Artificial intelligence – from overhyped to overwhelming: Demystifying the practical purpose of machine learning
Suggesting that machines are thinking is just one of the incorrect explanations of artificial intelligence (AI) that Tom Seddon, CEO-in-residence of Foundry.ai, sets out to debunk as he looks at how AI in travel and hospitality is making an impact.
“AI is simply data and maths, working in a sophisticated manner. But it is frequently overcomplicated and one of the most overhyped terms being used in travel and hospitality right now” warns Seddon. “Machines can have extremely sophisticated pattern recognition, and putting this to profitable use in a practical way is how AI is best harnessed for business success. When we think of AI we commonly focus on self-driving cars or robots creating incredible paintings, and whilst these are applications of AI, they don’t represent what can typically be usefully achieved by AI in business.”
AI comes into its own when applied to frequent decision processes. Its power is in better applying data to improve frequent decisions that make – or save – a business money. And of course this application must be practically useful. Decisions made day in, day out, such as resource and shift planning, food and beverage stock ordering, invoice monitoring and more are all the kind of tasks where AI can excel. An intuitively human decision, often made very infrequently, such as deciding whether or not to hire someone, is not usually a good use of AI.
Seddon explains, “Where there are lots of examples of data to be analysed, AI can be really powerful. But it relies upon reams of data, something that really only large companies can apply. In fact, the more data to analyse, and the AI impact keeps getting better.”
OVERCOMING SCEPTICISM – PLOTTING PATTERNS TO ENHANCE HOSPITALITY TECHNOLOGY
Given its reliance on data volume, Seddon is keen to point out that smaller hospitality businesses are right to remain sceptical about how AI can truly help their business. Without volume, AI is inaccurate and cannot bring tangible benefits.
“AI delivers an advantage for larger hotel chains. But not all chains are taking the advantage. There are multiple opportunities for those who embrace the many good uses of AI. For example, in deciding what to offer to customers based on your interactions with them. AI can identify patterns as to when pointing them to the website, sending emails or even phone calls are the right thing to do. But brands need to separate out the noise” outlines Seddon.
AI’s successful applications for the hospitality sector might be more skewed to larger businesses, but there are still ways that a smaller business can take advantage. Seddon warns that assessing the quality of the data is critical. “Small businesses must rely on someone else’s data to train the systems and, in this case, you have to be absolutely confident about the data quality.”
For businesses with the right data, AI in hospitality is an essential tool to be more efficient, drive greater profitability and enhance accuracy. It isn’t a new application. Hotel technology has been built on early iterations of AI for many years. The way data has been used in pricing and revenue management over the last 20 years is AI at its best, and today even more sophistication is possible.
FREQUENT DECISIONS – AI IN HOSPITALITY’S SUCCESSFUL ROLE
“AI is always best applied where there are very frequent decisions being made,” explains Seddon. “Importantly, these frequent decisions should have a profit impact such as in areas of resource planning. AI can analyse how many orders are made in the hotel restaurant at a certain time on a given day of the week. This aids planning people and shift patterns with greater accuracy and brings you closer to having the right number of people.”
“Predicting how busy an area of the hotel might be, and therefore how many staff are required, is an area where AI can learn from all other business units of the hotel and their historical performance. A hotel chain has the significant advantage of being able to analyse hundreds of thousands of transactions – not just from its restaurant – to build a more accurate picture.”
Being able to “learn sideways” from other hotels, not just the past history at one hotel, as Seddon refers to it, is a powerful route for AI. Technology can apply the data from other functions in the hotel, even competitors or hotels with similar circumstances to your own, to find patterns and highlight those that might be meaningful.
AI is good in functions like supplier bidding to find qualified vendors and allow more competitors to take part in a bid than a human might have time to assess. It also comes into its own in niche areas that require pin point accuracy, but can take people a lot of time. For example, AI can check for discrepancies on small invoices, something that many businesses cannot afford to employ a person to be dedicated to.
SEPARATING THE NOISE – WHERE AI DOESN’T SUPPORT THE CUSTOMER EXPERIENCE STRATEGY
But, there are other areas of a hotel’s business that seem to fit the frequency definition, yet do not deliver the right results.
One area where Seddon has seen AI applied to deliver accurate, but not necessarily meaningful, results is when it is used to analyse social media and reviews. “The key piece to remember about reviews is that the action lies back with the hotel. A hotel chain monitoring reviews will require an individual hotel to deal with any issue raised. So by the time the issue is spotted, flagged to a hotel and that hotel responds, it would have been quicker and easier for the individual hotel to monitor their own reviews. Even some of the largest hotels don’t receive huge volumes of reviews every day, and so it is manageable and reasonable to expect the ownership to be at hotel level.”
Seddon smiles as he recounts social media sentiment analysis going through hundreds of thousands of reviews, only to confirm that hotel guests prioritise cleanliness – something that every hospitality professional is of course already aware of. Whilst this is accurate insight, it is somewhat useless in that the AI is not telling business owners anything they don’t already know.
PEOPLE POWER – TRUST AND CONFIDENCE
When hotels are looking at their customer service experience, there are some areas of operation where humans remain trusted over AI. “Whilst using AI to support cleaning rooms sounds like it can work on paper – robot-style vacuum cleaners are commonplace and AI can use data to programme devices to carry out certain tasks – there is rightly some nervousness as to if this really improves the guest experience” explains Seddon.
“The key decision making factor for whether to employ AI is to ask if technology can truly deliver the same, or better, result than people and importantly do this cheaper than people. If AI can be more cost effective than people, and get to at least the same outcome, then it can really work hard for a business.”
Seddon continues, “When it comes to cleaning a room, the end result has to be absolutely perfect. No one wants to go into a hotel room and see a trace of another person having been there. We instinctively trust another human to carry out cleaning to this level of perfection rather than a machine.”
Our trust in AI has limits in all kinds of areas, with driverless cars a good example. “AI can excel at very complicated maths and patterns, but it is commonly the last 10 per cent of situations where it gets incredibly challenging” adds Seddon. “Five to ten years ago, it was reasonable to expect that driverless cars would be more mainstream than is actually now the case. AI can be programmed to carry out many of the tasks and challenges associated with driving a car but the unpredictability of, for example, an animal suddenly running out in front of a car, is a really difficult scenario for AI. There is no easy way to train a system to accurately predict this.”
For technology trends in the hotel industry, security and room access is another area where AI currently has a trust challenge. We are all fairly comfortable with our mobile device opening our room door – something we see more and more of – however hoteliers are drawing the line at using facial recognition. Seddon summarises, “The technology is there, but there are risks and consequences to think through. Hotels are righty greatly concerned about guest safety. And, at the moment it’s difficult to see using facial recognition can make hoteliers more profitable.”
WATCH OUTS – THE PITFALLS OF AI IN A POST-PANDEMIC WORLD
At the time of talking, the hospitality industry is beginning to plot how to surface and recover from an almost-global lockdown. Following a period of extensive restrictions on travel during the 2020 COVID-19 outbreak, and government enforced hotel closures, new ways of working in a world of social distancing where human contact is discouraged are beginning to emerge.
In Seddon’s view the reasons for embracing technology in the hospitality industry have been accelerated due to the unprecedented position all businesses are in. The exasperation of business challenges is one way the COVID-19 pandemic is pushing hoteliers to embrace AI. However, as all hospitality business work out how to operate in an environment with significantly lower demand, there are huge watch-outs for businesses. “People and systems can do a lot of unintended damage” warns Seddon.
“Pricing control is one of the areas usually determined by technology that needs to be very closely controlled. Most revenue management systems see low demand and respond by lowering prices. They also see competitors lowering pricing, and respond with what historically would have been more attractive rates. However, in these immediate months as the world learns to live with COVID-19, there is significant probability that someone travelling is in fact prepared to pay more, despite booking on short lead time. As we all scrabble out of a lower demand environment, there needs to be a focus on revenue management tools learning from more recent history rather than 12 months ago.”
Qualifying where AI is right for your business requires looking at how fast technology can do a good a job – or better – than people, and if it can help your business adapt to changing scenarios whilst also saving money. Expecting AI to be a machine doing intuitively human thinking is to overestimate its capabilities. The sweet spot is in applying machines to find patterns and trends about your business that people don’t have the time to spot. In an exceptionally tough operating climate, where hotels are forced to employ fewer people, the right applications of AI will really make an impact.