7 • a robust analytics infrastructure and MLOps (machine learning operations), with dedicated feature stores to rapidly deploy and scale ML models • new data pipelines and integrators, prompt stores, and vector databases to build or customize large language model implementations Along with granular transaction data, these additional data assets are the backbone of AI-powered decisioning to predict customer behavior in any channel and can be a true differentiator in today’s market. Decisioning To develop new targeted promotions and content through more robust targeting, companies can also benefit from refreshing their decision engines with new AI models. Their tasks include the following: • Promo propensity predicts the likelihood of a customer making a purchase due to a promotion, based on previous customer purchasing and engagement behavior. This can lead to better customer satisfaction and targeting with the right level of discount to improve margins. How a European telecom used gen AI to enhance marketing materials European telecom company recently boosted its marketing strategy by integrating a personalization engine that uses both AI and gen AI. The telco previously relied on mass promotions and a calendar-based approach. To better engage its customers, it set up a next-best-action engine, using a combination of multiple machine learning models to determine the most effective actions to suggest to each customer. The next-best-action mechanism predicted the probability of a customer accepting a specific action and the expected value if the offer was accepted. It then ranked these actions, optimizing for the highest expected value from a response. The company then tested a granular set of about 2,000 different actions by texting messages to customers. The marketing team then deployed gen AI– enhanced messaging for a handful of different campaigns to see how a more personalized approach would work. It sent messages to customers based on factors such as age, gender, and data usage. The copy blended general company messaging with campaign-specific features, making the offers feel more personal and relevant to customers. Guardrails were put in place to limit the length, tone, and content, ensuring that messages were concise, relevant, and respectful of user privacy. Over the course of a few months, the experiment showed that customers receiving the personalized messages from the gen AI–enhanced campaigns engaged and took action 10 percent more often than customers who did not receive personalized content. The telecom is now looking at more extensive use of personalized content across multiple marketing channels.