Generative AI and Language Models

The 20–30% EBITDA upside from casino AI is real—most properties just can't capture it. We have found that Gen AI is the unlock.

See a Gen AI analytics tool
Two casino employees in suits standing in front of slot machines, smiling at the camera.

The industry has been a slow adopter of AI approaches, largely because adopting AI often requires rethinking process. Our partners at Hex Gaming created a platform to instead align AI recommendations with operation by teaching LLMs to think like casino personnel.

We help operators develop campaigns, gamification, and offers which casino Gen AI can personalize.

In addition, we are helping casino operators wire for Gen AI which requires a specialized tech-stack.

Why is Gen AI such a game changer?

A good host knows understands how to engage and develop players. A good surveillance operator can spot a suspicious on a busy floor. The most valuable finance managers understand the spirit not just the letter of the guidelines. While traditional AI is great at spotting patterns, it can not recreate this thought process. Gen AI closes the gap, here how:

Experienced Managers Gen AI Recreates
Deciding how to solve a problem
  • Tree-of-thought planning
  • Crafting reasoning pipeline
  • Iterative feedback loops
Intuition about the situation
  • Routing questions to the right model
  • Access to pre-trained predictive models
Institutional knowledge
  • Access to guidelines and calculators
Player knowledge
  • Access to ratings and player preference
  • Access to tools to summarize and set objectives
  • Access to curated examples and a system finding the closest examples (RAG)

We built the casino industry's most advanced Gen AI analytics tool — converting our proprietary analytics into tools and exposing them to casino-specific Gen AI agents from HexGaming.AI.

Hex Insights (Rated patron database) See more examples here or try it yourself live

Where we can help

  • Gen AI requires a different technology stack than traditional AI.

    From configuring GPU servers, fine tuning large language models and then deploying those models on-prem, to standing up the technology needed for the shift from queries and case statements to Gen AI reasoning and ensuring compliance all requires a new approach to technology.

  • Pre-processing data: Converting raw data into granular trends Gen AI can interpret is the bedrock of Gen AI making meaningful recommendations.  We approach this challenge by pre-processing data in the same way we approach consulting engagements.  To make actionable recommendations, this pre-processing varies according to the use case.  For instance, with a marketing engagement we focus on gaps in specific behaviors like spend per day, churn and frequency whereas as with integrity we focus on unusual behaviors and suspicious correlations.

    Examples / RAG: The heart of Gen AI is providing ideal examples and tailored recommendations for specific situations. Crafting these examples requires deep casino experience

  • Tool Library: In order to craft realistic recommendations, the Gen AI model needs both access to appropriate data in a format it can understand (see above) and the skills to draw conclusions from that data. We have a deep library of analytical approaches — from data queries to advanced analysis to machine learning predictive models — all roads tested at properties across the globe. We've converted these techniques into tools a Gen AI agent can utilize.

    Reasoning pipeline and routing: Based on the objectives of the problem, the LLM agent needs to know which tools to utilize and how to break down a problem into steps and reason through that pipeline.

    The Gen AI model is free to then augment, mix and merge the tools in order to achieve the objectives.

    We maintain the only library of casino specific LLM tools.

    Visualization: We've moved our collection of casino-specific charts into a library that Gen AI agents can manipulate. This is the first such library in the casino industry.

  • Gen AI on its own, even with smart prompting, often struggles to produce realistic recommendations. What changes that is merging casino-specific predictive models into the Gen AI agent — bringing traditional machine learning to life. Through our advisory engagements, we've built a library of tested predictive models spanning promo sensitivity, lifecycle marketing, and more. These models have been converted and mapped so a Gen AI agent can call on them directly.

  • Beyond our work with Hex, who is commercializing casino-specific Gen AI agents, we've developed several bespoke Gen AI applications, including forecasting agents for casino staffing and table games pricing. We recently built a casino analytics agent combining Hex's LLMs with Differential's proprietary analytics IP which we make available license free to clients.

  • Compliance is fundamental to operationalizing Gen AI agents. It ensures that guidelines, safeguards, and even strategies are followed faithfully