A 12-week trial with 180 Indonesian vocational students (90 intervention, 90 control) found an AI system combining skills-mapping, chatbot mentoring and labour-market data cut career-path anxiety by 26.7% (p<0.001), with an 87%-accurate career-pathway matching model.
26.7 %
Reduction in career-path anxiety (12 weeks)
73.1
Career-path anxiety score, pre-intervention
53.6
Career-path anxiety score, post-intervention
87 %
Skills-mapping model accuracy
79 %
Students engaged in AI-mediated mentoring
2,400+
Chatbot interactions recorded
4.60 /5.0
Mentoring satisfaction score
500
Job postings analysed
Details
Maturity
Pilot
Promoter
Universitas Negeri Yogyakarta (Graduate School)
Period
12-week intervention, 2025; published February 2026
Keywords
career guidance, vocational education, machine-learning skills mapping, AI mentoring chatbot, labour-market intelligence
Context
An AI-driven career-guidance system combining skills-mapping machine learning, an adaptive AI chatbot mentor, and labour-market NLP analysis was piloted with 320 vocational students across three schools in Banjarmasin, South Kalimantan, Indonesia.
Objectives
The system aimed to reduce students' career-path anxiety and improve career and skills guidance through personalised AI mentoring combined with real labour-market data.
Activities
Over a 12-week period, 90 students used the AI system while 90 matched control-group students who did not use it were tracked for comparison; the system analysed 500 job postings and supported chatbot-mediated mentoring alongside repeatable skills-gap analysis.
Results
Career-path anxiety fell 26.7% in the intervention group (from a mean of 73.1 to 53.6, p<0.001). The skills-mapping model reached 87% accuracy (precision 0.85, recall 0.83, F1 0.84, ROC-AUC 0.89). 65% of students repeated the skills-gap analysis three or more times, and 79% engaged in AI-mediated mentoring across more than 2,400 chatbot interactions, with mentoring satisfaction rated 4.60 out of 5.0.
Conclusions
Because the comparison group was matched rather than randomised and the pilot ran for only 12 weeks across three schools, the results are promising but not causally conclusive; replication with randomisation and longer follow-up would strengthen the evidence base.
Implementation
Indicative cost
Low (< €50k) — Not disclosed in the source; as a small-scale academic pilot it is likely low-cost.
Time to results
Short (< 1 year) — 12-week intervention conducted in 2025; a preprint appeared on preprints.org in October 2025, with peer-reviewed publication in F1000Research in February 2026.
Staffing & skills
School counsellors and teachers at the three participating vocational schools who facilitated delivery, Research team affiliated with Universitas Negeri Yogyakarta Graduate School (study authors)
Conditions for success
Reliable student access to devices and internet for AI chatbot mentoring
Availability of local labour-market job-posting data to feed the NLP model
Careful matching of control-group students given the non-randomised design
Common failure modes
Small sample (90 vs 90) and non-randomised matching limit causal certainty
Short 12-week duration may not capture longer-term career outcomes
Where it fits
Governance type
academic pilot within vocational schools
Scale
3 schools / 320 students (180 in the intervention-control comparison)
Income level
lower-middle-income (Indonesia)
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.