A multi‑agent AI system designed to segment customers and generate targeted marketing strategies using e‑commerce behavior data.
Built as part of Google DevFest 2025 London, this project supported a 90‑minute interactive workshop I led: Agentic AI: Hands‑On with Gemini in Kaggle.
Participants learned how to build safe, scalable agent workflows using Gemini and Google Cloud tooling.
Workshop Objective
Demonstrate how agentic AI can automate customer analysis and translate insights into actionable marketing plans—showcasing practical, enterprise‑ready use cases for multi‑agent systems.
Business Problem
A UK-based online retailer is facing challenges in understanding its diverse customer base and optimizing its marketing efforts. With thousands of transactions across multiple countries, the company struggles to:
Identify meaningful customer segments based on purchasing behavior
Tailor marketing campaigns to different customer profiles
Avoid generic promotions that lead to low engagement and wasted budget
Ensure recommendations are ethical, inclusive, and aligned with customer preferences
Business Opportunity
The rise of agentic AI presents a transformative opportunity for online retailers to move beyond static analytics and embrace intelligent, goal-driven systems. By leveraging transactional data and Gemini-powered agents, the company can:
Unlock deeper customer insights through dynamic segmentation based on behavior and purchasing patterns
Automate personalized marketing strategies that adapt to each customer segment, increasing engagement and conversion
Scale decision-making with AI agents that reason, plan, and act — reducing manual effort and accelerating campaign deployment
Build trust and brand loyalty by embedding ethical guardrails that ensure fairness, transparency, and responsible recommendations
Empower cross-functional teams with a reusable, interpretable agent framework that supports experimentation and continuous improvement
Solution Objective:
Build a Customer Segmentation & Marketing Planner Agent that:
Segments customers based on behavior
Recommends tailored marketing strategies
Ensures ethical and responsible outputs
Dataset
For this project the 'Online Retail' dataset from the University of California, Irvine – Machine Learning Repository was selected. This dataset contains transactional data set which contains all the transactions occurring between 01/12/2010 and 09/12/2011 for a UK-based and registered non-store online retail.