The EMBA and MBA programmes at Fudan University School of Management offer two routes to the degree: the traditional thesis, or a practice pathway built on real management work. The practice pathway puts its weight on real management practice, the student’s own contribution, and verifiable results — but for every new student, the first question is a plain one: is this the right path for me? The 2026 cohort has more than 200 students, and that question is now being handed to an AI agent for a first-round answer.
The practice pathway asks each student to carry a real business project through their studies: a clear scenario, key actions of their own, data as evidence, and a company willing to grant authorization. Judging whether a student meets these conditions used to mean academic affairs teachers and class advisers reading through files and scheduling interviews, one student at a time — across two hundred people in a cohort, the cycle ran long and the bar was hard to keep consistent, because teachers did not always measure “qualified” the same way.
The first step of the project was to turn the school’s years of accumulated judgment into structured rules. The student profile defines five core dimensions — a real business scenario, control of the project, the student’s own key actions, results and data evidence, and feasibility within the study period — each with a score threshold. It also draws a red line: students whose data cannot be obtained, whose company authorization is unclear, or whose project cannot move forward during their studies do not enter the strong recommendation tier. Whether the student is a chairman, whether the company is a tech startup, how long the answers run — the factors most likely to mislead a judgment are explicitly excluded from the criteria.
The agent reads each student’s career history, company information and short-answer questions from the application, and returns one of four results: strong recommendation, strong recommendation candidate, interview confirmation, or on hold. The “strong recommendation candidate” tier is a deliberately designed buffer — for students with no obvious risk who still need one interview to confirm. Every result comes with the reasons for or against, recommended topic directions, and the gaps still open in the materials; teachers’ reviews of the labels and edits to the reasons flow back into the system, closing a data loop of identify, recommend, interact, follow up, and refine.
The diagnosis is not just for students. The project dashboard gives each of three roles its own view: programme directors see the overall distribution of potential, the reach-and-conversion funnel and progress by class; academic affairs teachers see student lists, potential labels and follow-up status; class advisers see their own students’ willingness to engage and the next communication step. Topic recommendations follow one format too — title, why it fits, the core practice question, materials that can support it, and items to confirm — helping students translate a real business problem into a practice topic within the scope of business administration.
We used to pick a few students by impression and focus our conversations on them. Now AI sorts two hundred people into four tiers with reasons attached, and we spend our time on the interview confirmations that genuinely need human judgment.
Chen, academic affairs teacher
The plan states its boundaries: the diagnosis serves as a pre-screen for the graduation format and as interview preparation, not as a substitute for the formal topic review; scores are a rule-based pre-assessment, and students with high scores or a “strong recommendation candidate” label still go through human review; when company-sensitive information is involved, de-identified data comes first, and students confirm how their materials may be used. The AI finishes the first-round screening for two hundred people and explains its reasons — the final choice stays with the students and their teachers.
Today, the practice-pathway diagnostic and topic-recommendation agent has completed simulation testing and iterations of its training data, and is about to go live for the 200-plus students of the 2026 cohort. For Fudan’s School of Management, what it solves is not “replacing teachers with AI” — it is making sure that every student, soon after enrollment, gets a first-round answer with reasons, topic directions and a materials list: am I the right fit for the practice pathway?
