Implementasi AI Agent Berbasis RAG untuk Klasifikasi Pelanggan KBLI 2025

Teddy Agustinus (Universitas Tarumanagara)
Jap Tji Beng (Universitas Tarumanagara)
Rahmiyana Nurkholiza (Universitas Bina Nusantara)
Vienchenzia Oeyta Dwitama Dinatha (INTI International University)

Jutisi: Jurnal Ilmiah Teknik Informatika dan Sistem Informasi
Vol 15, No 3 (2026): Juni 2026

The migration of the ERP system from SAP ECC to SAP S/4HANA requires customer master data to have a complete industry classification that complies with the 2025 Indonesian Standard Industrial Classification (KBLI) to support the implementation of the Business Partner model. This research aims to develop an AI Agent based on Retrieval-Augmented Generation (RAG) to automate the customer classification process within the Master Data Management unit at PT X. The research employs an implementational approach by building a RAG system that integrates the KBLI 2025 knowledge base, hybrid search, and a large language model to classify 235 active customers into five-digit KBLI codes. The system’s performance was evaluated based on the accuracy of the classification results. The research findings indicate that the system achieved an accuracy of 98 per cent and successfully mapped 50 customer line-of-business categories into 21 KBLI Level 1 sectors. These findings demonstrate that the AI Agent is capable of generating standardised customer master data and is ready to support the migration to SAP S/4HANA.Keywords: AI Agent; Retrieval-Augmented Generation; Hybrid Search; Customer Classification; KBLI 2025AbstrakMigrasi sistem ERP dari SAP ECC ke SAP S/4HANA menuntut data master pelanggan memiliki klasifikasi industri yang lengkap dan sesuai dengan standar Klasifikasi Baku Lapangan Usaha Indonesia (KBLI) 2025 agar mendukung implementasi model Business Partner. Penelitian ini bertujuan mengembangkan AI Agent berbasis Retrieval-Augmented Generation (RAG) untuk mengotomasi proses klasifikasi pelanggan di unit Master Data Management PT X. Penelitian menggunakan pendekatan implementatif dengan membangun sistem RAG yang mengintegrasikan basis pengetahuan KBLI 2025, pencarian hibrida, dan large language model untuk mengklasifikasikan 235 pelanggan aktif ke kode KBLI level lima digit. Kinerja sistem dievaluasi berdasarkan tingkat akurasi hasil klasifikasi. Hasil penelitian menunjukkan bahwa sistem mencapai akurasi sebesar 98% serta berhasil memetakan 50 kategori line of business pelanggan ke dalam 21 sektor KBLI Level 1. Temuan ini menunjukkan bahwa AI Agent mampu menghasilkan data master pelanggan yang terstandarisasi dan siap mendukung migrasi ke SAP S/4HANA.