The construction industry has invested heavily in digital tools for over a decade, but walk into any job site and the same issues persist. We’ve spent billions on BIM, digital twins and visualization software, but we’re still faced with rework, shoddy installation and crews checking measurements by hand. Look around and you’ll find measuring tapes, chalk lines and conflicting reference points all over the workplace. It’s a reminder that our digital investment still stops at the trailer door.
The problem is simple and expensive. We digitize the design but not the work. We build detailed models, but the way we build hasn’t kept up. Until construction digitizes its physical workflows, tools, and verification, digital twins will continue to fall short of their promise.
A broken thread: design versus production
Mike Tyson once said, “Everybody has a plan until they punch you in the face.” In construction, the model is the plan and the workplace is the punch.
Pour concrete and the column lines are slightly off. You start hanging the pipe and it collides with the existing structure. Teams follow the plan but make field adjustments to fit what was built before. We are digital in intent, but our field execution remains analog. The digital twin, which should reflect a living process, becomes a static file that loses value as soon as the real work begins.
This disconnect costs real money. On a facade project I worked on, two subcontractors shared responsibility for adjacent panel sections installed in an alternating sequence. Both used different design teams and slightly different assumptions. One team adjusted their field spacing to accommodate slab changes, while the other followed the layout drawings exactly. Both were well-intentioned, but those small tweaks turned into major mistakes. After comparing their work, the deviations exceeded what could be corrected in the field. An allowable tolerance of a quarter inch was converted to inches of misalignment across the floor line.
This also raises a larger question: what happens when that same facade conforms to slab conditions, but interior partitions have already begun? If the wall-to-stud connections have tighter tolerances than the concrete, the change creates a rework. If every trade was tied to the same live model that updated build conditions in real time, these issues would never reach the facility.
We are digitizing paper, not processes
These are not isolated errors, but show how disconnected digital coordination is from field execution. The data flow stops when it reaches the site. Without structured feedback, we rely on interpretation and memory. Even the best digital twin can’t fix what it can’t see.
As Hamzah Shanbari said, the industry is still “digitizing the paper, not the process.” Replacing blueprints with PDF is not transformation, it is translation. The software changed, but the work didn’t.
From projects to production systems
The real obstacle is not technology, it’s how we think about projects. Construction treats each job as unique, while industries like manufacturing or energy treat theirs as production systems. Todd Zabelle wrote, “Until we recognize projects as production systems and use operations science to drive project outcomes, we are doomed to fail.” Without visibility into production, there is no feedback loop, no fixes, and no real improvement.”
Manufacturing knows the cost of each delay because it measures performance as a single continuous system. Construction is said to be too complex for that, but the truth is simpler: manufacturing manages its process; construction manages the paperwork.
To manage construction as a measurable process, we need verifiable and structured data coming back from the field. Right now, that feedback depends on people filling out daily reports, checklists, or photo logs by hand. Some apps make it easier, but the information is still subjective; it is only as reliable as the person entering it. Without objective data capture, there is no reliable feedback loop.
In manufacturing, machine sensors and logs feed data directly into control systems, giving managers a live view of performance, quality and downtime. Some processes still involve a human quality review, but they are part of a structured system. The build does not have this mechanism. Each contractor acts as an isolated station in the process, largely due to risk mitigation issues. Contractual language encourages separation and self-protection. There are few shared standards for verification, so feedback remains fragmented.
Imagine if each workstation on a manufacturing line was an independent company, each with its own data, tools and risk profile. This is how construction works today. The system works, but only because it has to. The real opportunity lies in building that very structure through shared data, not control.
Market pressures that demand change
Two forces make this change inevitable. The first is work. For decades, experienced tradespeople made calm field solutions when drawings conflicted or conditions changed. This knowledge base is disappearing as veterans retire. Younger workers understand the software, but lack the experience to catch small bugs before they grow. Without data-driven verification, these errors now travel farther and cost more to correct.
The second is accuracy. The growth of hypercritical infrastructure projects (data centers, battery facilities and clean rooms) is driving owners to demand higher levels of detail. Many now require LOD500 deliveries as standard. But collecting, processing, and organizing as much verified and constructed information can cost almost as much as building the project itself. LOD500 provides the framework for true digital twins, but the cost and overhead make it impractical for widespread use.
Bridging the gap between data and work
3D scanners and robotic total stations are a good starting point, but they have limits. Full site scans can cost tens to hundreds of thousands of dollars per project and still require manual setup and interpretation. Point clouds often return unstructured data that is difficult to verify. Ground-penetrating radar faces the same problem: it detects utilities and voids, but the data is provided as flat anomaly maps rather than geometric models.
Most companies avoid interpretation because of liability risks. Even when data is collected, it arrives incomplete. Occlusions, limited coverage, and time intervals between scans leave critical areas undocumented. It’s like a bad manager checking in every once in a while to ask, “Is everything okay here?” while the thousands of changes in between are lost.
Automation changes this by changing data collection from periodic to continuous. Robots equipped with LiDAR, vision systems and location sensors collect data as they go about their work. Every hole drilled, support placed, or inspection completed is added directly to the shared construction model. Instead of working with occlusions, data is captured as it is being installed. The structure is documented as closed.
Because data is generated as work is done, it includes geometry, context, and time. This eliminates the pauses, manual registration and heavy post-processing required by traditional scanning. The documentation becomes a by-product of the work itself. Robotic data collection creates a continuous digital heartbeat for the project, similar to a manufacturing line, where deviations are detected early and accuracy improves over time.
With automated verification, owners can achieve LOD500 accuracy without the cost and paperwork burden that previously made it unrealistic. Robots, sensors and connected tools are finally closing the feedback loop that digital twins have always promised.
The way forward
The gap between model and reality will not be closed with more software. It will be closed through ongoing and verified feedback from the field. Automation provides that connection, allowing every part of the project to be measured in real time, producing data that is accountable and actionable. Instead of building a new digital twin at the end of the project, the twin is built as the work progresses.
Automation will not eliminate the need for skilled workers, but will elevate them by empowering them to make decisions based on real-time data. Automation replaces guesswork with testing, turns verification into collaboration, and transforms static building models into living systems.
The question is not whether construction can be digitized further. It’s if we can afford not to.
Conley Oster is partner–founder and COO of Raise Robotics. With a background in structural engineering and extensive experience operating cranes and rigs on complex construction projects, he now uses his field experience to advance practical robotic automation in industry.
