Generative AI should augment, not replace, human decision-making – but it can only do so with the right blueprint, writes Gaku Ueda, CEO of Mode
Although AI has the potential to slash planning time by 30%, adoption in the construction industry continues to lag.
However, there is an optimistic outlook towards AI, with 70% of project managers and quantity surveyors indicating that it can help deliver greater value.
Data quality concerns, complex workflows, and job anxiety are among the primary barriers to broader AI adoption. There is a reluctance to embed new tools due to a misconception that AI only further complicates construction processes.
There is a way forward, though, rooted in knowing what to look for when building an AI adoption strategy. That means asking the right questions from the outset and always keeping people closely informed.
Don’t digitise first and ask questions later
Any processes that involve AI need careful data management strategies. That’s not simply a matter of collecting data, which can lead to a mountain of unprocessable and inoperable insights.
Be wary of automation traps, where system failures are commonplace and costly. Construction projects are among the most complex, with project managers juggling information from multiple sources: suppliers, contractors, site crews, developers, regulators and more.
Generative AI’s reliability as a digital tool starts with data, but gaps and issues are often missed or overlooked. Additionally, data problems that are not immediately acknowledged and addressed will always snowball.
Inaccurate, incomplete and inoperable data have significant consequences. Just some of these include bottlenecked project timelines, angry suppliers waiting on accurate invoices, missed safety risks for on-site construction crews and overlooked building code violations. Relevant stakeholders, therefore, must verify that all applicable data is clear, trustworthy and complete.
Suppliers and contractors often have different formats for creating field reports or invoices that are difficult for a centralised AI platform to accurately process. As a result,
AI-generated reports and summaries are full of errors, making them unreliable for project decision-making. It can lead to chasing a lengthy paper trail, straining already-stretched teams, and stalling project trajectory, not to mention damaged supplier relationships.
Another example is missed safety risks, whether that’s misaligned scheduling so tasks like plumbing and electrical installations overlap, or overbooking crews on back-to-back
shifts. Are applicable requirements, such as OSHA and the Building Safety Act, being met with generative AI in the mix? What are potential dangers from an HSE standpoint?
Data also needs to be protected. Consider the origin and destination of the data. Is data usage fully GDPR compliant? What are possible security risks for databases? Getting on
top of these before actioning AI helps avoid problems further down the road.
Think beyond the here and now
A quick-fix mentality to AI is often a recipe for tunnel vision. Don’t just focus on immediate pain points; long-term considerations must also be accounted for. Truly thinking strategically about AI means stepping back and taking a thorough look at the bigger picture. How do these tools layer into existing tech stacks? How will their performance be monitored to drive tangible outcomes, like faster project completion and greater cost savings?
Crucially, AI consumes and produces a massive amount of data. Issues and insights can easily slip through the cracks. The right guardrails need to be implemented, such as regular audits to review performance against KPIs, and ensuring staff are familiar with leveraging and supervising the technology.
Importantly, while BIM is a foundational element for providing generative AI with insights in construction use-cases, there is a feedback loop between them. Generative AI can analyse BIM to recommend improvements or optimal sequences. Of course, this can only work if BIM is up to date, and no recommendations should ever be actioned without informed human reviews.
In fact, people should be closely involved with generative AI from start to finish. Staff must be well-versed in AI and data literacy so they can determine what it is used for and how its performance should be measured against goals such as cost reduction, time-saving and accuracy. They need to be able to closely monitor and contextualise outputs, whether that’s spotting scheduling issues, modelling errors, or ensuring ethical use and full compliance in areas such as HSE.
These considerations enable construction teams and project managers to understand how AI fits into existing workflows to achieve a smarter and safer industry. Ultimately, generative AI serves to augment human decision-making, not replace it, but it can only do so with the right blueprint—and that means asking the right questions.
The post Building a strong AI blueprint starts with smarter questions appeared first on Planning, Building & Construction Today.
