Scaling Industrial AI in Manufacturing
Indian manufacturers are investing in artificial intelligence (AI) and Virtual Twins, but moving these technologies from pilot projects to production remains a challenge [Reported/Alleged]. For Dassault Systèmes, the next phase of industrial AI adoption will depend on combining industry-specific knowledge with unified data platforms, protecting intellectual property (IP), and connecting digital skills with practical manufacturing experience [Stated].
Ravikiran Pothukuchi, Director, Enterprise Applications, Dassault Systèmes, and Azmathullah Mohammed, Senior Sales Director, Dassault Systèmes India, discuss why generic AI models may fall short in industrial settings [Reported/Alleged]. They explain how Virtual Twins can improve product development and operational efficiency, and what Indian manufacturers need to scale adoption.
Pothukuchi notes that manufacturing companies are conducting validation, pilots, and proofs of concept (POCs) while engaging with partners, but many are not yet putting solutions into production [Stated]. He asserts that AI still has to evolve to become a key aspect of the overall product development lifecycle [Stated].
The Limitations of Generic Models
Dassault Systèmes has taken an approach of developing native industrial AI to help customers incorporate industry-specific knowledge and know-how [Stated]. The company utilizes Virtual Companions to leverage industry-specific outcomes and objectives, aiming to make AI more industrialized and intelligence-based [Stated].
Mohammed argues that pilots will not succeed unless businesses identify a business problem, scope it, and determine how to leverage AI [Stated]. He states that running pilots simply to become familiar with AI is important but must eventually reach the next stage of production [Stated].
Mohammed describes generic AI, particularly large language models, as "librarians" that know a lot because they can access knowledge available on the web and elsewhere [Stated]. He claims the industrial world requires not just knowing, but doing, noting that actual intellectual property is captured by how organizations do things in the real world [Stated].