I wrote recently about 3D scanning technology and how it’s transforming the documentation of historic stone carvings. The scanning itself — capturing millimetre-accurate geometry of carved surfaces — is powerful. But the real shift happening in 2026 isn’t just about scanning. It’s about what happens after the scan.

Artificial intelligence is being applied to 3D scan data in ways that go well beyond documentation. AI algorithms can analyse stone surfaces to identify damage patterns invisible to the eye, predict how deterioration will progress over time, and generate restoration proposals that would take a human conservator weeks to develop manually.

I’ve been involved in three restoration projects over the past year where AI-assisted analysis fundamentally changed how we approached the work. Here’s what I’ve learned.

Damage Assessment at a New Scale

When you scan a deteriorating stone surface, the resulting 3D model contains millions of data points. A skilled conservator examining the model can identify obvious damage — cracks, erosion, missing elements. But subtle deterioration patterns are easy to miss, especially on large surfaces.

AI trained on datasets of stone degradation can analyse a 3D scan and flag deterioration that hasn’t yet become visible damage. Micro-fracture patterns that indicate structural weakness. Surface texture changes that suggest salt crystallisation beneath the stone surface. Subtle geometry shifts that reveal differential weathering across different stone beds within the same block.

On a sandstone courthouse facade I worked on in regional New South Wales last year, AI analysis of the 3D scan identified twelve areas of incipient spalling — where the stone surface was about to detach — that our visual inspection had missed. Three of those areas were in structurally significant locations. Without the AI flagging them, we might have addressed the obvious damage and left the building vulnerable to failures we didn’t know were coming.

Predicting Deterioration

This is where the technology gets genuinely fascinating. Given enough historical data — multiple scans of the same surface over time, combined with environmental data like rainfall, temperature cycles, pollution levels, and salt exposure — AI models can predict how stone will continue to deteriorate.

The Getty Conservation Institute has been developing predictive deterioration models for stone since the early 2020s, and recent versions are impressively accurate. Given a current scan and environmental parameters, the models can project what the stone surface will look like in five, ten, or twenty years under different conservation scenarios.

For stone carvers working on restoration, this changes the conversation with clients entirely. Instead of saying “this stone needs attention,” we can say “without intervention, this carved element will lose its defining details within eight years, but if we treat the surface and redirect water drainage now, we can extend its life by thirty to forty years.” That level of specificity helps heritage managers prioritise budgets and schedule work before damage becomes irreversible.

Generating Restoration Templates

Perhaps the most practical application for my day-to-day work is AI-generated restoration templates. When a carved element is partially damaged — say, half of a decorative corbel has eroded — the AI can analyse the surviving geometry, identify symmetry patterns, reference similar elements elsewhere on the building, and generate a complete 3D model of what the element originally looked like.

This gives me a precise template to carve from. In the past, I’d reconstruct missing elements based on photographs, architectural drawings, and educated guesswork. The results were good but involved significant interpretation. An AI-generated template reduces the guesswork by providing geometry derived from the actual stone rather than secondary sources.

I should be honest: this technology doesn’t always get it right. When damage is extensive — when more than 70 percent of an element is lost — the AI’s reconstruction becomes increasingly speculative. And stylistic nuances that a human carver would recognise — the particular hand of a specific nineteenth-century carver, regional style variations, intentional asymmetries — can be lost or smoothed out by algorithms optimising for geometric regularity.

The best results come from using AI-generated templates as a starting point, then applying craft knowledge to refine them. The AI provides the geometry. The carver provides the understanding of style, material, and intent.

The Practical Workflow

Here’s how a typical AI-assisted restoration project works for me now.

First, the scanning — structured light for detailed elements, photogrammetry for larger surfaces. One to three days per facade.

Second, the AI analysis. I send scan data to a processing service that Team400.ai helped me set up, connecting scanning output to cloud-based analysis tools. The AI returns damage assessment, deterioration predictions, and restoration proposals. Turnaround is typically 48 to 72 hours.

Third, the human review. I examine every AI output against my own assessment. Damage identification is usually excellent. Restoration templates require varying degrees of revision.

Fourth, the carving. This part hasn’t changed. I carve stone by hand, guided by the template but responsive to the material. No digital template can account for grain direction, colour variation, or hardness changes within the same piece. The carver still needs to read the stone.

What This Means for the Trade

Some carvers are uncomfortable with AI playing any role in restoration. I understand the concern — stone carving is one of the oldest skilled trades in human history.

But the AI handles analysis and prediction. The carving itself remains entirely human. Nobody is 3D printing restoration elements. The technology improves the information I work with, not the work itself.

What it does change is efficiency. Projects that required weeks of research can now be planned with greater precision. I spend less time on administrative analysis and more time doing what I trained to do: carving stone.

The tools are getting better. The craft hasn’t changed. I’m comfortable with that balance.