As construction firms increasingly integrate artificial intelligence into their operations, the challenges stemming from poor data organization have come to light. In my experience with Pacific Pile & Marine, the hurdles we faced weren't technological; they stemmed from a lack of standardized data practices.
When we transitioned to a comprehensive AI platform, I anticipated difficulties around the technology itself—choosing the right model and ensuring robust connections to our existing systems. Surprisingly, those aspects went smoothly. The real complications manifested when operational teams started utilizing the AI for actual tasks, revealing a startling truth: we were unsure of our own data.
One of the pressing questions that emerged was about the current revision of project submittals. Unbeknownst to us, AI couldn’t provide a definitive answer. Why? Because we lacked clarity and consistency in our own file management practices. The model was processing the data exactly as we had stored it—often in multiple locations with varying names and no clear indication of which version was authoritative.
Data Inconsistencies: A Reality Check
For years, experts on the ground relied on informal knowledge. The project superintendent knew which folder was the definitive source, and estimators understood that "final_v3_REV" trumped "final_FINAL." However, AI lacks this intricate, shared understanding and merely highlighted discrepancies that we had long navigated. This issue is a cautionary tale for other contractors beginning their AI journey: implementing AI will not resolve your data issues; it will unearth them.
Understanding the Core Issue
A recent article noted that the barrier to effective AI integration in construction isn't the technology itself, but rather the failure to establish a common linguistic framework across the construction process. The authors suggested an ontology—an overarching digital model that aligns language across various departments—as a solution. While I agree, my perspective suggests that a shared language must emerge from very basic conventions, such as folder naming and file handling.
If various teams misplace the same document under inconsistent names and locations, no software will reconcile those differences and enable comparability. Standardizing our folder and document naming was our initial goal; we aimed to ground our discussions in straightforward terminology before tackling more complex integrations.
Implementing Gradual Change
Instead of enforcing a broad, company-wide naming convention from the outset, we appointed one project manager to pilot a standardized approach on an active project. This method allowed us to evaluate the feasibility of the structure in real-time, yielding insights we could never have gathered in a conference room. A standard that emerges from the field rather than Christened in an office tends to resonate more effectively with on-site personnel.
This pilot approach also shifts the ownership of the standard. When a project manager champions the protocol through hands-on experience, it transforms into an operational necessity rather than an abstract IT guideline. Colleagues are more likely to adopt a system that has proven effective in real scenarios.
Defining Authoritative Data
Alongside refining the naming conventions, we needed to establish a clear definition of what constitutes authoritative information. Given the inevitable chaos of working project files, it became paramount for a designated individual to scrutinize the AI's outputs before treating them as the ultimate truth. For instance, to prepare for upcoming prequalifications, we’re generating a unified summary for each past project, ensuring it contains consistent and agreed-upon content.
Fostering Adaptability Over Technical Skills
Anthony Chiaradonna, Consigli Construction's CIO, succinctly stated that the most crucial shift in the industry isn't purely technical. Rather, it’s about how well teams adapt and embrace change without sacrificing quality. In our case, those who excel in using AI are not necessarily the most technology-savvy individuals but rather those willing to adjust long-standing habits regarding data management.
Our training programs have evolved to encompass not just generic technical skills, but also the vital details of where and how files are saved. This adjustment is essential when addressing pushback on AI’s effectiveness; the common issue isn't the AI itself—it’s the state of the data it has been fed.
Guiding Principles for Effective AI Rollout
Here’s what I would recommend to other construction professionals embarking on their AI journeys:
- Treat initial weeks of AI implementation as a rigorous data audit; every inaccurate response reflects an underlying file or naming issue.
- Establish consensus on the terminology to be used prior to selecting any AI platform, as these will shape how the AI interprets and interacts with your data.
- Test new naming conventions live on a project with a project manager who takes ownership of the initiative.
- Designate a person to validate what is treated as credible data, ensuring there’s a knowledgeable filter between raw files and AI outputs.
- Focus on adaptability in your hiring and training processes; the willingness to change existing workflows will be more critical than technical proficiency alone.
The chosen AI model may become outdated quicker than we anticipate, but the standards for folder naming and document management will endure, defining the success of any technology integration. Recognizing and addressing these foundational issues is key to navigating the complex arena of AI within the construction sector.
