At the Tech For Industry Show, a panel offered a deliberately broad take on industrial intelligence, featuring Sébastien Ory, President (AVEVA France), Ilaria Michelizzi, Southern Europe Presales Leader (AVEVA), Philippe Mège, Head of Digital Service Factory (Axens), Olivier Traore, industrial leadership lead (Accenture) and Laurent Champaney, CEO (Arts et Métiers ParisTech). The central message: industrial AI is just one component among others. Data quality and the ability to integrate frontline teams' know-how are just as decisive for a successful industrial digital transformation.

For Aveva, industrial intelligence refers to an organization's ability to gather and unify its data — engineering data, operational data, maintenance data — into a centralized system, in order to speed up decision-making at every level of the company and collaborate within a connected ecosystem.

Aveva published, in partnership with IMD Business School, a first Industrial Intelligence Report based on interviews with 275 executives. The main takeaway: industry's real challenge isn't so much access to the right resources as accelerating data sharing and its interconnection across the ecosystem.

Axens, which supervises and optimizes industrial units for refiners such as Total, BP and ExxonMobil, details a three-step method.

Data collection relies notably on OSIsoft's PI system (integrated into the Aveva ecosystem) and on Connect Data Service, presented as a key element for retrieving reliable, real-time data.

According to Axens, the most critical step isn't retrieving the data but making it usable: contextualizing it (which equipment, which process does it relate to?) and checking its consistency — for example, a temperature rise in a reactor should be consistent with a higher hydrogen flow rate. An AI layer automates this consistency check and discards data deemed unqualified.

Axens stresses the importance of adding a physical layer to purely statistical models, by building hybrid models: a thermodynamic model is continuously fed with real-time data, producing a true digital twin that is constantly recalibrated. This foundation then makes it possible to connect an action layer (advanced process control) to the customer's control system.

Sharing industrial data with a third party has long been a barrier. Axens describes an evolution over roughly ten years: from uncontrolled sharing, to strong reluctance, to a framework that is now more fluid, provided platforms meet expected security standards (ISO 27001, SOC 1/SOC 2). Industrial cloud is now widely accepted under these conditions.

For Accenture, connecting IT and OT isn't just a matter of technical interfacing: it's a transformation program involving different cultures, with distinct priorities and risk profiles (data accessibility and confidentiality on the IT side; physical safety and production continuity on the OT side).

Accenture observes a frequent tension between central management pushing for standardization to cut costs, and local sites seeking to preserve their autonomy and business context. The recommendation is a federated approach: the central level lays the foundations (standards, cybersecurity, platforms like Aveva Connect), while use-case prioritization remains at the local level.

One of the most concrete contributions mentioned: AI can be used to capture the tacit know-how of experienced employees and speed up the upskilling of younger ones — not replacing the expert, but freeing them up for higher-judgment tasks. The identified risk is forced adoption happening faster than the organization can truly absorb, which is why the explainability of AI-produced results matters.

Accenture cites four areas of benefits observed among companies engaged in this transformation: improved predictive maintenance, better resilience through a more predictive than reactive posture, gains in energy efficiency and sustainability, and faster scaling of use cases once the data foundation is in place (example cited: the Total Digital Factory, developed with Accenture).

For Arts et Métiers, the training question illustrates the same IT/OT tension: on one side, engineering schools historically centered on materials, processes and production systems; on the other, management schools more focused on data and decision-making.

Arts et Métiers chooses to keep teaching the fundamentals (materials science, manufacturing processes, sensors) while incorporating data quality and hybrid physics/AI models, notably through student projects run on the school's real industrial platforms (machine tools, forges, foundries, polymers and composites).

The school develops partnerships with companies to offer student projects grounded in real industrial cases, as well as post-graduate training built with professionals. One point of caution shared by the panelists: ongoing training for employees already in post remains an area still under-invested by companies, compared with management training.

In conclusion, each panelist summed up the key success factor in their view:

Industrial intelligence isn't just about adding an AI layer: it rests on quality data that is governed and contextualized, on successful IT/OT convergence, and on a collective ability to pass on and evolve industrial know-how. It's this combination that lets companies move beyond the pilot stage and create measurable value at company scale.

Watch the full panel, along with the other Tech For Industry Show replays, on our dedicated page.