- AFTM – Jérôme Bonnepart (Regional AFTM Delegate for Auvergne-Rhône-Alpes and Travel Manager at Arkema)
- CDS Group – Ziad Minkara (General Manager)
- DIMO Software – Jérôme Alla (Notilus Pre-Sales Manager)
- ESCAET – Julie Panadero (Professor)
- INSA – Vasile-Marian Scuturici (Associate Professor)
- Travel On Move – Guillaume Poulain (Editor in Chief)
Summary: In the business world decision-making processes have become extremely fast-paced and the entire software chain is being disrupted. Missing an A.I. cycle means exiting the market, which is what happened in the cases of Kodak and Nokia. Likewise, heavy industry’s implementation plans, to be rolled out over 15 years, are no longer fast enough. However, creating a strategy based on Artificial Intelligence leads to many complications: how to find the right time, the right investment, the right financing AND the right training program? This workshop focuses on the issues of data protection (GDPR), behavior simulation, decision-making aids, machine learning and expected productivity and efficiency improvements. Does Artificial Intelligence scare us?
GDPR:how to manage data?
When speaking of Artificial Intelligence the management of users’ data must be taken into account, especially following the GDPR law that came into force in May 2018. Jérôme Bonnepart (Arkema) found it very difficult to put this management into action: ‘We had to identify all the data used by the partners, no matter what tools were being used, and not only regarding travel. It was also necessary to make sure that our subcontractors were correctly managing the travelers’ data by securing the data themselves or ensuring their own subcontractors were doing the same. This covers many tasks, such as ERP, payroll, temporary employment management, banking data, etc). Proving that the data is always under control is very challenging. From November 2017, we worked with legal and judicial entities, as well as with the HR, purchases, and business departments. It became necessary to add security, traceability and destruction clauses to our supply contracts. The threat of being penalized made everybody get involved’.
A moral stance is what led legislators to promote awareness within companies. Julie Panadero (ESCAET) has put data at the center of her training program. According to her, the volume of data, the power of calculus and the cloud are what have led us to our current situation. ‘HR and IT departments must take responsibility, within the company, for the storage and use of data. Then again, the traveler often wants a highly customized product, and that is where the paradox lays. Should we ask for no information and offer a more basic product?’
GDPR introduces two fundamental notions to protect Europeans: the right to oblivion and the right to portability from one software company to another. Europe is behind the United States and China with regards these notions. The software company will have to return the data to the person who provided it and there is great pressure on these software providers to develop the ability to stock, encrypt and transfer data while respecting complex regulations that require big investments. Businesses, on the other hand, do not invest. Therefore, responsibility is being transferred while the objectives are not the same. As well as the client-supplier relationship, there is the question of the relationship between natural person and legal person; does the data belong to the traveler-contributor or to the company?
There are currently companies that are creating ‘fake data’ so as to invent behavior models for users travelling to China or the United States, for example. This is done because the data cannot be bought in these countries. The moral of the story is that data is the new black gold: whoever knows how to manage it will be rich tomorrow!
Machine learning, travel and prediction
Vasile-Marian Scuturici (INSA) explains that artificial intelligence can have a strong societal impact, specifically in the medical field. ‘A quarter of women are concerned about breast cancer. A radiologist needs to be able make a very detailed reading of the radiology test in order to give a diagnosis. In France, practitioners cannot provide an immediate answer; instead they must send the tests to an expert who can confirm or dispel the diagnosis, and then a third opinion is required. The lack of expert radiologists in France (there are fewer than 10) means that it takes between 3 and 6 months to receive a diagnosis, and the experts that do exist only have a few seconds to give the diagnosis, taking into account that approximately 10% of the cases are difficult to interpret. The role of the IT tool is to assist the radiologist by linking the image to other data thanks to algorithms created by machine learning’.
Guillaume Poulain (Travel On Move) mentions the case of IBM Watson: ‘This tool to which we pose questions in natural language is capable of searching through millions of websites, using the elements it collected from various sources to create a response, and producing the answer quickly – less than 3 seconds for 20 000 documents- but it is only an aid’. He explains that predictive tools began to appear in the travel world quite early in order to manage the prices of flights, for example, via machine learning and complex algorithms. ‘Flyer, the Californian company, predicts the price of an airplane ticket for a specific day and time with a precision of 90 %. It is interesting for companies with many travelers who can save up to 2 % per year. The prediction is equally interesting for the buyers and sales people, as an aid in decision-making. The business will therefore be successful partly thanks to prediction and algorithms’.
Jérôme Alla (DIMO Software) says that predictive tools allow him to anticipate cash flow needs and tariff negotiations. But Jérôme Bonnepart (Arkema) thinks there are limits: ‘When, for example, a newspaper sends their special correspondents to the field at the last minute, this is by definition an unpredictable situation. In my field of work –chemistry– we can predict pricing by analyzing the company´s usual volume and frequency, but we do not necessarily know the destinations’. According to him, companies do not have great knowledge about these tools yet or do not use them wisely. In the short-term, Artificial Intelligence will serve as a counter-power for airlines. It will accompany the traveler in order to predict his or her behavior: ‘For example, when somebody books a night at a hotel, a booking that includes access to a gym but not breakfast will be suggested to a specific profile, according to his or her behavior. Either way, businesses will come up with new methods in the mid to long term’.
Julie Panadero (ESCAET) thinks the solutions are as of yet too few and unknown. Beyond pronouncing the ‘buzzword’, who is capable of defining the term, of showing what it can specifically be applied to?
Productivity improvements and cost reduction
In the words of Ziad Minkara (CDS): ‘regarding predictive technology, a tool is of interest from the moment humans are the ones to write its rules. ‘Learn-to-learn’ means that, as of now, the machine is the one that learns. Previously the software company checked the user’s profile, the company’s regulations, and the traveler’s frequency and behavior. The next step was to try to produce the perfect solution for the entire chain: data was collected, objectives were assigned and results were produced. Now we no longer give the machine rules to follow, instead we give it examples. The machine creates the rules, and that leads to a problem with regards decision-making. Development becomes infinite and is no longer controlled by the company’. He adds: ‘by this I am referring to the airline or hotel industries, where prices vary constantly. Cost reduction is currently the main global objective. Software companies are investing in A.I., something which has been happening since 2011 with rules regarding travel policy, thresholds, geo-localization…but the machine is not yet the one to impose its choices through learn to learn. The software company’s aim is to satisfy the customer who wants to reduce their spending. Moving towards A.I. still scares us a bit’.
As to productivity improvements, Jérôme Alla (DIMO) highlights that A.I. can allow cost reductions in the management of tasks such as expenses reports, which is a time consuming task. Vasile-Marian Scuturici (INSA) explains: ‘OCR allows us to extract an image or text thanks to algorithms created by machine learning. In practice, the source image is not always of good quality: a train ticket travels too and can therefore be damaged, or certain characters can be concealed. OCR is no longer really a part of Artificial Intelligence. What we do not know how to do is related to A.I., apart from the recognition of handwritten text. The difficulty has now moved on to processing the characters and to replacing the missing information (such as the price). But how can we know if we are processing a train ticket and not a plane ticket? OCR in itself, however, no longer represents an obstacle’.
What about tomorrow? How will the offer be customized?
The role of the tool
Increased productivity in terms of ROI is hard to quantify for companies that want to use an A.I.-based solution. Jérôme Bonnepart (Arkema) outlines a series of important criteria, such as travel time, the traveler’s comfort and stress levels. ‘If we divide the time spent on calculating expenses reports into three, we will see that it is huge! This task is carried out on the train or in the hotel, during what we could call concealed time. We no longer need to be at the office. Then, we have the processing; there exists a risk that VAT errors may occur, but overall the company saves time. Often purchases are what can evaluate solutions’ profitability and create work hypotheses and establish rates. It becomes easier when the tool is rolled out in all fields and users make it their own. A.I. without the support of the software company is not very useful. There is a risk that it may be misused, which in turn leads to the risk of no one benefitting from it’.
Guillaume Poulain (Travel On Move) brings up the notion of ‘destructive creation’ in A.I.: ‘This is the first time that a technological development targets the intellectual domain. Will we now need internal accountants or lawyers because the machine may not fulfill its tasks? Internal corruption will be eliminated thanks to strict controls. Productivity will grow, but mostly in very developed positions’. Indeed, many businesses use shared service centers for compatibility, training, pay, etc. After having been centralized, they are now exported and processed abroad! Spending will be more controlled, but there will come a time when this control will be carried out solely by Artificial Intelligence. Should we be scared? Julie Panadero (Escaet) thinks that A.I. should allow us to re-focus on value-added tasks that cannot be carried out by the machine: ‘Now is the time for professions to focus on the essential’.
The role of chatbots
Referring to Artificial Intelligence is more and more becoming equivalent to referring to offer customization. How trustworthy are chatbots in the travel context? According to Guillaume Poulain (Travel on Move) ‘chatbots are an epiphenomenon and will not replace a travel agent or a TMC. They allow us to save some time on frequently asked questions and by automating tedious tasks, but they will soon be forgotten’. The Sam application signifies to him the world of the future: a personal virtual mobile assistant –much more powerful than Siri– that will allow us to exchange using natural language. He predicts that ‘we will no longer download apps; instead we will download vertical assistants that will provide solutions to situations that the traveler has not yet conceived. The assistant will be able to predict a delay due to a meeting, launch a follow-up action such as sending an alert and buying a ticket’. He estimates that good technology must be transparent, but must also respond to real life needs.
But does the autonomy gained on the one hand not suppress the will to learn (a modern language, for example)? A.I. certainly lacks empathy and creativity. But if it allows people to enter into dialogue, to become closer, this will be a valuable addition that we should not be afraid of. What we used to see as science fiction will become a business plan!
Let us be optimistic! A.I. will bring people closer together and create more space for emotional intelligence to develop! The closing sentence pronounced by Julie Panadero (Escaet) was: ‘A.I. will be able to create emotion and will provide human satisfaction in line with human needs’.