Casino marketing has changed little in fifty years. At the high end, marketing is high-touch but anecdotal and too often uneven. At the mid and low end, well-worn offers reward behaviors that were going to happen anyway. All told, most casino marketing programs are 30%+ inefficient. At some large properties, that's a $100M prize.

How do we make casino marketing more strategic while capturing the on-property experiences our players love, all within guardrails? Large language models are great at reasoning through problems at scale, but to solve those problems, the LLM needs to understand player preferences and propensities.

This is why integrating traditional predictive modeling with LLMs is the key to unlocking hyper-personalized offers.

The third leg is training the LLM to think like a host. We do this with a process called fine-tuning, and the result is an LLM that can break a problem down into sequential building blocks. Below is an example of the multi-step reasoning an LLM uses to craft a personalized offer; we are deploying this LLM at some of Asia's largest integrated resorts.