Come Tomorrow, Part 3
Vince Bailey / July 2019
It ain’t necessarily so...—George and Ira Gershwin
I spent the last two months’ installments of “Estimator’s Edge” making a case for my notion that the forthcoming emergence of artificial intelligence will almost certainly render the function of human estimators obsolete within the next five to 10 years. Having now heard protests from several readers, I might be willing to backpedal some on that rather bold statement. Most of those who refute my conclusion hang their argument on the belief that AI will simply make the estimators’ jobs easier—that, as with architects, the more powerful programs will relieve the estimator of some of the more mundane, repetitive aspects of our occupation.
I concede that my conclusion is speculative, but I still don’t believe it is far off base. Let me present a model of what I believe the advent of AI will bring to our industry. After reviewing the following scenario, thoughtful readers can then form their own conclusions.
My concept rests on a foundational assumption that the rapid development of programs that can store and manipulate vast amounts of data are presently being pursued and soon to be implemented. Let’s also assume that these databases are standardized and subscribed to by a consensus of contractors and architects. Not really a pipe dream when you consider the current monopoly that a few high-tech mega-corporations hold at present over information technology and retail sales (Facebook, Amazon).
Now, for the purpose of our commercial drywall example, let’s suppose that within one of these programs (let’s call her “Fat Lady”—we like to name our programs) is a codified base of all possible wall types (or ceiling types). Not a listing of general, most-used wall types like we work with now, but a catalog of every possible combination of components known to the industry—i.e., every possible 3 5/8" wall with multiple first, second, third options (25 gauge, 1 layer 5/8" each side, sound insulation full height, 1-hour rated. Then, the same, but non-rated. Then, the same, but without insulation … on and on, into the hundreds of thousands of possible combinations in assemblies). Fat Lady is capable of containing millions of such multiple-component assemblies.
We estimators currently apply or omit options from most frequently used model assemblies to determine an applicable wall assembly for what is indicated on the plans. But with Fat Lady, our vast database of the near future, a single code could be selected for the perfectly constructed assembly to address the required performance. With no variants to be determined, a critical step in the human estimating process is eliminated.
Let us further suppose that these assemblies are codified by performance specification. An AI program could be designed to select the appropriate assembly for the performance requirement. Take, for example, a hotel shell—multiple guestroom floors with “X” amount of square footage floor space. Hypothetically, an architect would submit his floor plan (a shell with no interior walls) to a program called “Wallmaster.” Wallmaster would divide the space into a certain number of guestrooms, corridors, lobbies and stair or elevator wells, and perform the layout of all partitions. But in addition, let’s suppose these programs “talk” to each other. Wallmaster could easily assign wall types per performance spec by way of a simple conversation with Fat Lady.
To illustrate, Wallmaster “knows” that demising walls between guestrooms require 1-hour fire rating and a sound transfer coefficient of 65. Wallmaster contacts Fat Lady and dials up the code for his required assembly. Fat Lady sends him a 6-inch assembly that includes two layers of 5/8" type X board on each side, 6-inch insulation in the cavity, and RC1 channel with a height that coincides with the deck height Wallmaster is working with. He similarly dials up the code for corridor walls and gets a similar assembly that complies with the performance spec for those walls. In less than 30 minutes, Wallmaster has laid out and determined all the wall types for the entire floor. Presumably, subsequent floors will be identical or similar to the initial one, and an entire tower can be designed in a couple of hours. No need for estimators to draw colored lines and construct wall types.
But there’s more. Suppose there are databases that track and maintain costs of material and labor specifically for the commercial market. These pricing programs are called “Labor Wizard” and “Material Wizard.” These programs are updated daily—Material Wizard by competing material yards, and Labor Wizard by historical data from previously completed projects. Both programs are cross-referenced by geographical location, with Labor Wizard also cross-referenced by construction type (health care, hospitality, gaming, etc.). By simply “talking” to the Wizard brothers, Wallmaster can determine an accurate cost for his wall types and, in short order, his entire build-out. All of this can be performed at the design phase. From here, it’s easy to theorize how this could impact project delivery, with the roles of contractors and subcontractors diminishing significantly.
I admit that there’s a lot of “supposes” and assumptions in all of that. But a perusal of the current literature on AI will bear out that these “supposes” may not be that far-fetched when applied to our industry and our profession. This is just my concept of what the future holds in store for estimators. But hey, as many of you have said, “it ain’t necessarily so.”
Vince Bailey is an estimator/project manager working in the Phoenix area.