Enterprise Information Technology Leaders need to effectively operate today while planning to support a changing future. This has never been an easy job and over the past two years it has been significantly more difficult. But 2022 and beyond does not appear to be a time horizon where IT Leaders can catch their breath. In fact, new and future demands for technology capabilities across every area of Casino Gaming and Hospitality continue to accelerate. Perhaps the most daunting of these new and future demands will be the need to create a set of IT capabilities that can support the growth of Artificial Intelligence (AI) and Machine Learning applications (ML).

The first AI/ML applications in Casino Gaming and Hospitality were implemented years ago with IBM-Watson powered AI-concierge application known as Ivy. Since those initial implementations dozens of other use cases for AI/ML have been created across the Gaming and Leisure industry. Currently the use cases are shifting from cognitive-based reactive responses to AI-based predictive and proactive applications.

Success in AI/ML depends on collecting, storing, and processing massive amounts of data. This foundational requirement for Enterprise IT Architecture is a major challenge for IT Leaders. Before we discuss the challenges and solutions we should set the stage with common definitions of AI, ML, and Deep Learning (DL).

AI is the broad science of mimicking human abilities. ML is the subset of AI that trains a machine how to learn. ML automates analytics model building using statistics, operations research, and physics to find hidden insights in data. DL uses large neural networks with many layers of processing units designed to learn complex patterns (think speech recognition as an example). It is not unusual in this early stage of AI to see these three terms used interchangeably but that is not technically correct. However, the challenges to IT Leaders are the same for any of these applications.

AI in Casino Gaming

It is not difficult to think of applications for AI that will benefit Casinos. Predicting future player behavior, fully understanding the lifetime value of a player, understanding player preferences to reduce player churn, optimizing the floor, and demand forecasting are just a few obvious examples. Those real-time applications could clearly optimize costs, improve player/guest experiences, and ultimately increase profits.

Other, near-real time or non-real time applications exist in Marketing. Database marketing based on the “Recency-Frequency-Monetary (RFM)” model has been the standard in casino marketing for many years. AI and ML methods for Marketing are becoming more attractive. Specifically, customer segmentation based on clustering algorithms such as K-Means or DBSCAN are showing promise when powered by Machine Learning. With the help of AI, casinos can improve their services and security while creating a more enjoyable experience for players and guests.

The Technology Challenges

For IT Leaders, one thing is clear. As AI moves more deeply into the organization and adoption is increased, it will demand significant computing resources and infrastructure costs. Building cost-effective platforms to run these intensive applications will be a requirement.

IT Leaders will need to work with Business Executives in the organization to build flexible IT capabilities and infrastructure. Given the large data stores, the need to apply huge amounts of processing power to those data stores, and the need to keep running all the existing applications and services – IT Leaders have their work cut out for them. The result will likely be an effective Hybrid Cloud Architecture that hosts or integrates with vendor software (with embedded AI).

The high-level technology requirements for successful AI adoption center around solving for the new stresses it will create in data management.

The collection and storage of the volumes of new data present one set of challenges. The location and type of storage will be dictated by where/how the data is collected and the latency of data transmission (either to storage devices or to AI application processing). These dynamics are likely to induce the need to create or grow edge infrastructure to reduce/eliminate data transmission across enterprise networks and/or cloud connections.

The critical success factor for good data management is the understanding of the data that you collect-store-process-create. As we will see below that understanding is the basis for creating universal data management policies and actually enforcing those policies across relevant data types.

We offer the framework below to help IT Leaders plan for this tremendously difficult task. Here are four attributes that your Enterprise Architecture Design and future Infrastructure must have to be successful. Following these attributes are six operational challenges that you need to have built into your future state architecture/infrastructure.

1. Effective
   It may seem obvious that the architecture needs ‘to work’ but the demands on the architecture will be unparalleled. Performance across data stores, processing, and end users will need to be in real time (a few seconds at worst) for predictive or proactive use cases. The architecture/infrastructure will need to be understood/supported/operated by a defined set of resources and skills. “It needs to work” has never been so difficult.
2. Scalable
   The platform needs to be able to scale to take on new applications, new datasets, new end users along with growth in existing applications. The danger to avoid is creating new ‘silos’ of infrastructure and/or vendor applications.
3. Highly Available-Resilient
   As the business begins to rely more on the AI/ML applications, downtime or delays will create business problems. Availability of the applications and data to the end users will increase in importance as adoption expands.
4. Cost Effective – Cost Transparent
   The architecture must be designed, and the infrastructure configured in a way that enables full visibility of cost of ownership and cost of expansion/growth. Pricing by usage (processing capacity or data storage size) is not unreasonable but needs to be fully transparent with the ability for IT Leadership to model the actual cost of ownership curve throughout the application lifecycle.

But wait, that’s not all. Those four attributes are critical success factors to a good Enterprise Architecture and Hybrid Cloud Platform. But you also will need to solve for the following six challenges.

1. High Computing Capacity
   Adoption of AI will create demands for high computing capacity. Deep learning involves multiple large data sets and processing complex algorithms. CPU-based compute environments may need to give way to GPU-based platforms. Computing capacity and density will grow, as will demand for high-performance networks and storage.
2. Storage Capacity
   The growth of storage and the location of storage (close to compute resources) will require a scalable set of storage capacity across your hybrid cloud. Real time AI will put the highest demand on latency across your storage and network. A foundational truth about AI applications is that they make better decisions when they are exposed to more data.
3. Network Infrastructure
   Networking is another key component to an effective AI architecture. Some AI algorithms are highly dependent on communications. Scalability is a high priority and that will also require high-bandwidth, low-latency networks. If your hybrid cloud spans regions you will need to ensure that your provider can offer consistent performance across all regions.
4. Security
   As your hybrid cloud architecture develops, consistent and effective security is critical. In many cases the data sets used for AI will include some of the most sensitive data in your enterprise. Breaches of these data sources would be a disaster.
5. Flexibility and Agility
   As AI is adopted across various use cases, the architecture/infrastructure will need to be able to flex into growth and it will also need to be agile enough to support new demands. Remember, the demands on your architecture are not going to be solely due to AI applications, your existing workloads and new ‘traditional’ workloads and services will continue to be added to your application portfolio.
6. Data Privacy and Protection
   In AI applications, the data is the source of any value created/derived. As pointed out earlier, success in AI applications is based on extremely large data sets that continue to grow. Your architecture is likely to have data sets stored on-premise, with cloud providers, with Software as a Service vendor, and even on network edge devices.

Solving this problem will rest on having a global policy-based approach to your data management. This includes knowing what data you are collecting-storing-processing and then setting management policies for each data ‘type’. The policies will inform the various Data Protection Tools to encrypt certain data or obfuscate certain data. The policies will also be guidelines to support retention/archiving practices and backup/recovery practices.

Some of the data will invariably fall under governance requirements (GDPR or other Consumer Privacy Laws). Anything other than a comprehensive understanding of your data and a policy-based approach to data management, data privacy, data protection will make it virtually impossible to have the ability to comply with some consumer privacy requirements.

The benefits of AI for casinos are basically endless. Using AI as a tool to predict the future enables organizations to improve almost every player and guest touchpoint while increasing operational efficiencies. The use of AI is going to become an important part of any casino’s evolution and data-driven strategy.

Don’t Forget Traditional Architecture

We have covered the attributes that your architecture needs to have in order to succeed – and – we have highlighted some of the challenges that need to be solved with your future infrastructure and operations. But do not forget the more ‘traditional’ elements of good architecture and IT operations.

As you design, build and manage – be careful not to build silos based upon specific use cases. Consistent and global policies, solutions, and practices are preferred. Be careful to avoid vendor lock-in where possible. Pay particular attention to reducing risk of obsolescence in any vendor selection, design, or choice. Most importantly, as you are solving these new problems make sure you are keeping the lights on and continually improving your current applications and operations.

There are as many opinions about the efficacy and ethics of AI as there are use cases to apply AI and create business value. AI has the ability to change lives for the better. Casinos are just one industry that can benefit greatly from AI.

Source: G&L Magazine Spring 2022 Edition