Probabilistic Tools in a Deterministic World
If you were to peel back the sociological onion of AI skepticism, or perhaps even fatigue in some cases, at the center of it you would find the culprit: a perceived mismatch of process and outcome.
Outcomes in our world are deterministic (or, so we like to think). You input work, you output results…you input variables into an equation, you get the same answer each time. In construction, it's even more imperative to have predictable & repeatable outcomes. Building codes and ordinances ensure that regardless of design differences, our built environment embeds a certain expectation of quality and identical systems.
Our processes, up until the recent AI era, have largely been deterministic as well. Manually measuring, quantifying, looking up answers, standards, guidelines. Then a new, powerful, but entirely probabilistic tool comes along in the form of artificial intelligence (our current iteration of it anyways).
The real root of AI skepticism in construction
The construction industry's skepticism of AI is not Luddism. It's an accurate intuition that probabilistic outputs don't map cleanly into a deterministic world. Validation differs from other industries that have quickly adopted AI, like software engineering, which has the luxury of immediate verification of AI output (does this code run, or not?). Construction is obviously different, with validation occurring weeks or months later on the job site, where you can't afford costly mistakes or scope gaps.
The new frontier
The winners of this era of software will be fighting on a new frontier, not over features, or pricing, or even service, albeit all are important. The winners will be the tools that can best reduce the chasm between AI output and confident outcomes.
In this new reality, the User Interface becomes the most important layer. AI features need to be framed and bounded within deterministic workflows—or they'll keep getting rejected by people who can't afford to be mostly right.
Read more about how we're taking this approach for MEP estimation at Canaveral.