AI for Engineering Knowledge Management
Model-based definition embeds PMI and GD&T in the 3D model so it becomes the single source of truth. Here is what engineers need to know in 2026.
·
⏱
7 min read
Michelle Ben-David
Michelle Ben-David is a mechanical engineer and Technion graduate. She served in an IDF elite technology and intelligence unit, where she developed multidisciplinary systems integrating mechanics, electronics, and advanced algorithms. Her engineering background spans robotics, medical devices, and automotive systems.
BOTTOM LINE
Model-based definition is the direction of travel, and the standards prove it is mature rather than experimental. Embedding PMI and GD&T in the 3D model removes the translation layer of the 2D drawing and makes the model the single source of truth, which improves inspection, quoting, and reuse. But drawings are not gone. ASME Y14.41, ISO 16792, and MIL-STD-31000 all accommodate hybrid and drawing based deliverables, so most teams will run mixed formats for years.
The win in 2026 is not picking a side. It is being able to read intent consistently across whatever format a part arrives in, so interpretation and recreation time stop eating your schedule. Get that right, and the single source of truth stops being a slogan and becomes how your team actually works.
For most of the last century, the 2D drawing was the contract. It told the machinist what to make, the inspector what to measure, and the supplier what to quote. The 3D model existed, but the drawing carried the authority. That arrangement is changing. Model-based definition, usually shortened to MBD, moves all of that authoritative information into the 3D model itself, so the model stops being a reference and becomes the single source of truth.
The idea sounds simple, but the shift touches how you tolerance parts, how you release data, and how every downstream team consumes it. In 2026 the question is no longer whether MBD is real. Established standards from ASME, ISO, and the United States Department of Defense already describe how to do it. The practical question is how to manage a world where some parts live as annotated models, some still live as drawings, and your team has to read both without losing time to interpretation.
This article breaks down what MBD actually is, where it beats a 2D drawing, where drawings still earn their place, and how to keep the engineering information legible no matter which format a part arrives in.
What model-based definition actually means
Model-based definition is the practice of embedding all the product manufacturing information directly into the 3D CAD model rather than capturing it on a separate 2D drawing. That product manufacturing information, often abbreviated as PMI, includes dimensions, geometric dimensioning and tolerancing, surface finish callouts, material specifications, and notes. In an MBD data set, those annotations attach to the geometry they describe, so a flatness control sits on the actual face it governs.
This is not a vendor concept. ASME Y14.41, titled Digital Product Definition Data Practices, was first published in 2003 and most recently revised in 2019. It defines how to present this information inside a digital data set. The international equivalent, ISO 16792 under the same Digital Product Definition Data Practices title, was first issued in 2006 and updated again, most recently in 2021. Both standards describe two ways of working, a 3D model paired with a 2D drawing, and a fully annotated 3D model with no drawing at all.
The distinction that matters is authority. In a traditional workflow the drawing governs and the model assists. In MBD the annotated model governs. When the model is the legal definition of the part, every team works from one object instead of reconciling a model against a drawing that may have drifted out of sync. If you have ever fought to find the right file in a crowded vault, the appeal of a single authoritative source is obvious, and it connects to a broader struggle described in why PDM search is broken and engineers cannot find parts.
IN PRACTICE
The connection to our PDM and using that as a data source is legit the best thing ever. I found three viable bracket options fitting my exact envelope constraints, in minutes, not days.
Eytan S., R&D Engineer
Where the 3D model beats the 2D drawing
A 2D drawing is a translation layer. The engineer projects a three dimensional part onto flat views, picks section cuts, and hopes the reader rebuilds the original intent in their head. Every projection is a chance to lose or distort information. MBD removes that translation step because the consumer sees the part as it is, with the tolerances attached to real faces and edges.
The advantages show up across the lifecycle in concrete ways.
Inspection plans can pull tolerances straight from the annotated model, which reduces the manual transcription that introduces errors on a coordinate measuring machine.
Manufacturing and quoting teams read the same geometry the designer built, so there is less back and forth over ambiguous views.
Tolerance reuse improves, because a feature carries its control wherever the model travels rather than living only on a drawing sheet.
MBD also reduces silent duplication. When intent travels with the geometry, two engineers are less likely to recreate near identical parts because one of them could not read the other's drawing. That waste is real and measurable, and it is the subject of the real cost of duplicate parts and how AI supports reuse. The cleaner your definition, the easier it is to find and trust an existing part before you draw a new one.
Where 2D drawings still earn their place
MBD is not a clean replacement for everything. The standards themselves acknowledge this. Both ASME Y14.41 and ISO 16792 explicitly support a hybrid mode, a 3D model accompanied by a 2D drawing, which tells you the committees expected mixed environments to persist for years.
Several practical realities keep drawings in service.
Suppliers and shop floors vary in their tooling, and not every partner can open or trust an annotated native model or a 3D PDF.
Legacy archives hold decades of parts defined only as drawings, and reconstructing each one as a full MBD data set is rarely justified.
Some review, markup, and contractual sign off processes are still built around a printable sheet that a person can hold and annotate.
The United States Department of Defense standard MIL-STD-31000, which governs technical data packages and was most recently revised as MIL-STD-31000B, reflects this pragmatism. It accommodates 3D technical data, including the 3D PDF format, while still allowing for traditional deliverables. The honest position for 2026 is that most organizations run a blend. Some parts are pure MBD, some are model plus drawing, and a long tail remains drawing only. Your workflow has to handle all three without forcing a costly conversion before anyone can do useful work, a theme that runs through good engineering knowledge management.
The hidden cost of a mixed format world
When some definitions live in models and others live in drawings, the friction does not disappear, it moves. It moves into the daily work of interpretation. An engineer evaluating an existing part for reuse may need to open a native model to read its PMI, then open a separate PDF for an older variant, then decode a scanned drawing for a third. Each format demands a different mental gear, and the information you actually need, a tolerance, a thread callout, a datum scheme, is buried at different depths in each.
This is the quiet tax of the transition. The data exists, but it is not uniformly readable. Recreation time creeps up because it is faster to redraw a feature than to hunt for how it was defined three formats ago. Design review slows because reviewers cannot quickly confirm that a model and its drawing agree, a risk explored in how AI design review catches errors before manufacturing. The promise of MBD as a single source of truth is real, but it only fully pays off when you can read intent consistently across the formats you already own, not just the ones you create going forward.
How Leo reads PMI and GD&T across models and drawings
This is where an intelligence layer changes the economics of the transition. Leo is an AI layer that sits on top of your existing PDM and PLM system rather than replacing it. It reads the engineering information where it lives, the PMI and GD&T embedded in annotated 3D models as well as the dimensions and tolerances captured on 2D drawings, and makes that intent searchable and comparable in one place.
The concrete value driver is interpretation time. Instead of an engineer manually opening three formats to confirm how a feature is toleranced, Leo surfaces that information directly, so a part defined as a drawing in 2015 and a part defined as a pure MBD model in 2026 can be compared on the same terms. That cuts the recreation work that happens when intent is too hard to retrieve, and it strengthens geometry-aware part search so an existing definition is found before a new one is drawn. Leo connects to engineering data through integrations that are available for SolidWorks PDM, Autodesk Vault, PTC Windchill, Siemens Teamcenter, and Arena PLM, which means it reads across the mixed archive you already have. For teams weighing tooling, the wider landscape is mapped in the best AI tools for CAD in 2026.
FAQ
Read intent across every format
See how Leo turns models and drawings into one searchable source
Leo is an AI layer on top of your PDM and PLM that reads PMI and GD&T from both 3D models and 2D drawings, with integrations for major systems.
Schedule a Demo →
#1 New AI Software Globally - G2 2026
Enterprise-grade security
Trusted by world-class engineering teams
