The Keys to AI Technology Leapfrogging

Adam Krob, Director of Information Technology, Field Audit and Process Improvement, Boh Bros. Construction Co., LLC

The construction industry continues to hold on to its reputation as a very late adopter of technology. Bernadette Salapare in Construction Times outlines seven obstacles to construction technology adoption, from legacy software to a lack of standardization of data types. These obstacles have prevented the industry from realizing the efficiency gains enjoyed by other industries, such as manufacturing or retail sales. With the emergence of artificial intelligence (AI), large language models (LLMs) and intelligent agents, the construction industry has a rare opportunity to leap ahead to a more advanced position relative to other industries.

The open question is: what will be different? How will LLMs, chatbots and AI agents succeed where other technologies have not?  The difference will have to be how construction companies tackle the human side of the technology productivity equation. There are three critical shifts construction should embrace for this leap: abandoning the insistence on uniqueness, building structured logistics for information and embracing 80 percent functional fit.

Abandon the Insistence on Project Uniqueness

The belief that every construction project is a unique undertaking has deeply influenced how we build systems, processes and data models. It has led us to underappreciate the similarity across projects and has prevented us from developing standardized ways of collecting and using data.

While the environments we work in are variable and complex, our approach does not have to focus on the outliers, but on the similarities. Predictable, repeatable processes create the foundation for clean, structured data models that remain consistent across time, projects and teams. This data consistency unlocks meaningful long-term analysis, benchmarking and AI-powered decision support.

Abandoning the focus on uniqueness does not require that we ignore complexity and nuance. Through techniques like mass customization using LLMs, we can tailor outputs without compromising the structure of data collection. Mass customization approaches will allow large language models to adapt to different conditions or requirements while keeping the chat or speech interface the same. These models serve as adaptive layers between humans and systems, absorbing differences in language, context and inputs while maintaining data integrity behind the scenes.

The key is to break down project elements into micro-level repeatable tasks and understand what is different, introducing new agents or prompts to collect the correct information and run the appropriate processes. Each project becomes a recombination of pre-generated elements that can adapt to the specific needs of a project, reducing the demand for customized data collection while increasing future predictability. Mass customization requires better and more tightly controlled information flows, which suggests a different approach to data collection.

Build a Structured Logistic Approach for Information

In the construction industry, job buyout, submittal review and delivery management are key components to effectively manage physical logistics. Contracts invest significant time and effort to get physical logistics right. Unfortunately, we do not invest in information logistics, a term I first heard from Oleg Kandrashou, the CEO of Cubby.

Data is typically entered manually, such as daily production quantities, then passed through multiple systems before reaching decision-makers. Delays and inconsistencies are common in this data gathering process. The submittal review process illustrates this inefficiency, with the process often spanning at least three systems: emails to share marked up versions with subcontractors, project management database entries to send to architects or engineers and spreadsheets to manage deliveries and delay tracking.

“There are three critical shifts construction should embrace for a leap ahead: abandoning the insistence on uniqueness, building structured logistics for information and embracing 80 percent functional fit”

A logistics-centric approach to data collection begins with defining inputs and handoffs just as we would in physical workflows. Each point in the chain, from field input to project management software to the executive dashboard, must be auditable, timely and structured. A submittal workflow in this logistics model will provide a single platform with visibility of versioned markups by all stakeholders, age of submittal at each stage and delivery estimates.

The rise of AI and agents makes this possible. Intelligent interfaces can provide real-time weather data for foremen when entering timesheets, validate that cost codes are open for labor and upload that information into the payroll system of record regardless of the origin of the information.

AI is not just an analytical layer, but the infrastructure of the information supply chain. While AI infrastructure allows complex projects data to be brought together, truly embracing the possibilities of AI will lead the industry to adapt the processes, as well.

Increase Tolerance for 80 Percent Functional Fit

Perhaps the most culturally difficult change is abandoning the myth of perfect fit of a technology tool to company processes. Construction has historically demanded that tools adapt 100 percent to existing workflows, often resulting in significantly customized systems, multiple overlapping solutions and disappointment in final products.

This demand for 100 percent fit is one of the significant factors for construction’s inability to keep up with other industries’ adoption of technology. As construction firms start to adopt AI more broadly, it will help the industry grow beyond this demand. AI provides adaptability, enabling more generalized systems to meet specific needs with fewer customizations.

This shift, like the other two, will be successful when perspectives change. With AI, teams can abstract from the technology and think about it as a reflection and enhancement of their own processes. This is why seeing AI in context of processes is essential. When people see immediate value from automation or data-driven insights, they can become advocates, even if the tool isn't perfect.

The success of Lean Construction demonstrates that repeatable, high-performance practices such as pull plan look-ahead schedules are possible even across diverse projects. The same mindset can be applied to technology: optimize processes, find tools that support 80 percent of the processes well and let AI handle the last 20 percent.

The Leap Is Within Reach

Construction has a rare chance to leapfrog using AI-driven tools, but it will require changes to mindsets, not just tools. AI is not a tool, but a way to supercharge labor. The best place to get started is by looking at the processes that are consistently problematic, featuring delays, poor data quality and excessive rework. Look at where there are commonalities with successful processes and inject AI interfaces, LLMs and agents to adapt for them. Executives can get started by identifying 2-3 high-friction processes and assessing how AI-enabled tools or agents could improve them within the next quarter.

The opportunity to use AI-driven technologies in construction is extraordinary. The industry should not settle for catching up. It must leap ahead or watch every other industry adopt and adapt to the AI revolution.