AI for knowledge transfer: An overlooked opportunity

Across industries, one of the most widely discussed applications of AI is augmentation: helping people work faster, automate repetitive tasks, and save time. That is valuable, and construction is beginning to see the benefits in areas like scheduling, reporting, and safety monitoring. But I think there is an equally important application that is largely overlooked: AI’s ability to help people articulate what they know, by asking the right questions.
Most AI applications in construction are designed to process information that already exists, such as documents, drawings, and datasets. There is a different application that I think deserves more attention. It is about using AI to help discover knowledge that has not been documented at all: the reasoning behind decisions, the judgment calls that experienced practitioners make every day, and the lessons that are learned in practice but rarely written down, from recognising early signs that a task is going to overrun, to knowing which approach works best in particular site conditions.

The difference matters. Knowing what happened on a project is useful. Understanding how and why decisions were made is where much of the value lies. That understanding is what allows an organisation to learn continuously, within the context of actual work, rather than relying on retrospective reviews or formal processes that often happen too late to capture the detail.

The construction skills gap is well documented. The CITB’s Construction Workforce Outlook 2025–29 estimates that the UK needs around 239,000 additional construction workers over the next five years. Over a third of workers are aged over 50. The CITB’s chief executive has acknowledged that without change, the industry risks losing valuable expertise as experienced workers retire.

The recruitment and training challenge is well understood. What is less discussed is the knowledge transfer challenge. This challenge extends well beyond construction. Industries such as healthcare, aviation, and energy are all asking the same question. When experienced practitioners retire, does their knowledge and experience leave with them? In a lot of cases, the answer is likely yes, because the mechanisms available to capture and transfer that knowledge are not always designed for the task.

The barrier is rarely willingness. In most cases, people are happy to share what they know. The barrier is more often friction. Capturing knowledge currently requires time, effort, and a format that does not fit easily into a working day. There is a timing problem: documentation tends to happen at the end of a busy shift or at the end of a project, by which point the detail has faded. There is a format problem: structured forms capture data well but are poor at capturing reasoning. And there is a priority problem: when delivery is under pressure, recording knowledge is the first thing that gets deprioritised.

This is where I think AI can offer something different. Recent advances in conversational AI have started to make this kind of low-friction interaction more practical within day-to-day workflows. Rather than asking people to write, AI can engage them in conversation. Rather than requiring a dedicated session, it can fit into existing routines. And rather than capturing only what someone chooses to report, a well-designed interaction can surface context and reasoning that would otherwise go unrecorded.

The potential here goes beyond documentation. If knowledge can be captured at the point of work, structured, preserved, and made accessible, it creates the basis for continuous organisational learning. Not learning in the abstract, but learning within situated practice: understanding what works, what does not, and why, in the specific context of real projects and real conditions.

That kind of learning is difficult to achieve with traditional approaches. It depends on capturing knowledge while it is fresh, from the people closest to the work, in a way that does not add to their burden. AI does not replace the expertise. It provides a mechanism for transferring it: from individuals to the organisation, from one project to the next, and from one generation of practitioners to the next. Once knowledge is captured and structured, it becomes something that newer practitioners can learn from, whether through searchable records, AI-assisted onboarding, or simply having access to the reasoning of experienced colleagues who are no longer on the project.

The conversation around AI in construction is rightly focused on efficiency and productivity. I think it is worth also including knowledge, specifically, how we ensure that what experienced people know does not disappear when they retire, and how AI might help with that challenge.