation, Sneha said compliance was a central design consideration. The current academic version of the tool uses fabricated airline data, but the system follows a retrieval-based architecture. “It only responds using information that has been explicitly provided to it. In a real-world setting, approved company manuals and policies would be the only data source,” she said.

The tool also uses role-based access controls to limit what information a user can see. “If someone asks a sensitive or restricted question, the system does not provide an answer,” Sneha said. “Instead, the query is flagged so that it can be addressed through proper channels.”

She said designing these boundaries was the most challenging part of the project. “In aviation, speed matters, but accuracy and compliance matter more,” she said, adding that they had to be very clear about what the AI should answer, what it should restrict, and when it should escalate.

Sneha said the project showed how AI could be used to reduce operational load without replacing human judgement. By converting complex organisational documents into guided conversations, she said such tools could support safer and more efficient onboarding in large, regulated organisations where delays directly affect operations.