INTRO

Entertaining the Human Race

MGM Resorts International, an S&P 500 global gaming and entertainment company, operates 31 distinctive, Las Vegas-inspired destinations around the world—truly living by its mission to entertain the human race.

Before this initiative, MGM had already implemented an AI-powered SMS chatbot to support Guest Services at two flagship Las Vegas properties: Bellagio and ARIA. However, the Product team saw an opportunity to create a more engaging and integrated experience by expanding its capabilities within the MGM Rewards app.

SUCCESS METRICS

In 2023, the SMS chat system at Bellagio and ARIA handled over 20,000 guest conversations. Any inquiry the AI couldn't understand was escalated to a live agent. At the time, the baseline chat containment rate—the percentage of conversations resolved without human intervention—was 40%.

The Product team aimed to boost this containment rate to 60% and scale the AI chat experience across all U.S. properties, with a projected capacity to handle 2.1 million conversations annually. This expansion would enhance the guest experience through faster, more consistent support while also improving operational efficiency across the organization.

SMS AUDIT

Identifying Use Cases, Gaps, and Opportunities

To define the scope of in-app chat, we began by analyzing existing SMS use cases and identifying which ones could be enhanced through a richer, native app experience.

We reviewed conversational data to help prioritize features for the MVP. For example, towel requests emerged as the top housekeeping-related inquiry:

I also identified friction points within the current SMS conversation flow, and in collaboration with our content strategist, proposed enhancements to improve clarity and user satisfaction.

COMPETITIVE ANALYSIS & SECONDARY RESEARCH

Identifying AI Chat Emerging Best Practices

Although AI is a rapidly evolving field, I drew on cross-industry chat experiences and insights from theuxofai.com to establish practical guidelines for designing an effective AI chat experience.

Here are some key takeaways:

• Top prompts combined with free-form text input offered the most flexibility for generative AI interactions

• Phone-tree style navigation—commonly used in rule-based chatbots—resulted in rigid, less helpful user experiences

• Response rating mechanisms were essential for ongoing model training and improvement

• Clearly indicating whether users are chatting with an AI or a human helped manage expectations and build trust

USER RESEARCH

Understanding What Guests Expect From a Concierge Service

Next, we set out to understand hotel guests’ expectations of a concierge—both in person and through chat. The existing SMS chat experience at Bellagio and ARIA had never been formally tested or set up to gather user feedback, so we also wanted to evaluate whether it was meeting guest needs or if there were clear opportunities for improvement.

A key question that emerged from our competitive analysis was:

Should we focus on doing one thing exceptionally well (e.g., dining recommendations), or offer a broader range of services across multiple verticals?

The answer: a little bit of everything.

Through user research, we found that guests expect concierge support to provide:

• Assistance with both pre-trip and in-trip requests

• Clear, direct, and natural language communication

• Proactive suggestions, not just reactive answers

DEFINING THE NEW EXPERIENCE

Crawl. Walk. Run.

Given the technical complexity of the project, we needed to adjust our ambitions to align with engineering capacity. While we explored bold, future-facing ideas—such as advanced dining recommendations—we adopted a crawl, walk, run approach to deliver a realistic and impactful MVP:

USABILITY TESTING

Testing for Best Approach

One of our key challenges was introducing users to the full breadth of chat capabilities without overwhelming them. However, insights from our competitive analysis and secondary research suggested that presenting specific, high-demand prompts was more effective than mimicking traditional phone-tree navigation.

To find a balance between the two, our UX researcher, Maria Gonzales, helped us test a concept that surfaced top prompt categories, which led to specific, high-demand prompts.

Here are the study results:

THE LAUNCH

Delivering a Richer Concierge Experience

We were excited to roll out the enhanced chat experience to MGM Rewards app users and begin collecting real-world feedback. Below is a look at the final build and key features:

Here’s one of the many prototypes created for this project; this one primarily focuses on the prompt interaction:

Learnings

This was a complex, fast-moving project built on emerging technology that evolved in real time. To navigate this effectively, I had to:

• Stay flexible and adapt quickly as requirements shifted

• Overcommunicate with team members and stakeholders, and work closely with the conversational tech team to ensure ideas were feasible and to course-correct efficiently

• Learn and apply Figma variables to prototype rapidly and streamline collaboration with our content strategist—special thanks to Jonathan Miller for not only driving content strategy, but for also spotting the opportunity to use variables for managing conversational content at scale

↑ Back to Top