By Dave DeFusco

Finding the right answer inside a university can sometimes feel harder than taking a final exam. Policies live in long handbooks. Procedures are buried in PDFs. Students and staff often don’t know who to ask, and even experienced administrators can spend hours searching for one small detail.

That everyday problem inspired Katz Assistant, a new project developed by Ashish Yakub Beary, a student in Katz College’s M.S. in Data Analytics and Visualization. The project uses artificial intelligence in a practical, human-centered way: it lets people ask questions in plain English and get clear answers drawn directly from official school documents.

“At its heart, Katz Assistant is about saving people time and frustration,” said Beary. “Instead of reading through dozens of pages, you can just ask a question and get the information you need right away.”

Most organizations store knowledge in static files—Word documents, PDFs and guides that don’t change often and aren’t easy to search. Katz Assistant takes those documents and turns them into something interactive.

"The system reads the documents once, learns what they mean and then remembers how different ideas are connected,” he said. “After that, it can talk with users the way a helpful colleague would.”

Behind the scenes, the assistant builds a structured map of information that understands not just words, but relationships. For example, it learns that “email guidelines,” “communication rules” and “messaging policies” are related ideas, even if they use different wording. That means users don’t need to guess the exact phrase used in a handbook to get a useful answer.

Katz Assistant was designed specifically for academic operations. Program managers can ask about policies. Graduate assistants can get guidance on procedures. Students can finally ask questions without worrying about who the right contact person might be.

“The goal was to meet people where they are,” said Beary. “Everyone already knows how to ask questions, so I built a system that understands questions instead of forcing users to adapt to the technology.”

The assistant even has a name for its intended user—“Charlie”—to reinforce the idea that this is a supportive tool, not a cold piece of software. One of the most important features of Katz Assistant is that it knows its limits. If the system doesn’t find strong enough information to answer a question, it doesn’t guess. Instead, it clearly says so and directs the user to a person who can help.

“That was a design choice I felt strongly about,” said Beary. “Trust matters. If the system isn’t confident, it should say that and point you to a human expert.”

Each answer also comes with a confidence score, giving users transparency about the reliability of the response. That extra layer helps people decide when an answer is good enough, and when to double-check.

From the user’s perspective, Katz Assistant looks simple: a clean chat screen where you type a question and read the answer. Underneath that simplicity, however, is a carefully designed system that balances speed, accuracy and privacy.

The assistant runs on open-source technology and works locally, which means sensitive internal documents don’t have to be sent outside the institution. It also remembers past questions, so repeated queries can be answered almost instantly.

For administrators, there’s a dashboard that shows how the system is being used, what questions come up most often and where information gaps might exist. Over time, this data can help improve both the assistant and the underlying documentation.

David Li, director of the M.S. in Data Analytics and Visualization, sees Katz Assistant as a strong example of how graduate-level education should connect theory to practice.

“This project shows how data analytics and AI can be applied responsibly to solve real operational problems,” he said. “Ashish didn’t build technology for its own sake; he built something that directly improves how people work and learn.”

Li also emphasized the broader lesson. “Katz Assistant demonstrates that advanced tools can be made accessible, ethical and useful at the same time,” he said. “That’s exactly what we want our students to be able to do.”

While Katz Assistant is already functional, Beary sees it as a foundation rather than a finished product. The system can be expanded with more documents, improved understanding and even better answers over time.

“For me, this project is about possibility,” said Beary. “If we can make knowledge easier to access in one school, we can do it anywhere.”