Dr. Maor Farid, Co-Founder & CEO at Leo AI
Dec 31, 2025
Everyone's talking about AI in manufacturing. VCs are pouring money into it. Jeff Bezos just launched Project Prometheus, a $6.2 billion AI venture focused on engineering and manufacturing for aerospace, automotive, and computing. Market leaders like Dassault Systèmes, Siemens, and PTC have been announcing at their annual conferences that "AI will revolutionize how products are made."
But for many in our industry, there's a lingering feeling that AI in manufacturing is a bubble. Big promises, impressive demos, but no real change on the horizon.
So which is it?
The Gap Between Promise and Reality
We're seeing a genuine surge in interest from manufacturing companies across industries and geographies. They want to design faster, get products to market quicker, and reduce costly errors. However, the results from major CAD vendors have been... underwhelming.
Take Dassault Systèmes. At their 3DEXPERIENCE World 2025 conference in Houston (February 2025), they announced AURA, their AI assistant for SolidWorks.The promise? A revolutionary tool that would transform how engineers work.
The reality as of December 2025?AURA is still in beta, available only on the 3DEXPERIENCE cloud platform, which serves roughly 1% of their customer base. The 99% still using the desktop application? They're still waiting. And the functionality so far is limited to text-based chat that answers questions about how to use SolidWorks. That's a far cry from understanding your designs, your standards, or your company's best practices.
Why Manufacturing Is Different
Meanwhile, AI has already transformed workflows in other industries.Software development has GitHub Copilot and Cursor. Legal has Harvey.ai. These tools work because those industries operate in the textual domain. Code and legal documents are text. They can plug directly into LLMs (Large Language Models) like ChatGPT, Claude, or Gemini with a relatively low technical barrier.Manufacturing is different.
In our world, the most critical data type isn't text. It's CAD: 3D models, geometry, assemblies, tolerances, constraints. Generic AI tools simply cannot read or understand CAD files. They can summarize a document or help you write code, but they can't interpret an assembly, check a tolerance stack, or understand why your team prefers a specific vendor's components. Mechanical engineering demands spatial reasoning and specialized knowledge that general-purpose language models don't have.
There's also the cost of mistakes. In software, you can push a fix in minutes. In legal, you can revise a document. In manufacturing, changing a part that's already in production can cost hundreds of thousands of dollars. A recall can cost millions. According to industry data, the average product recall costs companies between $10 million and $50 million in direct expenses alone, not counting brand damage and lost sales.
This high-stakes environment naturally makes our industry more cautious about adopting new tools.
The Trust Problem with Generic AI
Here's something that doesn't get discussed enough: in engineering, every decision should be backed by a legitimate source. Internal guidelines, engineering handbooks, industry standards. Where does ChatGPT get its information? According to OpenAI's own disclosures, about 22% of GPT-3's training data came from Reddit posts with 3 or more upvotes. The rest comes from web crawls, books, and Wikipedia.
So when an engineer asks ChatGPT "how many bolts should I use to seal a pressure vessel rated for 500 MPa?" and it confidently answers, "5 M8 bolts, no O-rings needed," that answer might be drawing from an anonymous Reddit comment. No engineering context. No verified expertise. No traceability. In aerospace, medical devices, automotive, or defense, trusting that kind of output is a serious liability.
Two Camps in the Manufacturing World
Right now, manufacturing companies are splitting into two camps:
Camp 1: Wait and See These companies are watching from the sidelines, waiting for others to figure out AI first. They tried ChatGPT, saw that it couldn't deliver engineering-ready results, and concluded that AI isn't ready for their industry. Many of these are more conservative organizations, risk-averse, or operating in less competitive markets.
Camp 2: Figure It Out Now These companies saw what AI did to software development and legal, and they're actively working to capture that advantage for themselves. They understand that being early comes with costs, but they also know that falling behind has costs too. Both camps have legitimate reasons for their position. Not every company faces the same competitive pressure. Not every product development cycle demands the same pace. But here's what Camp 1 often misses: specialized AI tools built specifically for engineering are already delivering results.
What Engineering-Specific AI Can Do Today
Leo AI, for example, uses a proprietary AI model that can actually read and search CAD files based on geometry and technical specifications. It connects to both your organization's data sources (CAD files, PDFs, internal documentation) and external engineering references, including over 1 million standards, guidelines, and technical books. Companies using Leo are already seeing:
34% reduction in design errors
30% reduction in redundant design work (because engineers can actually find existing parts and previous designs)
~18% faster time to market
5+ hours saved per engineer per week
These aren't hypothetical benefits. They're measured outcomes from companies like HP, Scania, and Intel.
The Risks to Consider
Using AI today isn't without risk.
Data Security If you use generic AI tools like ChatGPT, you risk exposing your IP. These tools may use your inputs to train their models. Engineering-specific tools like Leo AI are SOC 2 certified and GDPR compliant, providing enterprise-grade security that keeps your data protected.
Being the Avant-Garde Leading always has a cost. You're going down a road that isn't yet standard in your industry. It requires a firm decision from leadership that learning new tools and skills is critical to gaining a competitive advantage.
Think about the companies that adopted SolidWorks in the 1990s or Pro/Engineer in the 1980s. The ones who invested the time and effort won. The others - not sure if they're still with us.
The Real Decision
Before you bring AI into your organization, understand this: it's not primarily a time investment.It's a strategic decision.
Your leadership needs to decide: are we willing to lead? Are we willing to invest in training, even though AI tools aren't taught in engineering programs yet? Are we willing to be early?
Leading has its costs.
But before you answer, consider this: your competitors have probably already started.
