Are we getting too excited about AI’s impact on construction?

Barely a day goes by without news of an innovative AI-powered product or service that promises to revolutionise the construction industry.

Indeed, Construction Briefing itself has been full of news of AI advances: the term “artificial intelligence” brings up 606 article results on the constructionbriefing.com website alone.

But is its potential impact on the construction industry being overstated? A recent research briefing from Oxford Economics suggests that perhaps it could be.

Oxford Economics looked specifically at the potential productivity gains from generative AI (GenAI) and other similar technological advancements. GenAI, which is capable of generating text, images and audio is likely to enhance productivity gains and drive economic growth over the medium term, it forecast. But the distribution of those productivity gains is not going to be even across different industry sectors.

And construction is one sector where the impact on productivity from GenAI is going to be significantly less than in others like services, Oxford Economics said.

That’s because GenAI will prove most useful when it comes to automating cognitive and routine tasks, the authors of the research briefing said.

In industries like construction and agriculture, where there is a more significant element of physical and manual work, GenAI’s ability to boost productivity is likely to be more limited.

‘Not a wholesale takeover of tasks’

“The reality is that the vast majority of tasks done by people in construction are out in the real world where GenAI lacks the ability to go out and touch and move things,” Nico Palesch, a senior economist at Oxford Economics and one of the authors of the report, told Construction Briefing.

“New tech might help on the margin, but it’s not the same as the wholesale takeover of tasks done by people who sit in front of computers all day. GenAI can already automate much more in the virtual world,” he added.

Incorporating sectoral adoption rates moves information to the top of the rankings (Source: Oxford Economics/US Census Bureau)

In fact, construction sits second from bottom on a ranking that scores different industry sectors’ potential benefit from AI, combined with their current adoption rate (see chart above).

That’s not to say that there is no potential benefit of GenAI to construction but that it’s much lower than other areas like pharmaceuticals, machinery and chemicals, according to the ranking.

“Sectors we expect to benefit the least from GenAI adoption sit mostly in the production side of the economy, with agriculture and construction the lowest ranked. These sectors still have scope for AI-driven productivity gains but to a far lesser degree than the service sectors,” the authors of the briefing note said.

The research comes off the back of previous work by Oxford Economics that used machine learning to analyse the exposure of over 18,000 tasks to GenAI automation. It then mapped those tasks to job roles and then to industries, using US labour market data. That process then enabled Oxford Economics to assess the GenAI exposure of each industry.

Disproportionate impact on certain roles

Where GenAI does have an impact on the productivity of construction, it will fall more heavily on the relatively few office and managerial roles rather than on the larger proportion of manual roles, according to Oxford Economics.

The research briefing also acknowledged that there are factors and technologies that it has not included in its sectoral benefits index. These could potentially boost productivity in sectors like construction in ways unrelated to automating tasks.

Harder-to-automate jobs in construction make up >70% of sectoral employment (Source: Oxford Economics/BEA)

For example, it highlighted the potential for GenAI to analyse production processes and reduce the time and effort involved in constructing or re-tooling factories. AI-driven software that helps robots’ visual recognition capabilities could also help them to act more freely in different contexts and while such benefits are currently speculative, they could eventually drive productivity in production sectors, the authors pointed out.

And there’s another benefit to the construction industry from AI, albeit one that does not directly relate to automation of tasks in the sector, according to the briefing. That comes from the construction of many more data centres, which are being used to train and operate computationally intensive AI models.

The authors recognised this as an important source of new work. “We expect that the decades-long trend toward increasing data centre construction will only accelerate in the coming years as GenAI adoption picks up. This will be accompanied by a buildout of energy infrastructure: data centres are power-hungry, and new electricity-generating infrastructure will be needed to power them,” the authors said.

The report tempered this forecast by highlighting the fact that the activity associated with the expansion of data centres will still only be a relatively small share of overall construction activity, however.

“In the US, for example, in 2023 data centre construction was only 2% of the size of total nominal residential construction, a drop in the bucket which, as our latest US construction forecasts show, is not expected to fundamentally shift over the coming years,” it concluded.