Herzen University Develops an AI Assistant for First-Year Students
Herzen State Pedagogical University of Russia has unveiled a prototype AI assistant designed to help first-year students navigate official university language and understand what it means in everyday situations.
For students just starting university, the biggest challenge is often not academic work but figuring out administrative procedures. Herzen State Pedagogical University of Russia aims to address this problem with a prototype conversational assistant. Its goal is straightforward: bring together information scattered across more than 60 pages of administrative documents and explain it to newcomers in plain language.
The University Learns to “Listen”
The Pomoshchnik pervokursnika (First-Year Student Assistant) has a relatively simple design. It draws on organized information about university rules, schedules, faculty members and course materials. A separate skill module for the language model defines how it should interact with users: how to conduct a conversation, find relevant information and formulate answers. But the most important feature is its clearly defined limits. The assistant is not allowed to make up dates, classroom numbers or payment amounts. If the documents do not contain the information a student needs, it must explain whom to contact rather than offer a plausible-sounding but unverified answer. Language models often state incorrect information with confidence, so the developers deliberately built caution into the assistant's operation.
The team initially tested the assistant in four use cases. The prototype passed all 19 automated tests, while its average response time fell from 59 to 36 seconds. The developers caution, however, that these results should not yet be considered conclusive. The system still needs to be tested in additional scenarios and put through another round of evaluation.
The Race to Make AI Answers Verifiable
Herzen University's initiative is part of a broader wave of projects aimed at supporting students at Russian universities. Over the past three years, higher education institutions have stepped up their use of digital tools to attract incoming students and help them navigate university life. In late 2024, Sberbank introduced a Telegram bot powered by GigaChatMax for applicants, students and faculty. It can answer questions about coursework, scholarships, exams and administrative procedures. Among the first universities to join the project were St. Petersburg State University, the Russian Presidential Academy of National Economy and Public Administration, Peter the Great St. Petersburg Polytechnic University and Immanuel Kant Baltic Federal University.
In spring 2025, the St Petersburg Federal Research Center of the Russian Academy of Sciences (SPC RAS), together with Sberbank, expanded the capabilities of its assistant. The developers said users could search for information across the websites of 18 universities in 10 Russian regions. The project demonstrated that a single platform could serve multiple universities.
In fall 2026, Saratov State University began using an assistant for prospective students in a live operating environment. The tool searches official documents by meaning and provides links to the source materials alongside its answers. It can handle up to 80% of routine questions on its own, with responses taking between 3 and 15 seconds.
What sets Herzen University's project apart is its emphasis on verifying answers and assessing their quality. It tackles a challenging question: it is not enough to establish whether a system can answer a question. It is also important to determine whether the answer is supported by the original source.
Keeping Information Current and Defining the Limits
The AI assistant prototype has been successfully tested in four scenarios, but it will eventually need to handle thousands of questions during the peak admissions period. Schedules and payment amounts also change over time, so the information in its knowledge base will need regular updates. If the system is connected to student accounts and personal data, strict access rules will be essential.
In spring 2026, the university adopted rules governing the use of AI in teaching and research. They call for responsible use of the technology, protection of personal data and continued human oversight. As a result, administrative actions with legal consequences, such as submitting applications or changing entries in official records, are unlikely to be handed over to a conversational interface in the near future.
Turning Bureaucratic Language into Plain English
Importantly, universities are beginning to treat access to administrative information as an engineering challenge.
“At the university, administrative information is spread across several documents and regulations that together run to more than 60 pages and are primarily intended for staff in university departments. The agent allows students to interact with these materials in natural language and receive concise instructions: what they need to do, by when, whom to contact and which provision of the regulations supports the answer,” explains Mikhail Streltsov, an associate professor in the Department of Geology and Geoecology at Herzen State Pedagogical University of Russia.
The assistant does more than reduce the number of phone calls students need to make. The prototype saves them from having to learn the language of university regulations during their very first week on campus. The system is not yet fully mature, but its direction is clear: a reliable university AI assistant must be able to admit when it does not know the answer and point students to someone who can help.
