Skip to main content

Generative AI and Sensitive Hotel Data

With the contracts, regulations, and expectations involved when handling sensitive data, there are some things to keep in mind when implementing a generative AI tool.
laptop displaying personal data
Advertisement - article continues below

ChatGPT and other generative AI tools continue to create buzz in the hospitality industry. Hoteliers are uncovering more about where it works, where the security and privacy issues are, and how to absorb the enhancements that have come along since its initial splash. 

ChatGPT, launched by OpenAI, became wildly popular overnight and galvanized public attention. (OpenAI’s DALL·E 2 tool similarly generates images from text in a related generative AI innovation.) Top-tier analytic firms such as Gartner are following this technology closely and have been for some time. The firm has released several pieces on generative AI, including how it can contribute to business value, impacted industries, best practices, and some interesting predictions.


Gartner sees generative AI becoming a general-purpose technology with an impact similar to the steam engine, electricity and the internet. “The hype will subside as the reality of implementation sets in. Still, the impact will grow as people and enterprises discover more innovative applications for the technology in daily work and life.” Its predictions reveal that the technology is primed to increasingly impact enterprises over the next five years: By 2024, 40 percent of enterprise applications will have embedded conversational AI, and by 2025, 30 percent of enterprises will have implemented an AI-augmented development and testing strategy. 

For now, the clear use case seems to be anywhere there is a natural fit of public or non-sensitive data with a smooth, natural language interface. Hoteliers and their third-party vendors are broadly adopting the technology for functionality that meets this clear use case. They’re either building it in-house or leveraging solutions from the numerous technology startups providing specialized functionality around generative AI tools. 

Some examples of generative AI in use today:

  • Customer service
    • Vastly improved customer interactions for automated assistance
    • Virtual concierge
    • Multi-lingual support
  • Summarizing, categorizing, and building recommendations from reviews and feedback. While somewhat sensitive, much of this data is already public.
  • Targeting messages and audiences within marketing.

Any area of the hotel where the backbone is natural language will continue to evolve and expand. The challenge that continues to exist, however, is for solutions using, or built on, private and proprietary data. Any unauthorized parties must not see this data. To protect this type of information and satisfy security and regulatory requirements, there must be tight control over where the data lives and who has access to it.

More recently, the technology backbone has emerged that can provide the framework for those in the data-sensitive world - a dedicated instance. OpenAI, Microsoft, and many other vendors using generative AI have recently made this available. This is the first step to unlocking the possibilities, as it allows sensitive data to stay on the ranch. Unfortunately, these options come with a hefty price tag for most at present. 

As time marches on, and time is marching quickly, these costs will come down. Those with an appropriate ROI (return on investment) proposition can and will be able to justify these costs. With a dedicated instance, innovators can feel the level of confidence needed to proceed. Though this is an emerging technology, it would still be wise to tread carefully with sensitive data.

Tread Lightly, But Carry a Big Stick

With the contracts, regulations and expectations involved when handling sensitive data, there are some things to keep in mind when implementing a generative AI tool. 

  • Isolate unnecessary data from the generative AI instance. Only put the minimum required into the data store and model to realize the value provided.
  • Minimize and anonymize sensitive information exposed to the system as much as possible. This is especially true for personally identifiable information (PII).
  • If the ability and expertise are available, consider deploying an on-premises or private-cloud-based instance. This gives more control over the data and minimizes exposure.
  • Ensure the proprietary data used with the language model complies with relevant laws and regulations, such as data protection and privacy laws.
  • Ensure your chosen generative AI solution complies with customer contracts and terms and conditions.
  • As with any solution…
    • Use secure, encrypted storage.
    • Use secure, end-to-end transport protocols.
    • Implement access controls to restrict data access only to authorized personnel.

With dedicated instances, preventative caution, and solid ROI propositions, the hospitality industry is sitting on a gold mine of possibilities. 

There are only two major hurdles to building solutions based on this data. One is the expertise. Because this is a new technology, the personnel are expensive and hard to find. Most generative AI tools promote themselves as easy to use, but when building proprietary models on proprietary data, the level of expertise required increases dramatically. 

The second hurdle is the cost of the dedicated instances themselves. As the technology in this field advances, these system costs will come down, and the availability of competent, affordable workers will increase.

For larger vendors and hoteliers, current costs, as a factor of revenue and potential cost savings, put proprietary solutions within grasp today. Most others will continue to focus on solutions that take advantage of the public cloud versions of generative AI tools, such as ChatGPT until the costs come down. 

While waiting, it would be wise to start planning. Take baby steps. Hire someone with expertise and begin identifying the business cases. Investigate areas that can affordably be used today, whether with non-sensitive or anonymized data, that can safely risk exposure. Regardless, those dealing with large amounts of sensitive data must pay attention to technologies like ChatGPT and make plans. For those that don’t, expect to be behind those who already have.


About the Author

Rob Landon is VP of Engineering of Knowland

This ad will auto-close in 10 seconds