Connecting AI chatbots such as Claude or ChatGPT to SolidWorks can automate CAD commands, but does this translate into meaningful engineering productivity? Leo AI argues that executing modeling operations represents only a fraction of the product development process.
Leo AI cofounder and CEO Maor Farid compares general-purpose chatbots connected to CAD software with specialized AI platforms designed for mechanical engineering.
Using Model Context Protocol (MCP) plugins, chatbots can perform operations such as sketching, extruding, and applying fillets. However, Farid argues that these integrations lack sufficient understanding of geometry, engineering standards, company-specific requirements, and manufacturing constraints.
Leo AI takes a broader approach through its Large Mechanical Model, reportedly trained on more than one billion CAD assemblies. The platform incorporates engineering guidelines, previous projects, component libraries, and industry standards to support workflows extending from initial requirements to manufacturing-ready assemblies.
Its capabilities include engineering calculations, component selection, CAD modeling, design reviews, and manufacturing preparation. Manufacturing drawing generation and finite element analysis are undergoing testing, with broader availability planned.
The company cites deployments involving Scania, Nvidia, BMW, Bosch, and U.S. military engineering teams. According to customer testimonials, Leo identified a design error in one minute that five engineers had overlooked. Another engineering team reportedly reduced three weeks of analysis to 30 minutes.
Data security is another differentiator. Leo AI reports SOC 2 Type II certification, isolated customer environments, and policies preventing customer information from being used for model training.
Farid maintains that AI adoption should prioritize measurable engineering outcomes rather than faster CAD interactions. However, the comparison reflects Leo AI’s commercial perspective, and its performance claims require independent validation.
The article highlights an emerging distinction in engineering AI: automating software commands versus supporting the broader technical decisions required to develop reliable, manufacturable products.
