For two years, companies have focused their efforts on a single question: which artificial intelligence model should we choose?
ChatGPT, Gemini, Claude, Mistral… Comparisons have multiplied. Executive teams have evaluated performance, tested conversational assistants, and begun integrating AI into their processes.
Yet this question is becoming secondary.
The real transformation is no longer solely about models. It now concerns the infrastructure that will enable building, training, governing, and evolving them.
Project Tapestry, unveiled by the AI Alliance with Yann LeCun as Chief Science Advisor, perfectly illustrates this evolution. This project doesn’t propose a new chatbot to compete with ChatGPT. It opens a much deeper reflection: can we build artificial intelligence differently than through the concentration of data, infrastructure, and decisions in the hands of a few players?
For leaders, this question goes far beyond technology. It directly touches their company’s governance.
Model Selection Becomes a Strategic Decision
Since the arrival of generative AI, many organizations still think as they did when adopting new software: they compare features, negotiate licenses, and select a vendor.
This approach is now showing its limits. Choosing an AI model now means choosing a technological dependency, a data management policy, a governance framework, and an evolution capacity. In other words, the choice is no longer just about IT. It becomes strategic.
Leaders must now ask themselves a different question:
What Will Happen If Our Vendor Changes Its Terms of Use, Pricing, or Access to Certain Features Tomorrow?
This concern goes far beyond Project Tapestry. It affects all companies that are gradually building their processes on artificial intelligence they don’t control.
A Third Path Emerges
Today, two models dominate. The first consists of using platforms developed by major international labs. This approach offers remarkable performance and rapid implementation. In return, it creates dependency on a single vendor.
The second consists of developing one’s own sovereign model. This strategy offers more control but requires considerable investments in computing, data, and research.
Project Tapestry proposes a third path. The idea is to allow multiple organizations to build a common foundation together, which each can then adapt to their own needs. This collaborative approach doesn’t challenge sovereignty. On the contrary, it seeks to make it economically accessible by pooling some of the investments. It’s therefore not just about technological innovation.
It’s a new way of organizing the production of artificial intelligence.
Data Becomes the Company’s True Asset
This evolution confirms an underlying trend. General-purpose models are gradually becoming accessible to everyone. However, business data remains unique to each company. That’s what will make the difference tomorrow.
Companies capable of documenting their processes, structuring their knowledge, and organizing their governance will have a lasting competitive advantage, regardless of the model used.
Conversely, organizations that still consider their data as a mere byproduct of their activity risk gradually losing their ability to create value.
The real strategic asset is no longer just artificial intelligence. It’s the quality of the information that enables it to produce relevant decisions.
Governance Becomes a Competitive Advantage
Artificial intelligence is gradually ushering companies into a new phase of their digital transformation. For a long time, governance mainly concerned data.
Tomorrow, it will also need to integrate models, AI agents, technological dependencies, and responsibilities associated with automated decisions.
The questions are numerous. Who decides on model selection? How do we ensure the quality of data used? How do we avoid excessive dependency on a vendor? How do we ensure traceability of decisions produced by an AI agent? How do we maintain the ability to change technology if the context evolves?
These questions now directly concern executive management.
What a Leader Can Do Today
Companies don’t need to wait for Project Tapestry’s maturity to act. However, they can already take several concrete actions. Map their dependencies on AI vendors. Identify the data that constitutes their true competitive advantage. Build architectures open enough to replace a model if necessary. Establish AI governance involving executive management, business units, IT, and legal functions. Finally, consider that choosing artificial intelligence is now as much a matter of business strategy as it is of technology.
A Battle That Goes Far Beyond Project Tapestry
The significance of Project Tapestry may not lie solely in its future success. It reveals above all that the industry is entering a new phase. Yesterday, competition was about models. Today, it extends to infrastructure, data, governance mechanisms, and the rules that will organize their development.
For leaders, the question is therefore no longer simply which conversational assistant to adopt. It’s about determining what technological foundations they want to build their company on for the next ten years.
This is probably where the next great battle of artificial intelligence will be fought.
