Pembuatan Fitur Chatbot pada Website E-Commerce PT Ratu Bio Indonesia

Abstract

PT Ratu Bio Indonesia masih melayani konsultasi calon klien secara manual melalui admin, yang menyebabkan keterbatasan kapasitas layanan dan keterlambatan respons di luar jam kerja. Pendekatan pengembangan yang dipilih adalah metode agile agar sistem dapat menyesuaikan perubahan kebutuhan secara bertahap. Hasil penelitian meliputi pembuatan fitur chatbot berbasis aturan dengan algoritma pencocokan berbasis skor yang dilengkapi normalisasi teks dan kamus sinonim, perancangan use case diagram, relasi antar tabel, dan activity diagram, serta pengembangan menggunakan Laravel, React.js, dan Inertia.js. Sistem menyediakan eskalasi ke live chat dengan admin, dashboard pengelolaan aturan tanpa mengubah kode program, serta pencatatan pertanyaan yang belum terjawab. Pengujian black box terhadap sepuluh skenario menunjukkan seluruh fungsi berjalan sesuai rancangan, sedangkan pengujian akurasi terhadap 30 pertanyaan uji memperoleh akurasi sebesar 86,67%. Penelitian ini menitikberatkan pada percepatan dan konsistensi layanan konsultasi maklon. Disarankan penerapan pemrosesan bahasa alami agar chatbot mampu mengenali maksud pertanyaan di luar cakupan aturan.

PT Ratu Bio Indonesia still handles prospective client consultations manually through admins, which leads to limited service capacity and delayed responses outside working hours. The development approach chosen is the agile method so that the system can adapt to changing needs incrementally. The research results include the creation of a rule-based chatbot feature with a score-based matching algorithm equipped with text normalization and a synonym dictionary, the design of a use case diagram, table relations, and activity diagrams, as well as development using Laravel, React.js, and Inertia.js. The system provides escalation to live chat with an admin, a dashboard for managing rules without modifying the program code, and the recording of unanswered questions. Black box testing of ten scenarios showed that all functions ran as designed, while accuracy testing of 30 test questions obtained an accuracy of 86.67%. This research focuses on accelerating and maintaining the consistency of contract manufacturing consultation services. The application of natural language processing is recommended so that the chatbot can recognize the intent of questions beyond the scope of the rules.