Revenue Managers will disappear in the near future

Posted by JaniP on 2018-05-10

The Question

At HSMAI Europe ROC and DOC event in April 2018 an interesting question was presented by hosts Kelly McGuire and Steven Dow, namely will AI completely replace revenue managers at hotels in the near future? The fun part of this presentation was the role of the audience as a voting body. Everybody had to take a stance on one of two sides of the hall, on the right-hand side those who believe the answer to this question is YES and on the left-hand side those who believe the answer is NO. What was most interesting is the fact that many of the people in the audience actually stayed in the middle of the hall. Why?

As always with topics regarding future, the reasons for this behavior are many. But the most important in my opinion is the lack of understanding of AI (artificial intelligence). And I certainly don't want to be offensive to anyone or underestimate anyone's intelligence, but I believe many professionals in the hotel industry haven't really met AI in real life. The closest encounter was probably the product of Hollywood's imagination presented as a superior entity in some SciFi movie or maybe Amazons' Alexa. Unfortunately, the truth is far from that perception. In case you wonder - I stood on the left-hand side and I don't count myself by any means as an expert in AI.

The Reasons

Because the opportunity appeared to justify my decision I took it and counted 3 reasons why I think AI will not replace Revenue Managers in the near future:

  • The data quality that is needed in order for an AI algorithm to learn successfully is not at the appropriate level due to heterogenous environments, lack of data integration and many manual data handling operations,
  • The unpredictability of the hotel industry environment can cause "black swan" events which can distract AI algorithms into producing bad outcomes,
  • AI's ability to understand the concept of the future is non-existing, since "entities" without a physical body cannot develop self-awareness which is the key ingredient that unlocks understanding of the future and its relation to the present.

Of course, I got a nice reply from the right-hand side audience in the form of following answers (disclaimer: time difference between creating this blog post and the event might affect my memory of the exact answers as presented by Lennert De Jong so I apologize for any misinterpretation of the exact words in advance):

  • The data quality will be no issue in the near future due to new technology development,
  • The "black swan" events will be all included into the algorithm as they will be learned and handled appropriately within a couple of months,
  • As Tesla is demonstrating with their autonomous cars and Google with its Deep Mind AI, there is almost no task that AI cannot perform better than human regarding data handling and predicting outcome.
OK - all of the above seem very valid arguments at first sight and since at the time of the event there was no place nor time to start an excessive debate around these issues, I didn't provide any reply.

The Arguments

However, I can use the power of internet and the privilege of blogging to reply and follow-up the debate here:

  • The data quality can be solved with "new technology" only in the way of implementing a common technological architecture or platform, where all the data is controlled and maintained with the same rules, the same master data definitions, the same formats and the same structure. Implemented at each and every hotel in the world. In other words, pure technological "communism". Since that is very unlikely to happen in a competitive business world (and because this is a really bad idea) the data quality issues will still be around because different people/businesses will use different solutions and technology, some of them quite old. Therefore, the challenge of data integration performed and maintained by people will not go away anytime soon.
  • The unpredictability of the hotel industry environment is on a completely different complex level than the problem of voice recognition or gaming. The first one being Amazon's Alexa needs to find the right answer among the finite number of answers by learning the finite set of rules regarding speech and word recognition. And in order to learn the right answer there is a very strict rule to what the right answer is so the algorithm can "learn" right from wrong. The second case being Google's Deep Mind or more specifically AlphaGo is an AI which can outperform a human by playing very complex game of Go. Again, to teach this algorithm the "right" from "wrong" it takes a very strict and limited sets of rules involving huge amounts of combinations. The key issue comparing to the hotel industry, or any other business for that matter, is that a definition of the "right" answer is very vague. In other words, AI usually solve multivariable problems with single solution, business on the other side is a multivariable problem with multiple solutions. This poses a real issue for the process of teaching AI. In case of previously mentioned AIs there were vast numbers of people involved in the teaching process who gathered the data, structured the model and performed the learning process. In the case of hotel business, who will teach AI which answer is correct if answers to the same question can be different? If we entitle AI to solve specific issues with a single solution (like forecasting the demand) of course the algorithms will provide much better solutions than a human but the final decision of business tactics used upon that information is still better and cheaper by far performed by humans.
  • Strategic thinking is one of the key properties of humans over nature. There is no other Being in this world who applies the possibility of the future outcome to current actions. In philosophy and psychology this property is called consciousness (yes I can already see a huge distress within a philosophy and psychology communities) and that property is impossible to create within a bodyless algorithm. Because of this reason again AI will almost certainly replace several human operations within RM practice but it will not replace the human completely since business is never successful without strategic thinking.


So, there it is. The reply I couldn't put "on the table" at HSMAI ROC. But at least we had a lot of fun during this presentation and I hope I am not the only one who thinks these kind of discussions are far more important than just for general amusement purposes.