Industrial AI is running into a problem that more capable models cannot resolve on their own, because many factories, energy installations, and utility assets are described by technical records that no longer correspond neatly with what engineers encounter in the field.

Norwegian industrial-software company Aize has acquired Paris-based Samp, combining its engineering and operational-data platform with AI technology that converts laser scans and other reality-capture material into structured information about physical assets. Financial terms were not disclosed.

Samp’s Shared Reality platform is deployed across more than 500 industrial facilities, particularly in energy and water. Its technology turns 3D scan data into structured asset environments and connects physical equipment with engineering information including tags, drawings, and piping and instrumentation diagrams.

Aize approaches the same information problem from the opposite direction. Its software brings engineering, operational, and enterprise data into a common visual environment and is used by companies including BP, ExxonMobil, SBM Offshore, Aker BP, and Aker Solutions, while the business itself is owned by Norwegian industrial investment group Aker.

Bringing the two products together is intended to cover both relatively new assets, where detailed engineering models exist from construction, and older facilities where years of maintenance, modifications, and incomplete documentation have created a widening gap between the designed asset and the physical one.

That distinction becomes important once AI is introduced into engineering and maintenance work. A system can process drawings, equipment histories, sensor data, and maintenance records quickly, but the quality of its recommendation still depends on whether those sources describe the installation that actually exists.

Brownfield assets expose the data problem

Large industrial sites tend to accumulate information across engineering applications, maintenance systems, document repositories, spreadsheets, scans, and specialist operational tools. Some records date back to original construction, while others reflect years of changes carried out by different contractors and operating teams.

Centralising those records does not automatically make them accurate. Organisations can move decades of documents into a modern platform without resolving whether they still reflect the equipment, pipework, instruments, and layouts on site, while a digital twin built from outdated information simply reproduces the same error in a more sophisticated interface.

Samp’s proposition is to treat physical reality as another source of evidence. Laser scanning creates a current three-dimensional representation of a facility, after which its AI technology identifies and structures equipment and links it to existing technical records. Where documentation and reality disagree, the discrepancy becomes visible rather than remaining buried until engineering or maintenance work begins.

The technology therefore addresses a less conspicuous part of enterprise AI: reconciling operational data before organisations automate decisions on top of it. Generative interfaces attract more attention, but engineering systems depend on unusually precise relationships between data and physical assets, leaving less room for a model to be approximately correct.

Aize says it will remain a software company rather than move into the scanning business itself, using specialist reality-capture partners to gather the source material that Samp’s technology can structure. That approach should allow the combined platform to work with existing scanning suppliers rather than requiring customers to adopt a vertically integrated service.

AI moves into the asset record

The acquisition also points towards a broader change in industrial software, where AI is moving beyond being an assistant layered over documents and becoming part of the machinery used to build and maintain the underlying information model. Instead of asking a model to summarise an engineering drawing, organisations can use computer vision and machine learning to determine what equipment is present and connect it with corporate records.

That can affect maintenance planning, engineering changes, remote work, project handovers, and safety-critical preparation because teams spend less time manually comparing drawings with field conditions. The gains, however, depend on how reliably automated recognition works and how organisations govern corrections when machine-generated asset information conflicts with established records.

Aize and Samp both emphasise integration with existing systems, reflecting the difficulty of replacing the specialist applications already embedded across industrial environments. A platform that tries to become the only repository for every engineering and operational dataset would face a lengthy migration problem, whereas an integration layer can connect those systems without requiring customers to remove them all at once.

Aize is also preparing the next generation of its own platform for general availability by the end of 2026, with changes intended to make it easier to work across 3D assets, engineering drawings, and industrial information. Samp’s technology will therefore enter a product that is already being redesigned rather than being bolted onto a static software estate.

Although the acquisition is rooted in industrial technology, the underlying constraint is familiar across enterprise AI: organisations want automated systems to make decisions before they have resolved whether the data describing their operations is trustworthy. In an office workflow, that can produce an incorrect answer or inefficient process; around physical infrastructure, the distance between the database and reality can be considerably more expensive.

By buying Samp, Aize is betting that improving industrial AI starts with making the digital record more faithful to the asset itself. Whether that approach scales will depend on how easily reality capture can be kept current enough to become part of everyday operations rather than another model that gradually drifts away from the facilities it was built to represent.