AI for Design Quality & DFM
Can AI handle ASME and ISO standards-backed engineering calculations? A look at where AI helps, where it fails, and how engineers keep calculations traceable.
·
⏱
9 min read
Dr. Maor Farid
Maor Farid is the Co-Founder and CEO of Leo AI, the first AI platform purpose-built for mechanical engineers. He holds a PhD in Mechanical Engineering and completed postdoctoral research at MIT as a Fulbright fellow. A Forbes 30 Under 30 honoree and former AI researcher and Mechanical Engineer in an elite military intelligence unit, Maor leads Leo AI's mission to transform how engineering teams design better products faster.
BOTTOM LINE
So can AI handle ASME and ISO standards-backed calculations? It can be a powerful assistant, but not an authority. AI is excellent at finding references, setting up problems, reusing prior work, and catching obvious errors, and it is poor at guaranteeing that a result reflects the current code and a sound engineering decision. The teams that benefit most use AI that ties its answers to real, reviewable sources, keep a qualified engineer accountable for every result, and treat verification against the standard as a required step. Used that way, AI makes standards work faster without making it riskier.
Engineering calculations are not like ordinary math problems. A pressure vessel wall thickness or a bolted joint capacity is not just a number, it is a number that must trace back to a recognized standard, use the right allowable values, and survive review by someone who can be held responsible if it is wrong. So when engineers ask whether AI can handle calculations backed by ASME and ISO standards, they are really asking whether AI can be trusted where the cost of an error is measured in safety, not style.
The honest answer is nuanced. AI can accelerate parts of the calculation workflow and reduce certain mistakes, but it cannot replace the standards, the engineering judgment, or the accountability that codes require. This article looks at where AI genuinely helps with standards-backed calculations, where it falls short, and how to use it without compromising rigor.
Why Standards Make Calculations Hard for AI
Codes like the ASME Boiler and Pressure Vessel Code are large and precise. The ASME code spans 11 sections with many divisions and subsections, and it prescribes specific approaches to design calculations, material allowables, fabrication, inspection, and documentation. ISO standards add another layer, governing everything from general tolerances to quality processes. A standards-backed calculation is correct only if it uses the right edition, the right clause, and the right inputs.
This is exactly where a general AI model is weak. A large language model generates the most likely text, which can look like a correct calculation while quietly using an outdated allowable, the wrong safety factor, or a clause that does not apply to the case. The structure of standards also matters: ASME Division 1 relies largely on standard design rules and manual calculations, while Division 2 permits non-standard designs verified with finite element analysis. An engineer knows which path applies. A generic model does not inherently know, and may blend them.
There is also the question of inputs. A calculation is only as good as the material properties, load cases, and boundary conditions behind it, and those often live in datasheets, prior reports, and supplier documents rather than inside the AI model itself. An engineer who pulls the wrong yield strength will get a wrong answer no matter how good the method is, and a general model has no reliable way to know which datasheet applies to your part.
The lesson is not that AI is useless here, but that calculations governed by codes demand traceability that generic generation cannot provide on its own.
IN PRACTICE
The search in Teamcenter has always been a weak point for us. If you don't know the exact part number or file name, you're basically not finding it. Leo changed that. I can describe a part geometrically or by function and it finds relevant parts from our own history, not just from an external catalog.
Verified User, Defense and Space Enterprise
Where AI Genuinely Helps
Used carefully, AI removes a great deal of friction around the calculation itself. The value is less in producing the final number and more in everything that surrounds it: finding the right reference, setting up the problem, and catching obvious errors early.
Several uses hold up well in practice:
Finding the right reference. AI can quickly point an engineer to the relevant standard, prior calculation, or internal guideline, which is often the slowest part of the task.
Setting up and explaining. AI can lay out a calculation method, list required inputs, and explain a clause in plain language, which speeds up less experienced engineers.
Reusing prior work. Most calculations resemble ones a team has done before, so surfacing a vetted past calculation is faster and safer than starting fresh, a benefit related to broad engineering part reuse.
Sanity checks. AI can flag results that fall outside expected ranges, catching transcription and unit errors before they propagate.
In each of these, AI assists the engineer rather than replacing the standard. The engineer still selects the method, confirms the inputs, and owns the result, which is exactly how it should be for code-governed work. Our broader overview of AI for engineering calculations goes deeper on these workflows.
Where AI Falls Short
The failures matter as much as the wins, because in standards work a confident wrong answer is more dangerous than an obvious gap. There are a few areas where AI should never be trusted blindly.
First, AI can fabricate. A model may cite a clause that does not exist or apply a formula from the wrong context, presenting it with the same fluency as a correct answer. Second, standards change, and a model trained on older material may not reflect the current edition. Third, AI does not carry accountability. A code calculation often needs a responsible engineer to stamp it, and no model can take on that legal and professional responsibility.
These limits are why verification is not optional. The same discipline that catches design problems before they reach the floor, discussed in our piece on design for manufacturability, applies to calculations: every AI-assisted result must be checked against the actual standard and an engineer's judgment before it is used.
How to Keep AI-Assisted Calculations Traceable
The way to get AI's speed without its risks is to insist on traceability at every step. A traceable calculation can be followed from the result back to its inputs, its method, and the specific standard it relies on. That is achievable with AI as long as the workflow is set up correctly.
Four practices make the difference:
Prefer retrieval over generation. Use tools that point to a real standard, a real prior calculation, or a real internal document rather than inventing an answer from nothing.
Record the source. Capture which edition and clause were used, so a reviewer can confirm it independently.
Verify against the standard. Treat the AI output as a draft to be checked against the code text, never as the final authority.
Keep the human in the loop. Ensure a qualified engineer reviews and owns the result, especially where a stamp is required.
This is where an AI intelligence layer that sits on top of a team's own data is more useful than a general chatbot. Leo AI connects to the engineering knowledge a team already holds, so it surfaces prior calculations, parts, and documents from the company's own history rather than guessing. Tying answers to real, reviewable sources is what makes AI compatible with standards work, and it is closely linked to good engineering knowledge management.
A Practical Workflow for Standards-Backed Calculations
Putting it together, here is a workflow that captures AI's speed while respecting the demands of ASME, ISO, and similar codes:
Define the problem and identify the governing standard and clause before touching any tool.
Use AI to retrieve the relevant standard text, prior calculations, and required inputs, and to explain anything unfamiliar.
Perform or reproduce the calculation, using AI for setup and sanity checks rather than as the source of truth.
Verify the result against the standard and against a related calculation, such as a tolerance stack-up analysis where dimensional limits interact.
Document the method, inputs, edition, and clause, and route the calculation for review and sign-off.
This workflow keeps the engineer firmly in control. AI compresses the slow, repetitive parts, while the standard and the responsible engineer remain the final authority on whether the calculation is correct.
FAQ
Calculations Backed by Real Data
Tie AI answers to your own standards and prior work.
Leo AI surfaces prior calculations, parts, and documents from your own systems, so engineers move faster while keeping work traceable. Start your free trial today.
Schedule a Demo →
#1 New AI Software Globally - G2 2026
Enterprise-grade security
Trusted by world-class engineering teams
