It seems every day there’s another announcement surrounding some preternatural feat that artificial intelligence (AI) has accomplished. Within the hospitality world, the number of technology vendors introducing (or claiming to introduce!) AI products and features is also mushrooming. It’s all a bit overwhelming.
In our previous article for the last issue of Hotelier Magazine, we tackled the question of where to start. There are some clear and present applications like chatbots, AI voice assistant, guest sentiment analysis and workforce automation tools with underlying machine learning (ML) functionality.
The underlying motto is ‘move fast and break things’. Deploying solutions right now will always be messy. Platforms don’t work exactly how you want them; they may not have the integrations you want or have full data structuring within those interfaces; they may be cost-prohibitive for current budgets.
However imperfect a solution may be, deploying now is still better than getting left behind. You need your teams to be comfortable with a faster cycle of implementation, from vendor research and RFP through to staging, training, go-live and ongoing refinement. You need your teams to bring forward ideas for how to better use the software you are already paying more and to identify repetitive tasks that should be automated. This is no longer just about AI but about having business process innovation (BPI) permeate your entire culture.
As background, here are some applications to get you in the mood:
- Chatbots, virtual concierge and omnichannel inboxes
- AI voice assistants to replace interactive voice recordings (IVR)
- Review sentiment analysis and conversational surveys
- Dynamic pricing, dynamic availability and prearrival upselling
- Rate planning, demand forecasting and market intelligence
- Adaptative marketing and communications content
- Real-time service recommendations based on guest profile data
- Staff scheduling, team learning and upskilling
- Preventative (and now predictive) maintenance
- Energy management systems, food waste reduction and kitchen management
- Cybersecurity detection as part of a zero-trust architecture
But that’s how you get off the ground. Where is this all going? What’s the grand vision for ML, AI, the use of large language models (LLMs) and becoming a learning organization?
To present one future with a strong possibility of becoming a reality, think back to the switchboard operator or executive assistant whose primary purpose is to take incoming calls, understand the recipient’s request and its context, and then route the recipient through to where they need to go.
If we envision AI working in a similar manner, we would have a master AI program that would act as an orchestra conductor – an AI orchestration layer as it’s being called – whose job is not to complete each task, but to simply interpret then direct the incoming request to the right bot within the entirety of the agentic workforce that has been trained to execute this task to perfection.
These bots may be proprietary ones developed in-house using enterprise data for a specific automation task or a form of robotic process automation (RPA). They may use an LLM with an additional layer of refinement or lensing known as retrieval-augmented generation (RAG). They may natural language processing (NLP) to interpret your back-of-house procedures or those coming in from guests. Or they may interact with outside web2.0 entities or APIs via model context protocols (MCPs).
Oh, the acronyms! Given how complex hotel operations are, there will undoubtedly be hundreds of different AI agents or bots that will be developed in the coming years, with many available for out-of-the-box purchase as part of an agentic workforce that you can assemble to increase operational efficiencies. But at the core of all these will be the conductor, the switchboard operator, the orchestration layer.
Are orchestration layer tools available today? Yes, they are expensive, even for those hotel companies with hundreds or thousands of assets under management. But the rule of technological democratization still applies; we just cannot say when exactly these tools will be offered at a cost that works for small groups or independents.
For now, though, understand where this is all going and align your innovation cycle so that you aren’t stuck in a wait-and-see pattern. Implement specific-task AI tools now to get your feet wet in order to realize gains today, and then newer and better solutions will become apparent as the years march on.
