How does generative artificial intelligence fit in with Cegid Notilus' technological solutions?
Laurent Lassure - It changes the game by bringing this notion of creation. AI is starting to arrive on the mission order generation and, consequently, the related expense reports. This is on the Cegid Notilus roadmap, and it's something that's going to arrive later this year, this summer I think. Let's take the example of a traveller who is used to booking his hotel on Booking or his train ticket on SNCF Connect. Well, the Notilus AI, having analysed their email inbox and detected a train ticket and a hotel night in Lyon on such and such a date, will be able to suggest creating the travel order for this journey. While a travel order is compulsory in the public sector, in the private sector it is one of the administrative burdens that companies avoid, contenting themselves with travel information. The benefit of a mission order automatically fed by AI offers the hierarchy the possibility of giving official validation upstream and get an idea of the cost of the trip. This means that, if necessary, travel arrangements can be adjusted if the fares charged on a given date are too high. In addition to this budgetary aspect, knowing where the employee is is also important for safety and duty of care reasons.
But is AI something new in this universe?
L. L. - Well, not exactly. For example, taking a photo of a receipt to enter an expense claim, it was already artificial intelligence. But AI is now enabling us to go even further. At the beginning of the year, we replaced the old OCR functionality used by Cegid Notilus for many of our customers. Previously, items were identified using character recognition and machine learning. Thanks to developments in AI, we can now put in place LLMs, broad language models, to analyse the content of a ticket in greater depth. Previously, a VAT value of 20% was used to estimate alcohol consumption. Now we can actually read the contents of the receipt, picking out the word beer or the name of a well-known brand. It's up to the company to decide whether or not to authorise reimbursement, but it's important to be able to highlight this information. In the same way, artificial intelligence was already there for a number of years in fraud controlAI is used to verify values. Today, AI makes it possible to carry out more detailed analyses, to check whether the traveller was on holiday when the expense claim was submitted or, for a full tank of petrol, to compare the number of litres with the category of vehicle. An 80-litre tank of petrol for a Clio makes you wonder.
What other developments are in the pipeline at Cegid Notilus?
L. L. - From June onwards, we will be upgrading the reporting section, integrating artificial intelligence into the data analysis process. This will enable us to generate dashboards and reports in natural language. Through a prompt, in the same way as in a GPT chat room or a Gemini. For example, you can ask the solution to identify the company's ten biggest travellers. From there, the AI will fetch the information for each employee, the amounts of their expense claims and build the report. The result is available in three seconds, without having to go and find the information yourself, then generate an Excel file, then create a pivot table. As a corollary benefit, users are no longer obliged to know the data model or which field to find which information in. By requesting an analysis based on the amount before tax, the latter will be directly retrieved.
Hence a multitude of possible applications, I suppose?
L. L. - The idea is to be able to simply produce all tables or graphs The more in-depth the analysis, the more precise the questions that can be asked. The deeper we go into this analysis, the more precise the questions we can ask, for example, how many employees deserve to have a Liberté card, based on the number of trips they make.
In short, it's a travel budget analysis for everyone...
L. L. - Managers can query the solution directly. However, while the AI is supposed to have access to all the information needed to respond to requests, it does not give everyone access to everything. Each user can query it according to the scope of their rights, in compliance with RGPD regulations. But, in fact, a manager will be able to have information about his team, find out who has a trip planned on any given day, and estimate whether he is in control of his budget. Travellers themselves can also use it. As we are in the process of recovering more and more carbon footprint data by integrating the reservations section of the travel agency or SBT, travellers can even monitor their personal footprint, and find out how far they have travelled by train or plane. This is information that we already have in Notilus, but we are making it easier and more flexible to access.
After mission order generation and reporting, in what other areas do you plan to integrate artificial intelligence?
L. L. - These first steps will enable us to move on to other stages. The next one, which is already in the pipeline, concerns travel policy, with a view to resolving one of the major shortcomings in this area at present, its communication. Not everyone knows where it is, or which corner of the server it is in. Sometimes, a traveller may be surprised to be reimbursed for only part of their meal expenses because they weren't aware of the existing ceilings. Hence the desire to work on communicating these elements, initially a little like a chatbot that you can ask yourself to find out about travel policy and reimbursement limits. The first chatbots for travellers should be available from the end of this year.
Could this information be delivered more proactively?
L. L. - Eventually, AI will be able to push out this information without having to ask for it. By analysing the context, because it knows that an employee is going to Berlin, it will be able to inform him or her that, to get from the airport to the city centre, the travel policy allows taking a taxi or not. With a connection to the mission orders, using all the different information integrated into Notilus, she can call back at the end of the last appointment. train time, seat number reserved. Let's take another example, that of car fleets: by analysing your next appointments and the movements of other employees, AI will be able to suggest car pooling. Or invite you to take a free vehicle from the company fleet rather than your own car, to avoid mileage costs.
Would you say that Cegid Notilus is ahead of the game when it comes to integrating AI into its solutions?
L. L. - It's always a bit difficult to position oneself. I wouldn't say we're ahead of the game, because there are a lot of solutions out there today that incorporate AI. The difference that Cegid Notilus can make is that through its functional coverage. In other words, an expense claim pure player may be ahead of the game in terms of AI in this area, but it's not going to open up directly to the travel or vehicle fleet sector. In fact, I think that in terms of the changes we are making to reporting, we are the first to offer them in the business travel sector. With this global vision, I hope that Cegid will provide enough elements, not to revolutionise travel management, but to make it more efficient and more serene.
Let's look to the future. What new applications do you foresee?
L. L. - We're only at the beginning. In my opinion, we will see the emergence of intelligent agents who will acting for you. By identifying in emails that a traveller has an appointment in Strasbourg at such and such an address, they will be able to ask travellers if they want them to book a train ticket and the nearest hotel for them. I'm getting ahead of myself by a few years, but I think these are the trends of the future.
