Hailiang Kefu launches eShengya, an OpenSpan-powered AI career planning platform

Hailiang Kefu has launched eShengya, a service platform built on its OpenSpan career-planning model that brings an AI career planner into family decisions. A QbitAI report on November 5 describes a system that first learns about a student's academic performance and interests, then confirms the family's goals before combining education knowledge and regional policy into advice on subject selection, extracurricular development and higher-education planning.

Unlike a general chatbot, eShengya splits the consultation into roles and steps. It extracts facts from the conversation to build a student profile, identifies the intent behind a question, routes it to a specialist agent, and then invokes predefined Skills that structure follow-up questions, retrieval and reasoning. The report shows roles including a talent-development specialist, an admissions adviser and a study-abroad consultant.

The answers are organised around an "independent decision map", a visual representation of the OpenSpan knowledge graph. Paths start from learning foundations and talent development, run through school stages, connect to admissions routes and extend to university majors and careers. Students with a target can work backwards to the subjects it requires; those without one can explore from subjects they are good at and interests they enjoy.

The report contrasts the platform with general AI using two examples. Asked what a fifth-grader's interest in science might mean, eShengya keeps probing family budget and sports preferences before suggesting a hands-on direction such as drones or robotics alongside a low-cost sport within a 10,000-yuan budget. Asked how a student scoring 490 could get into university, it checks admissions routes and eligibility requirements before discussing score improvement rather than jumping straight to a study plan.

The team also built its own evaluation rubric for career-planning conversations, scoring profile completeness, understanding of needs, accuracy of domain knowledge, quality of planning reasoning and service communication. In one internal test of version r43 covering 89 dialogues, with a score of 2 out of 3 treated as usable, the overall usability rate was 92.1%.

Keeping data current is central to this kind of product. The team says the system connects to public data from exam authorities and university admissions offices, with update mechanisms and expert review, so knowledge and rules can change when policy does.

Commercially, basic planning on eShengya is free, aimed at families — including those in county-level areas — who struggle to obtain sustained professional guidance. Hailiang Kefu's approach is to productise the methodology built up through human consulting and reach more households at a lower marginal cost.

The value of a vertical AI product like this does not lie in how fluently it chats, but in whether its advice is grounded, fits a specific family, and can be revised as circumstances change. The next questions are whether a free tier can cover the cost of maintaining knowledge and policy data, and whether the internal usability rate holds up in real, long-term use.