Why construction’s risk problem isn’t about finding risks. It’s about losing them.

The register exists. The problem is everything that happens next.

Every serious construction business has a risk register. Most have years of lessons learned, detailed records of what went wrong, what was missed, what cost them. Teams are good at building these. They can run to thousands of line items, carefully categorised and maintained.

The problem isn’t documentation. It’s distribution.

Quinn describes what happens as a trickle-down effect. Project leaders carry elements of the risk knowledge through procurement and package award, passing fragments down through the layers. But each handover strips away detail. By the time it reaches the workforce on site, the original risk can be four or five steps removed from where it was first identified.

At that point, people aren’t referring to a register. They aren’t thinking in terms of risk at all. They’re doing the work the way they did it on the last project.

The breakdown, as Quinn puts it, isn’t in identifying the risks. It’s in making sure they’re consistently mitigated. And that process falls over somewhere along the chain, on almost every project, in almost every business.

What this costs for construction

The cost of unmanaged risk in construction is not abstract. The Get it Right Initiative (GIRI) estimates risk at up to 21% of build value, around 15 to 18% lost before completion, and a further 3 to 5% emerging during post-completion rectification, where the same issues frequently reappear on subsequent projects.

That last part matters. It’s not just that risk costs money on a given project. It’s that the industry keeps paying the same cost, on the same risks, because the feedback loop doesn’t close in time to matter.

Quinn estimates that feedback loop at seven to ten years. A major project might run for two to five years, but the causes of issues are often not fully understood until long after completion. By the time the lessons are captured and codified, the team has moved on, the project is a memory, and the same risks are sitting unmanaged at the start of the next job.

The industry, as Quinn puts it, still relies on learning by being burnt. You don’t need to touch the hot plate every time to know it’s going to hurt. The risks are already known. The problem is that knowing them at a company level and acting on them at a site level are two very different things.

Why “individual brilliance” is not a system

The current state of risk management in construction relies heavily on what Quinn calls individual brilliance, someone in the team taking ownership, pulling what’s relevant from the register, and making sure it’s managed through the life of the project.

In reality, that rarely happens consistently. And it was never a system. It was a workaround that the industry has come to accept as normal.

The problem with individual brilliance is that it doesn’t scale, it doesn’t transfer, and it disappears when the individual does. When the experienced project manager moves to another job, so does the institutional knowledge they were carrying. When the quality manager who knew which risks to watch for on this type of project goes on leave, the catch net goes with them.

AI-powered knowledge management systems can analyse the challenges encountered, decisions made, and outcomes in previous projects to offer guidance for new ones, establishing learning bridges across different teams, minimising redundancy, and preventing repeated mistakes. That’s the structural shift the industry needs: not a smarter individual, but a system that doesn’t forget. ResearchGate

What AI should actually do here for project teams

Most of the AI conversation in construction right now is about speed. Chatbots, summarisation tools, generative interfaces that help people do existing work faster. That’s not nothing, but it’s also not where the highest value sits.

As Ryan Treweek explains, the problem construction has is a shortage of institutional memory that actually reaches the people doing the work. The highest-value application of AI in construction isn’t a smarter interface. It’s a system that takes what the industry already knows, the lessons, the risk registers, the hard-won patterns from past projects, and makes sure it reaches the right person, at the right moment, before it becomes a problem.

The vision Treweek describes is one where, by the time work reaches site, the worker doesn’t need to think about the register at all. They’re following a process that has already been aligned to the known risks. The catch net is already in place. The intelligence is embedded in the way the work is delivered, not sitting in a spreadsheet that nobody opens.

That’s a meaningful distinction. It shifts AI from a productivity tool to a risk discipline, which is, as the article’s title suggests, data that can see around corners.

The most effective frameworks integrate AI with institutional knowledge, ensuring that digital insights align with on-the-ground realities. The technology alone isn’t the answer. The foundation it runs on is. Construction Dive

The role of consistent data capture

There’s a prerequisite to all of this that doesn’t get talked about enough: the data has to be there in the first place.

Making AI-driven risk tools effective depends on clearly defined processes that are consistently applied across projects. That consistency is what underpins improvement. And that means not being afraid to raise issues.

Quinn is direct about this. There has been a tendency in the industry to downplay or overlook non-conformances and defects, to patch and move on. But those issues, properly captured, are what builds a stronger risk register and lessons learned process. If you aren’t capturing them, you’re missing the opportunity to improve. You’re also ensuring that the feedback loop stays broken.

This is one of the more uncomfortable implications of the AI in construction conversation. The industry can’t shortcut to intelligence. The data foundation has to be built through consistent, honest capture of what’s actually happening on site, the inspections, the defects, the NCRs, the tickets. All of it. Every phase. That’s what gives AI something real to work with.

Connecting intelligence across the full project life

The real value, as Quinn describes it, comes from connecting intelligence across the full life of a project rather than allowing it to fragment across stages and teams.

Inspections, defects, tickets, project documentation, every phase becomes a source of insight. And all of that feeds back into the risk process, refining how future projects are set up and making the same issues less likely to repeat.

That’s how you create continuous improvement that doesn’t depend on one person’s memory or one team’s institutional knowledge surviving the handover. The risk doesn’t get lost in the layers. It’s embedded in the way the work is delivered.

Construction companies in 2026 are increasingly treating risk as a condition that can be identified early rather than an outcome managed after an incident. The shift from reactive to proactive isn’t a technology question, it’s a data question. And the data starts on site, one inspection at a time.

The conversation the industry needs to have

Damien and Ryan’s Inside Construction piece is worth reading in full, not because it’s a product story, but because it frames the problem correctly. The industry doesn’t need more risk registers. It needs the ones it already has to actually work.

That means fixing the distribution problem. It means building systems that don’t rely on individual brilliance. It means capturing issues honestly rather than papering over them. And it means thinking about AI not as a layer you bolt on top of broken foundations, but as a tool that should help make the knowledge the industry has already earned available to the people who need it most.

Construction risk management is well understood. Making it work is the challenge, and it’s one the industry is only just beginning to take seriously.

Read the full cover story ” How Visibuild is using AI to help builders see around corners.