Artificial intelligence appears as a "must" with which every company must count to not be left out of the digital transformation, allowing them to automate their operations to propose a more adjusted offer to the client and in the shortest possible time. The heading of online travel agencies is no stranger to this search.The objective is simple: to analyze large flows of information through the learning of the machines so that the tourist packages offered to the user are as personalized as possible. And, in this way, get an increase in sales.Takeoff, which billed $ 398 million in the first nine months of last year (has not yet released its fourth quarter results) began some time ago to develop the branch of machine learning -more known as machine learning- of photos.Its objective: to identify images of poor quality, stretched or unclear among the offer of more than 500,000 available accommodations within its platform."The need was born from Content management, which works on the moderation of content generated by our partners," says Leandro Malanadrini, director of Product Management & UX of the company, iProUP."The problem is that they started to find images of poor quality or that generated false expectations in the users, in this context, it was decided to move forward with this problem and develop a solution", he adds.Five years earlier, the company had created a Research & Development area to adopt artificial intelligence and big data tools, which in 2017 had changed its name to Data Science.It also had an interdisciplinary team of ten people specialized in machine learning, whose main utility is to allow the automatic realization of operations on large sources of information through data banks."Standards were predefined and they identified the most frequent problems, among which pixelated and very low resolution photos stood out, as a consequence, we opted to attack the moderation of the images from the perspective of their quality, in search of a robust solution. , reliable and effective, "continues the executive."As soon as this process was started, the first question was: what is a good photo?" To answer it, it was necessary to build a manual of photographic style, based on the preferences of the users, "says Malandrini.And he continues: "Once this question was defined, the next and most important challenge was to tell a machine, without doubt, the prototyping process was arduous, it required the aggregate and the continuous improvement of observables of the image to make the algorithm cash".The implemented method is known as "supervised learning". It requires a database with all its elements labeled according to what is sought to learn."In the case of the company, this meant having a set of images with a score assigned from 1 to 5, based on the opinion of several people, with the purpose of using public image banks in a first version", he explains.However, the results did not meet expectations because the criteria used to determine the score of the images differed from those required for the Despegar graphic universe."It took nearly 1,400 photos and almost all the Content team to rate them from 1 to 5, according to the parameters that the service returned, making it possible to build a set of data that best represented the images and criteria. ", he explains.Currently, the service allows identifying distortions known as noise, ringing (when the edges of the figures are ring-shaped), blurring and blurring."In this way, two major milestones were achieved that changed the mode of moderation of photos in Despegar.In principle, you can quickly recognize the status of the quality of the photos available on the site, traveling up to 900,000 images in three days", says Malandrini.The director points out that the other important achievement was "getting the order of photos of accommodations with storytelling: the images tell the user the story of a hotel, anticipating how their experience will be there".With operations in Argentina, Brazil, Colombia and Mexico and a turnover of more than 400 million dollars, Almundo is another of the online travel agencies that uses artificial intelligence, also since 2017.In your case, to recommend to travelers who access your platform those products in which they might be interested."When they enter the web from any device or app, they find offers of destinations that are similar to them, and when they search for accommodation, they are shown in the first results the ones that are most relevant to them. Almundo, a customization algorithm enters the scene, which assigns the ideal expert for that traveler, "says Lautaro González, CIO of the company.Complementarily, according to the executive, all the email marketing campaigns and push notifications of the company are based on artificial intelligence algorithms. The purpose is to try to provide relevant content to each of the clients."All the developments were inhouse, we chose to do so because of the wide variety of frameworks and open source libraries that are available for anyone to use," says the manager.He adds: "Many of these technologies were created by the pioneers in the area, such as Netflix or Amazon, and then released to the community, but each model we used must have been trained with data from our own domain, to adapt it to our industry".The team that carries out this process of implementation, in the case of Almundo, is composed of two developers and a data scientist."The main barrier we had to overcome was the conformation of the team itself, as the profiles are scarce and with high demand in the market," says González.The results are, so far, satisfactory. "We tackled the evaluation of different aspects: in offline campaigns, the conversion was improved between five and eight times, while in the personalization of the offers it doubled", he points out.This is: the number of users who click on an advertising piece in relation to the number of impressions displayed, expressed as a percentage."In ordering and classifying hotels, all iterations are carried out with AB testing, which is why the versions that remain are those that achieve the best results measured in conversion," he points out.He concludes: "For this reason, the changes in the algorithms are made continuously, considering views, margin, conversion, customer's DNA, etc. In the first version of the hotel sorting, using customization algorithms, the conversion improved in 13 percent. "Other technologiesAs iProUP informed, in addition to artificial intelligence, online travel agencies are also betting on other innovations to offer new experiences to customers before they buy a tour package."Functions such as dimensioning the size of a hand luggage, to know if it should be dispatched, or exploring a hotel or its rooms and virtually checking its characteristics are examples in which we are working to see what value they could bring to users," he confesses. Malandrini, from Despegar.Having the possibility of knowing digitally and in a more "real" way a destination, a room or any component of the trip could be very useful in the decisions that customers have to make."These technologies allow us to extend this previous research into an experience that involves all the senses, that you can live a little while as if you were in the place you want to know," says Thomas Allier, CEO and co-founder of Viajala, iProUP.He adds that, if these tools are integrated into airports and airplanes, tourists can start their adventure from the moment they leave their home."Once the traveler landed, you can improve your experience in hotels, restaurants and during tours of the city through interaction with maps, menus and smart rooms," he summarizes.In Despegar they believe that, in addition to virtual or augmented reality, there are technologies that will gain prominence. "There are new trends that will generate a strong impact in the coming years, such as VoiceAssistant.""Through the barcode scan of the DNI, we achieved an automatic loading process for passenger data, thus using the Despegar app, customers can streamline the purchase process and prevent errors in data loading. ", they complete.The new tools of the digital economy do not stop their march or discriminate against any item. Everything is to provide customers with a better experience, in addition to offering the tourist proposal that best suits their demands. And always, before the competition does it.