Professional Development of an AI-Driven Chatbot for Customer Service in Nigerian Commercial Banks

Nigerian commercial banks handle millions of routine customer queries daily through call centres and branch visits, creating operational bottlenecks and high service costs. AI-driven chatbots offer a scalable digital channel for routine customer service, but existing deployments at Nigerian banks lack natural language understanding for Nigerian English, Pidgin, and multilingual customer communication patterns. This study developed an AI-driven customer service chatbot framework specifically designed for Nigerian commercial bank customer service contexts. A conversational AI design methodology using Rasa Open Source framework was applied. A domain-specific intent taxonomy was developed covering 42 intent categories drawn from analysis of 8,000 customer service interaction transcripts sourced from a Lagos-based tier-2 bank. Training data was augmented with Nigerian English, Pidgin, and code-switched utterance variants for each intent, generating a total of 24,000 training examples. An LSTM-based NLU model was trained achieving 91.3 percent intent classification accuracy on a 20 percent held-out test set. A named entity recognition component handled account numbers, transaction references, and branch names with 88.7 percent F1-score. A multi-turn dialogue management system maintained context across up to eight conversation turns without user re-prompting. The chatbot was integrated with a mock core banking API for balance inquiry and transaction history retrieval. Usability testing with 30 bank customers showed 84 percent first-contact resolution for routine queries. The study recommends live deployment at a single branch as a hybrid human-AI escalation pilot and collection of 100,000 real interaction examples for model retraining.

Keywords: AI chatbot, customer service, Nigerian banking, Rasa NLU, conversational AI