The thesis: the model is not enough
According to François Bohuon, Managing Director EMEA at Palantir, speaking at the World Manufacturing Forum 2026 organized in Milan by the World Manufacturing Foundation, companies are mistaken if they focus their efforts on choosing the most advanced artificial intelligence model. Many began by experimenting with large generative language models: tools that, the manager acknowledges, increased individual productivity in summarizing documents or contracts, but had a marginal impact on industrial processes and revenue.
For Bohuon, these models are a computational brain without an operational body and heart: they do not know the deep structure of the individual company and are not integrated into the systems where decisions are made. Energy, microchips, data centers and the LLMs themselves are now commodities accessible to anyone, so they do not guarantee lasting advantages.
Ontology and the Alpha factor
The heart of the proposal is the corporate ontology, described as an intelligent operating system and semantic layer of the organization: a dynamic digital twin that models plants, facilities, machinery, materials and logistics flows at maximum granularity. Through it, data from IoT sensors, ERP, MES and emails are linked to business logic and action systems. The infrastructure must unify three elements: data, logic and action, because data without logic is noise, logic without action generates information overload, and logic with action but without data produces chaos.
The central concept is the Alpha factor, borrowed from financial language: the know-how and proprietary data that make a company different and unique in the market. For executives, there are two options. The first is flattening the enterprise, that is, sending data to external models via prompts: a fast path at first, but one that commoditizes the company and eliminates its advantage. The second is integrating the models into the proprietary ontology: a more demanding path, but according to Bohuon the only one capable of generating cumulative value and preserving technological sovereignty and ownership of the Alpha.
The Spirit Airlines case and the limits of the argument
In support of his thesis, the manager cites the Spirit Airlines case: after the bankruptcy, a trove of 100 million emails, 500 million chats, pricing strategies and procedures was reportedly sold at auction for $10 million to a technology operator to train its own models, with legal transfer of the company's Alpha. Bohuon then indicates five levels of technological and operational maturity, without the available material detailing their contents, and three adoption principles: continuous agility, development backward from bottlenecks flagged by line operators, and focus on measurable economic return.
It remains to clarify the verifiable scope of some claims: the details of the five-level scale are not disclosed, the Spirit Airlines case is reported as the manager's reconstruction and is not accompanied by documentary corroboration in the text, and no examples are given of companies that have already followed the proposed path with measured results. The overall picture remains that of a position stated by a Palantir executive, which should also be read as a commercial argument in favor of a proprietary infrastructure. The thesis on governance, however, is clear: according to Bohuon, sovereign AI is not an IT project that can be delegated to the IT department, but a strategic priority for managers, aimed at amplifying human skills and elevating operators to more strategic roles.
