What is AI at Nomadia?

Artificial intelligence is making its way into all aspects of human activity. What about at Nomadia? Franck Laugier, Data Scientist responsible for AI development at Nomadia, explains how the company integrates AI advancements to enhance team efficiency and improve customer performance.

Franck LAUGIER, Data Scientist responsible for AI development at Nomadia

• What is the current position of AI within the organizational structure of the Nomadia Group?
• What challenges are inherent in deploying AI in companies?
• How can we ensure that an AI integration project is successful?
• Do Nomadia’s software products incorporating AI require users to acquire new skills?
• Artificial intelligence fascinates, but also raises concerns and fears. To what extent are these fears justified?
• What precautions does Nomadia take to maintain full control of its AI tools?

What is the current position of AI within the organizational structure of the Nomadia Group?

Nomadia has logically placed artificial intelligence within the “Core” R&D division, which serves the other six R&D divisions. Unlike the other divisions, the one responsible for AI does not work on a specific product, but on modules designed to be integrated into Nomadia’s various products and services.

This cross-functional positioning makes it clear that AI is not a product – and certainly not an autonomous one – but a set of techniques and tools that we use both to optimize our internal processes and to enrich the functionalities of Nomadia’s software and the user experience. This enrichment is achieved by integrating AI components into existing products with two primary objectives: first, to minimize low-value-added activities, and second, to help users make better decisions daily, whether tactical, strategic, or purely operational.

What are the challenges inherent in deploying AI in companies, and more specifically in a group like Nomadia?

rom a general perspective, companies are all required to question the role and place of AI within their activities and organization. The first challenge is not implementation, but understanding what artificial intelligence actually is today. This understanding is essential to dispel the myth of “magical” AI or “one-size-fits-all” AI and focus on the most relevant use cases for each company. Nomadia has approached this similarly: we’ve identified over 50 use cases that helped us build a roadmap aligned with our business priorities and customer expectations.

The second challenge for all companies is moving from intention to actual implementation and deployment of AI products. Since machine learning algorithms became widely accessible, many studies show that 50% of internal AI/data projects fail to move beyond the prototype stage, either due to lack of resources and skills, or because of disappointing results. We’re no longer fully in that situation, mainly because AI is increasingly entering companies through software like ours. As a software vendor, we integrate AI components into our products that allow our clients to work faster and more effectively, without requiring specific AI skills.

The difficulty for us as a business software provider is that we don’t start from scratch. We’re not building AI “out of the blue.” We introduce AI components into existing products and ecosystems, which requires collaboration among all technical teams at Nomadia – developers, hosting, data teams, etc. – to create an integrated solution that is both stable and simple for clients.

How can we ensure that an AI integration project is successful?

For Nomadia, as for any company, we can never be entirely certain that an AI project will deliver the expected results. This highlights the importance of conducting a feasibility study for each project and problem, which, in addition to validating technical feasibility, must answer two critical questions to justify the investment: How much will it cost versus what it will save? And within what timeframe?

Once this proof of concept is validated, several iterations to refine the models are generally required.

I emphasize that there’s nothing magical about AI. It’s simply statistics and mathematics. As impressive as it may seem, a generative AI like ChatGPT is just statistics. But one of the strengths of current AI, thanks to neural networks and learning algorithms, is the ability to ingest billions and billions of data points to produce this statistic faster than millions of computers and billions of human brains could. This is how we can obtain predictions almost instantly that would take statisticians 100 years to calculate!

This also means that for an AI project to succeed, certain competencies must be brought together, which not all companies can afford but which are essential for a software vendor. I’m thinking of data scientists, whose key skill is knowing which algorithm can address a particular problem. I’m also thinking of data engineers. They play a crucial role because they’re the ones who will integrate data science into existing ecosystems to deliver effective, seamless solutions.

Do Nomadia’s software products incorporating AI require users to acquire new skills?

Integrating AI extends the capabilities of our software but doesn’t require any particular skills from the end user. Automatic correction of intervention reports using generative AI, time prediction for interventions, or recommendations for relevant actions in a given business context are all there to assist users and save them time.

The same is true internally, for example, with the AI-based search tool we created for our support teams. Where it once took several minutes to find the right information, this system searches all the knowledge gathered by the support team over decades in a few fractions of a second, pulls up similar tickets, and ranks the solutions provided. It’s a time-saving measure that translates to hours for the teams, and also improves the relevance of the responses. AI is here to make things easier, faster, and more efficient, not to complicate them! That’s the approach we take, at least.

Artificial intelligence fascinates, but also raises concerns and fears. To what extent are these fears justified?

Given its ability to process billions of pieces of information efficiently, fears can be justified. One example is the surveillance of our public spaces through video. From algorithmic video surveillance for “the right cause” to a generalized social control that threatens individual freedoms, it’s unfortunately just a small step.

Another major risk concerns the truthfulness of information. Social media is already full of images and videos created by generative AIs. Whether it’s texts or images, how will we differentiate real information from that generated by AI for manipulative or criminal purposes in the future? This is a question that currently has no answer.

We must therefore be extremely vigilant and adopt behaviors that minimize risks. For example, avoid including personal data in your conversations with generative AI models, and if you use such tools in the workplace, be sure not to disclose strategic information or data subject to GDPR.

What precautions does Nomadia take to maintain full control of its AI tools?

This is an extremely important point. For reasons of confidentiality and independence, our first rule is to do everything we can with open-source tools and internally.

Next, to train our AI models, we work with anonymized data sets, which avoids non-compliance with GDPR and addresses concerns from our clients who are particularly focused on the security and confidentiality of their data.

Finally, we do not host anything intended for the European market outside of the European Union. By adhering to these rules, we can provide guarantees to our clients and continue offering solutions that leverage AI wherever it can simplify, accelerate, and optimize what their teams do.