Interview with Antoine De DECKER, Director of INCEPTIVE

Can you describe your company’s business?  

Inceptive is an artificial intelligence engineering and consulting firm founded in November 2017. The company develops and industrializes AI models for its customers. This service offering can be applied across a company’s entire value chain, so the use cases are very varied.
In parallel with this consulting activity, inceptive has developed its own in-house tools to build and develop new AI models.
We have also designed Robo Fabrica, a SaaS chatbot platform. It works with technologies called conversational scenarios. These are not complex technologies like chatGPT, but they consume little electricity, and the risks of hallucination are virtually non-existent. This platform already handles 2.5 million conversations a year. Customers include INPI, the French Ministry of Health and an insurance broker. The chatbot solution we offer guarantees the sovereignty of our customers’ data.

Who are the key partners in your success?

The presence of the MIAI cluster (Multidisciplinary Institute for Artificial Intelligence) and ENSIMAG, the INP group’s engineering school, is essential for us and for the Grenoble Alpes region as a whole.ENSIMAG offers multi-disciplinary courses and trains the data scientists we recruit.Thanks to the MIAI and the admirable work of its director Eric Gaussier, there is a real emulation here at Grenoble Alpes, generating a large number of very high-level articles.
We have also established a strategic partnership with OCI, an ESN based in Alsace with 1,500 employees and national coverage. Although their offer is very rich in the field of IT, our expertise in artificial intelligence and development is perfectly complementary.On the strength of this partnership, we have set up a co-marketing process: when OCI identifies an AI opportunity with one of its customers, it contacts us.We are then in a position to present our AI expertise, direct the customer towards BPI France (Banque Publique d’Investissement) support schemes, and structure solid, fundable AI projects.The “BPI label” associated with these projects makes it considerably easier to obtain additional bank financing.

How would you define artificial intelligence? 

Artificial intelligence is a complex and evolving field, and we prefer to use the term artificial intelligence system. This technology has experienced spectacular growth thanks to the increase in computing power and the massive availability of data. The arrival of ChatGPT popularized it. AI is a technology that enables us to program differently. Conventional” programming transcribes known rules into coded language. AI offers a different way of programming when the rules are complex, unknown and the parameters too numerous. The AI system is asked to find the rule from training or learning data. The two main uses of AI are classification and prediction. In industry, for example, classification will be used for quality control of manufactured parts, while prediction will be used for predictive maintenance.

What are the challenges a company needs to anticipate if it is considering investing in AI-based technology bricks?

It’s essential to emphasize that the success of an artificial intelligence project relies heavily on data. Rigorous data collection and structuring are essential for reliable results.
The training data needed to train an AI system are essential elements of the model’s performance. The more data we agglomerate, the greater the predictive potential. The quality of the data supplied by our customers is another fundamental element. The data collected by companies has a great deal of value: it’s information, but it’s also the reflection of know-how.

How do you help companies meet the challenges of AI?

Bpifrance offers particularly interesting support with the IA Booster France 2030 program. This program provides 50% funding for a complete audit and implementation of the AI solution.
We start with an initial business data audit and a few days’ immersion in the company. Within 30 days, we draw up a personalized roadmap and prioritize the most promising projects according to rigorous criteria (R.O.I, data requirements, technological risk and cost). This preliminary phase brings a fresh perspective to the company and establishes a relationship of trust.
If the project is validated, we develop the artificial intelligence solution: the customer entrusts us with part of its data, we design the model, and measure its performance on the customer’s premises with blind tests (using data not used in the design of the model). If the model is reliable and validated, we integrate it into the company’s information system. These assignments last between 4 and 12 months, rarely more than a year.
In all our projects, we prioritize customization (each solution is tailored to the specific needs of each customer), data confidentiality (we implement rigorous measures to protect sensitive information), and ethics (where we ensure that AI is used in a socially responsible manner).

What are the benefits of AI for the business world? Do AI systems have more impact in some sectors than others?
Can you give some concrete examples?

We address a wide variety of sectors and players, and on very different use cases, because the benefits of AI are manifold: improved productivity and quality, reduced costs, increased customer satisfaction… AI enables companies to gain in competitiveness and differentiation, and allows public players to free staff from certain repetitive tasks to devote human time to higher value-added tasks.

In the agricultural sector, we have developed an agricultural yield benchmarking tool accessible to all members of a cooperative (our customer). Using AI models, farmers have access to customized, in-depth statistical analyses in a matter of seconds. In the Energy sector, we developed a predictive model for the State of Montenegro, aimed at optimizing the management of energy exchanges with its 4 neighboring countries, thus helping to stabilize prices and reduce the environmental impact of energy production. In the field of Education, we are currently participating in a European Erasmus + project to create an AI-based interactive tool. It features immersive role-playing games simulating professional conflict situations, with the aim of improving students’ “soft skills”. In the public sector, we have set up chatbots for town halls, enabling citizens to obtain rapid answers to the most frequently asked administrative questions, and freeing up the time of town hall staff for more qualitative support and advice…

In industry, AI is used in a wide variety of fields, from production to sales management. In particular, it can be used to implement predictive maintenance (anticipating breakdowns and optimizing maintenance operations), improve quality control (thanks to computer vision, manufacturing defects are detected in real time, rather than on batches), and optimize the use of information stored and internal to the company (large language models can be used to create knowledge bases accessible to all employees, facilitating information sharing and problem solving). AI can also be used to assist support and sales departments (AI can generate personalized summaries of customer history from e-mail exchanges, customer segmentation, automation of repetitive tasks, help with drawing up quotations, etc.), train employees (internal chatbots to answer employees’ questions and provide personalized training), create personalized after-sales service (AI generates personalized technical documentation for each product, facilitating installation and maintenance).

At a time when the protection and quality of personal data has become a major issue, what guarantees do you offer?

We ensure the confidentiality and sovereignty of our customers’ data by processing it exclusively on infrastructures located in France. Our AI models are trained on strictly anonymized data and are not reused for other projects. We have put in place a robust contractual framework, in collaboration with lawyers specialized in intellectual property, to guarantee the protection of the data and models generated. This approach, which we adopted in 2018, positions us as a benchmark player in terms of regulatory compliance and digital sovereignty issues.

We try to detect problems related to data quality or organization very quickly. This is one of our diagnostic objectives. In some projects, we have suggested adjusting collection protocols. This may involve increasing the frequency of measurements, standardizing sensors or optimizing forms: we will work closely with the customer to implement these improvements.