An Explainable End-to-End AI Framework for Marketing Automation and Lead Generation in Small and Medium Enterprises

Abstract

Small and medium enterprises (SMEs) still face weak lead-generation outcomes, with only about 20–30% of generated leads eventually becoming customers. Because of limited resources and technical capacity, many SMEs cannot use the same AI solutions adopted by larger firms. This paper presents and evaluates an integrated nine-phase AI-driven marketing automation framework designed to support lead scoring, qualification and business decision-making in SME settings. The framework includes data ingestion and exploratory analysis, preprocessing of imbalanced data using the Synthetic Minority Over-sampling Technique (SMOTE), a unified preprocessing workflow for six classifiers: Logistic Regression, Random Forest, XGBoost, LightGBM, Support Vector Machine and Multi-Layer Perceptron, SHAP-based explainability, NLP-based sentiment and intent extraction, a real-time conversational lead qualifier, and a Return on Investment (ROI) dashboard. The framework was tested on Kaggle's Leads (EdTech) dataset and the UCI Bank Marketing dataset to evaluate its performance across inbound digital marketing and outbound telemarketing contexts. On the EdTech Leads dataset, XGBoost achieved an Accuracy of 92.9% and a ROC-AUC of 97.2%. On the imbalanced Bank Marketing dataset, LightGBM achieved an Accuracy of 90.5% and a ROC-AUC of 92.7%, while also training faster than the competing models. SHAP analysis indicated that Tags, Lead Profile, Engagement_Score and Avg_Dur_Contact were among the most influential predictors. Overall, the results suggest that combining predictive modelling, explainability and conversational qualification can improve lead prioritization for SMEs and provide managers with clearer evidence for allocating marketing effort, estimating expected returns and reducing dependence on manual screening in resource-constrained sales and marketing environments with greater consistency.