The future of building inspections can’t stay manual. Meet the startup changing that.

Building inspections have changed little over time. Inspectors still walk the perimeter of tall residential blocks and commercial towers, scanning façades with binoculars or point-and-shoot cameras while taking notes on clipboards. For areas that are hard to reach, rope access teams are often brought in, adding cost and safety risks.

“The conventional means of building inspections have always been very manual,” says Shaun Koo, Chief Executive Officer and Founder of Singapore-based startup H3 Zoom.AI. “Reports are usually not comprehensive, and from inspection to final report, it can take anywhere from one to two months.”

These constraints have raised questions about whether inspections are producing the information owners and property managers actually need to act.

“It’s a process that’s very centred around experience,” notes Shaun. “It’s also centred around limited sampling instead of full coverage when assessing a building’s condition.”

Shaun on a construction site with the H3 Zoom team. 

As buildings age and inspection requirements tighten, these limits become more difficult to ignore. In countries such as Singapore and parts of Europe, inspections are required on fixed schedules, leaving little tolerance for slow reporting or gaps in coverage. 

H3 Zoom’s Interior Inspector platform is responding to these limits by capturing visual data across entire buildings and applying AI-based computer vision and geospatial analysis to assess building conditions.

The platform processes high-resolution pictures and video to identify defects, then ranks them by severity and location. This gives inspection teams a consolidated view of exterior and interior conditions to see where issues cluster, track how they change over time and prioritise maintenance work.
 

The platform starts with visual data captured using drones or other imaging devices. It ingests high-resolution images, video, 3D spatial data and LiDAR (Light Detection and Ranging) data, then applies computer vision, segmentation and multimodal models trained on more than 3 million proprietary datasets.

These models can identify more than 30 categories of defects, including corrosion, spalling, deformation and subsurface delamination (augmented by thermal bridge analytics), which is often difficult to detect through visual checks alone. Each defect is assessed in context, including its severity and how it relates to surrounding conditions.

While many existing tools stop at showing a 3D model or a virtual walkthrough, Interior Inspector supports decision-making.

This includes determining which defects need immediate remediation and which can be addressed during planned maintenance cycles. The platform links visual findings to safety implications and maintenance timing, as well as the financial implications of delaying repairs.

A snapshot of SGInnovate’s office from H3 Zoom’s Interior Inspector. The green line represents the path walked, and the grey dots represent a 3D visualisation of the space. 

This visual analysis can identify unsafe conditions on construction sites, helping teams spot workplace hazards that are often missed during manual checks. To make findings usable without technical expertise, they’re presented through an interface built for everyday operators.

“A nice digital twin doesn’t really help property or facility managers solve problems. Most systems either can’t detect defects at all, or they leave the interpretation to the user.”

Interior Inspector also answers practical questions. Where is the problem located? How serious is it? What happens if nothing is done this year? How much is that likely to cost later?

A capture from H3 Zoom’s Interior Inspector. The yellow dots represent minor defects; red dots indicate more serious defects. 

A recent inspection illustrates how the platform is used in practice. 

In a large Southeast Asian property developer’s handover workflow, issues were captured inconsistently across teams, and the same defect would often be reported multiple times with different photos, then pushed straight to contractors, creating repeated site visits, delays, and avoidable rework. 

In these manual, sampling-based inspection processes, the information gathered is patchy and requires even more checks before decisions can be made with confidence. 

These developers worked with H3 Zoom’s Interior Inspector to streamline processes. Site teams captured visual data once, and the Interior Inspector automatically organised defects by location and ranked them by severity, so managers could quickly see where issues clustered and what needed attention first.

The results were presented in a simple, operator-friendly view that answers practical questions like: Where is the problem? How serious is it? What happens if we delay fixing it?

“Instead of chasing repeated screenshots and duplicated tickets, the system helped teams see the real pattern, what’s happening, where it’s happening, and what to prioritise next.”
 

When Shaun began working on the idea in 2016, the technology focused on geospatial information systems and the interpretation of unstructured images captured by drones.

Early versions of H3 Zoom’s platform focused on basic defect detection to gauge demand. Feedback from property managers and contractors helped the team refine its direction.

“They told us that what they needed was help deciding what to do next,” says Shaun. This feedback drove the development of more sophisticated visual language models that focused on diagnostics and prioritisation. 

Sample report for clients from Interior Inspector.

Making those models useful for decision-making depended on access to data that wasn’t available. “There wasn’t a dataset we could just download,” says Shaun. “If we wanted this to work, we had to go back to the fundamentals of data acquisition.”

The H3 Zoom team began conducting inspections themselves. They paired software subscriptions with inspection services to gather high-quality data while building commercial traction.

Over nine years, this produced more than 3 million proprietary datasets.
 

As the platform matured, H3 Zoom formalised the processes behind its inspection methodology. The company has secured two patents, one in Singapore and one through the European Patent Office. They cover the methodology for using drone platforms for inspections and feeding the data into a common data environment for analysis and review.

The intellectual property, developed in collaboration with the Jurong Town Corporation in H3 Zoom’s early days, protects the know-how behind capturing, processing and accessing inspection data. Together with its proprietary datasets, this gives the company defensibility as it expands.

This focus on improving how inspections are carried out is one reason SGInnovate chose to invest in H3 Zoom, participating in the company’s Series A funding together with JRE VENTURES and M7 Holdings.

Shaun (centre) with the H3 Zoom team, SGInnovate, M7 Holdings and JRE VENTURES team. 

With this backing, H3 Zoom is preparing to expand its presence in Japan and Hong Kong while continuing to grow in Singapore. The company is also building a stronger commercial function and extending its use of robotics to support inspection workflows.

One upcoming development is an AI-powered virtual assistant that will sit on top of the existing platform and help clients draw deeper insights from the inspection data they already have.

For Shaun, the motivation remains personal.

“What keeps me going is seeing an idea move from the lab into real-world use,” he says. “Working with regulators and industry stakeholders who are willing to change how things have always been done in building maintenance, that drives me forward.”

Learn more about Singapore’s fast-growing deep tech community and the startups that SGInnovate supports.