Travel: the AI frontier

Authored by Carl Michel, an Operating Partner of The Strategy Exchange, and originally published at

This week, I got an email from a large and well-known online hotel provider offering me some competitive options to book a hotel in Porto. Nothing extraordinary in that and yet it was completely and totally irrelevant. I had visited and stayed Porto a few months ago, so why was I being lured back? It’s a charming city but it is 100% possible to see all the wonderful sights over a long weekend and to be so fulfilled as to have no (or very little) intention of returning for many years. And yet somehow my once-off interest had been logged by the hotel booking site as someone who was a repeating visitor.

The result is that I now know that the booking site (known in the industry as OTAs, or Online Travel Agency) doesn’t really know me, despite a lot of booking “data”. I am just a customer, an order-giver, but nothing more. It’s as if your travel agent had suggested the same trip to you after six months – at the very least you would give him or her an odd look, perhaps you might suspect something worse.

Of course systems are not people. They do not apply the cognitive reasoning that we call understanding or empathy to suggestions. There is no heuristic – or learning – in what they suggest.

These transactional systems often apply a very poor common sense filter: we have all clicked on a cheap flight to Vienna thinking we will save £10 only to discover it goes via Kiev with a 12-hour duration or that it involves an overnight in Brussels. Real people wouldn’t never make such suggestions – unless they are feeling particularly malicious – but systems do.

That all may be now changing – in a long-heralded process whereby artificial intelligence (AI) begins to creep into the travel industry.

AI is the next big thing, it is coming soon and will be crucial in setting the next frontier in travel, which has been for the last decade still happily stuck with worn-out themes such as the impact of mobile, while others still wonder if Big Data is a potential goldmine or just quicksand that is hard to monetize.

Huge leaps in AI have come about due to vastly increased computing power, the fact that this power can be rented from the likes of Amazon or Google and that we can now get results from so-called artificial neural networks very much faster. AI has allowed recommendations to become relevant and accurate: it has enabled deep learning to mimic (and even surpass) humans in certain cognitive tasks. The most celebrated moment was the AI system AlphaGo that beat the Grand-master of the Chinese board game GO earlier this year.

The good news first – AI will not (at least in the foreseeable future) spell the end of travel agents.

Advice, suggestions and some creativity will still be sought out. And delivered to us by entities with a pulse. Indeed, technology has always created more jobs than it has destroyed so the only uncertainty is who will be a winner or a loser. Empathetic, smart travel agents will continue to thrive.

The bad news – anything in travel that is more routine and requires the fast assimilation of lots of data can be rendered much more effectively by an artificial neural network. And better still you can now talk to a Siri or Cortana and get the results quickly. No need to type and search in Google and then trawl through myriad customer reviews to see what the hotel thinks about dogs, vegans or noisy babies; no need to cross-check climate charts to see if Mozambique is rainy in January (hint: it is); no need to explain that you would rather fly with a certain airline provided it is not more than 10% more as you are collecting miles on them, your personal AI concierge will soon find you a pretty good trip that will meet all your criteria, plus offer you a few great add-on suggestions and also let you know which of your friends has been there/done that before or even lives nearby.

Like anything, AI needs to feed off a large volume of data points, so that it progressively learns and gets better at matching your needs. Initially it’s much like walking into a travel agency for the very first time and mumbling what you want. But if you articulate with a lot of qualifiers the things that you need, the stuff you would like, those aspects that don’t bother you and highlight all your pet hates, the result you get will be much better. It all depends on how long you are willing to invest in self-disclosure and the ability/willingness of the agent to “figure you out”.

But AI has an advantage: it doesn’t forget you. So unlike your travel agent who may move branch or simply forget, AI has your details and will build on them. It is thus quite conceivable that in perhaps 5 years from now, we may say that “Wow, no one understands my travel needs like Siri or Cortana does” (or whatever new personal assistant names we have by then… perhaps even Firedrop!).

In the shorter term, AI will replace travel agents who are doing no more than sifting through lots of data for you. Agents get bored, they can’t sensibly gather all the information or know everything, they may hide the odd awkward data point and so they will offer you a good or satisfying solution in the 20 or so minutes you are sat opposite them or on the phone. In contrast AI applies rules that, from countless observations, proxy for common sense filters and can simply draw in and process a lot more data, which is then weighted and filtered. You may not get any truly wonderful surprises but it will be optimised and you will definitely avoid any nasty shocks, which matters a whole lot more to nearly everyone.

They can also learn dynamically from your tastes: whether you like it a bit more upmarket; a bit nearer to the sea; a bit less red in the décor and effortlessly present a new set of options that would leave your travel agent biting his or her nails to the quick as they see yet another half hour slip away.

Early trials of concierge systems have resulted in mixed results. But this is simply a matter of time and about getting more data to learn from. The quality of results will inevitably improve. It takes longer to do the machine learning with 100 variables, such as in travel, than for much more defined product sets. So a short-haul flight will be considerably easier than a Round-the-World ticket than would be an adventure trip to Iran. And in some cases, AI may simply act as a triage point, saving a lot of wasted time for all as you gradually move from a vague idea of a beach holiday under £5000 to something much more precise and tailored, which an agent with a smile, charm and cup of tea can convert into a sale. But then again, maybe by then Cortana or Siri will have their own friendly personality traits!

AI is coming – so get ready for the next frontier in travel.


Leave a comment

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.

One thought on “Travel: the AI frontier”