Editable AI Geometry Is Promising, but It Still Needs a Production Check

AI image generators can produce an impressive building concept in seconds. The problem begins when the design team asks a practical question: can we edit it?

Illoca is addressing that gap with Plamo, a browser-based tool now in beta. According to AEC Magazine, Plamo can turn sketches, images, drawings and natural-language instructions into editable geometry using boundary representation and parametric logic.

That distinction matters. A rendered image may communicate an idea, but it does not provide the geometry needed to test dimensions, relationships or alternatives. Editable output has the potential to become part of an actual design workflow.

The opportunity is strongest in early-stage work, but the output should still pass a production check before it is trusted on a live project.

Where editable generation could save time

Early design involves repeated exploration. Teams test massing, circulation, system zones, site relationships and the effect of nearby buildings. Many of these studies do not need fully detailed BIM content, but they do need geometry that can be adjusted.

AI-generated parametric geometry could help create:

• early building massing options;

• simplified models of nearby structures;

• context for shadow and solar studies;

• preliminary space or system layouts;

• quick alternatives for design discussion.

The benefit is not that AI produces a finished building. It is that the team may reach a useful starting point faster and spend more time evaluating alternatives.

For example, surrounding buildings often need to be modeled only well enough to study their effect on the site. If a tool can generate reasonable, editable context geometry from available drawings or images, the designer can refine what matters instead of building every mass from scratch.

Editable does not automatically mean production-ready

The word “editable” can cover a wide range of quality. Geometry may be technically modifiable and still be difficult to use downstream.

Before generated content enters a live project, we would check four areas.

1. Geometry

Are the forms clean, closed and dimensionally reasonable? Are there overlapping surfaces, tiny fragments or unnecessary complexity? Can the geometry support the intended analysis or modeling task?

2. Coordinates

Does the content use the correct origin, orientation, units and project coordinates? A model that looks correct in isolation can create significant problems when linked into a coordinated environment.

3. Parameters

Are important dimensions and relationships actually controlled by meaningful parameters? Are parameter names understandable? Do changes behave predictably, or do they break the form?

4. Downstream editability

Can the receiving team modify the geometry in its normal authoring workflow? Does it remain usable after transfer, or does it arrive as a heavy collection of generic objects? The real test is not whether the source tool can edit the model. It is whether the next person can continue the work without starting over.

The checking effort must remain lower than the modeling effort

Generated geometry creates value only when the time saved in creation is greater than the time spent cleaning and rebuilding it.

This will vary by use case. A rough mass used for a shadow study can tolerate more simplification than geometry intended to influence structural grids or building systems. Teams should define the required level of reliability before generation begins.

A simple acceptance checklist can help:

• intended use is stated;

• units and coordinates are confirmed;

• key dimensions are checked against the source;

• geometry is reviewed for errors and unnecessary complexity;

• parameters are tested through several changes;

• transfer into the downstream tool is tested;

• a responsible person approves the content before project use.

A better starting point, not an automatic final answer

Plamo represents an interesting shift from AI-generated pictures toward AI-generated design objects. That is closer to what AEC teams need.

Still, the value of the tool will be proven in the handoff between generation and production. If the geometry is clean, correctly located, meaningfully parametric and easy to continue editing, it can shorten early exploration. If it requires extensive repair, the apparent speed disappears.

The right approach is neither blind acceptance nor automatic rejection. Use generated geometry where it provides a faster starting point, then check it according to the consequences of its intended use.

For help developing reliable BIM content and model-quality workflows, contact DDG.

Sources

• AEC Magazine report on Plamo

• Illoca product site