Most industrial repair delays do not start with a broken wrench. They start with a photo. A technician at a substation sends images of a failing bushing; the remote engineer cannot tell whether the crack runs behind the flange; another call produces more shots from nearly the same angle—and the outage window is gone before anyone shares the same mental model of a three-dimensional asset.

That mismatch is the quiet bottleneck of industrial operations, argues David Fattal, co-founder and CTO of Leia Inc. Infrastructure has grown more complex—distributed energy on aging grids, production lines mixing robots with thirty-year-old machines—while the media people use to share what they see remains stubbornly flat: photos, markups, reports, and video calls that force every recipient to rebuild spatial reality in their head.

Substation and process assets are three-dimensional; flat field photography still forces remote experts to guess scale, depth, and what sits behind the nearest surface.

Where the time actually goes

Industry analysts estimate hands-on work may be only 30–40% of total repair time. The rest is detection, response, diagnosis, coordination, and verification—acts of interpretation. Sensor alarms, field photos, and after-repair evidence all have to survive a journey from site to decision-makers. Cameras collapse depth, hide what sits behind the nearest surface, and give weak scale cues. When the technician guesses the wrong angles, the job rarely fails loudly. It fails as a return visit—wrong parts, incomplete information, mismatched skills—before the truck ever left the depot.

Treating repeat visits as a scheduling problem misses the point. They are often the price of communicating three-dimensional problems through two-dimensional channels.

Cabinet rows and switchgear geometry are exactly where photo markups break down—and where shared spatial models change who can judge clearance and access.

Spatial AI plus immersive displays

Fattal’s path out is not more megapixels. It is two maturing stacks. First, Spatial AI that can turn a sufficiently complete walk-around capture from a phone or inspection camera into a measurable three-dimensional representation—geometry for comparison with drawings, change tracking for corrosion and settlement, and a living record updated visit by visit. Second, switchable immersive displays on everyday laptops, monitors, and tablets that present genuine depth instead of forcing viewers to infer it from shading and motion.

When depth arrives on devices teams already own, adoption becomes a software decision rather than a fleet buy of dedicated headsets. Control of perspective moves from the person who captured the scene to the person who needs to understand it: the remote engineer looks behind the flange herself, checks clearance, and measures. Dispatchers planning against a current spatial model can confirm parts, access, and skills before the truck rolls.

Capture quality still decides whether a spatial model is usable for measurement—or just another pretty mesh nobody trusts for clearance calls.

From better calls to institutional memory

Immediate wins are fewer truck rolls, faster diagnosis, and design reviews where stakeholders share one spatial context. The compounding value is knowledge transfer: a veteran’s walkthrough preserved spatially beats another written procedure in an industry retiring faster than it can replace expertise. Capture by capture, operators build spatial memory that connects to digital twin and asset-management investments already on the books—alongside the same control and protection ecosystems plants already spare, from Honeywell process and safety portfolios to rotating-equipment monitoring lines such as Bently Nevada condition monitoring.

Opinion

Industrial operations did not stall for lack of data. They stalled because the tools for sharing what a person sees strip the dimension that matters most. Restoring shared spatial understanding will not replace PLCs, historians, or work-order systems—it will make those systems argue about the same physical truth. Organizations that treat Spatial AI capture as infrastructure, not a demo, will cut misunderstandings first and wonder later how they ran critical assets on photographs.