The Empire State Building was designed in roughly two weeks. Modern pre-construction often runs for years, and most of the damage gets baked in early: by the time a project is 20% designed, around 80% of its cost is already locked. Ediphi CEO Dustin DeVan, who built and sold BuildingConnected to Autodesk for $275m, argues the whole paradigm is wrong. Design should not lead. Cost and schedule should. Here is why the design-first model keeps producing $200m gaps between owners and GCs, and what a cost-first workflow would actually look like.
Construction loves to flagellate itself with one specific statistic. The Empire State Building went up in 13 months. A comparable tower today might take five to seven years. Everyone nods, mutters something about productivity, and moves on.
The number nobody quotes is the one that matters more. William F. Lamb drew up the Empire State Building in about two weeks. Pre-construction on a comparable modern tower can take years. So when people complain about how slowly we build, they are usually pointing at the wrong half of the project.
That is the argument Dustin DeVan, founder of Ediphi and previously the founder of BuildingConnected (acquired by Autodesk for $275m), made on the latest Bricks, Bucks & Bytes weekly rundown. And it has uncomfortable implications for almost everyone working in commercial construction today.
Construction has not actually gotten less efficient. Pre-construction has. And because pre-con sits upstream of every other decision, the rot spreads.
Pre-con is where the time goes
Construction got safer. Pre-con just got slower.
It is true that modern construction is slower than 1930s construction. It is also more or less correct that this is mostly fine. As DeVan pointed out, the Empire State Building cost 30+ lives. The Salesforce Tower cost zero. Modern safety protocols, MEP coordination, code compliance, sprinkler systems, accessibility, fire protection: all of it adds time, and almost none of it is up for negotiation.
The part of the timeline that has genuinely ballooned, without any equivalent justification, is what happens before anyone breaks ground. Schematic design. Design development. Construction documents. Bidding. Value engineering. More value engineering. Another round of value engineering because the first two did not get the price down enough. According to the AIA’s standard phase structure, this whole process exists to derisk construction. In practice, on a lot of commercial projects, it does the opposite.
Ask any GC who has worked a meaningful number of commercial jobs whether they have ever spent months in pre-con with an owner on a project that simply could not be built for the money available. They will all say yes. They will also tell you that nobody on the design side believed them when they flagged it at the start.
Why design-first breaks the budget
You cannot value-engineer your way out of a feasibility problem.
Here is the mechanic DeVan walks through. Designs today get drawn in Revit. By the time the design team hits 20-30% complete, somewhere in the region of 80% of the eventual construction cost is already baked in. Not the design cost. The actual construction cost. The materials, the systems, the spans, the floor-to-floor heights, the structural choices that downstream trades will price against.
Then the GC gets the drawings and starts pricing them. And the gap between what the owner thought the project would cost and what the GC says it will actually cost is rarely small. DeVan described one project where the owner thought a build would land at $450m. The GC priced it at $650m. That is a 44% miss, baked in before anyone wrote a check. The fix is a multi-month value-engineering exercise to claw back the difference by removing scope the owner wanted in the first place.
You shouldn’t work on a bunch of designs that were never feasible. It’s crazy.Dustin DeVan, founder and CEO, Ediphi
The deeper problem is that this is not a freak occurrence. Bent Flyvbjerg’s research on major transport projects found that 9 in 10 went over budget, with rail projects averaging 45% over the original estimate. Building projects are not immune. The standard industry response, value engineering, treats the symptom. The disease is that the pricing conversation arrives too late to influence the decisions that drove the cost.
What clients actually buy
Cost and schedule, dressed up as design
The provocation in DeVan’s argument is this: in commercial construction, clients are not really buying design. They are buying cost and schedule, and tolerating the design conversation because that is how the industry is structured.
Think about how owners actually evaluate projects. Multifamily developers run pro formas on unit mix: how many studios, how many one-beds, how many two-beds. Data center operators care about cost per megawatt. Healthcare owners think in exam rooms, patient rooms, and square footage by department. Hotel developers track cost per key. Parking garages get measured in cost per stall. None of that requires a Revit model. All of it can be priced before a single elevation gets drawn.
The implication is uncomfortable for the design profession. If most commercial work is cost-and-schedule first, then the conventional pre-con sequence (program, schematic, DD, then pricing) is solving the problems in the wrong order. The builder should be in the room before the designer, with a working price for the program, plus or minus 10%. Only once the owner accepts that envelope does it make sense to spend money on detailed design.
Hear the full debate on Bricks, Bucks & Bytes
Dustin DeVan and Foundamental’s Patric Hellermann go deeper on why pre-con is broken, what venture is funding, and where the real money is being left on the table.
Listen to the episodeWhere pre-con is not broken
Single-family residential, and anything where aesthetics are the product
The cost-first argument does not apply everywhere. DeVan was clear that this sits on a spectrum. At one end of that spectrum is a custom single-family home, where the owner cares about how the building looks more than almost anything else. Design has to lead there, because the design is what the client is buying. At the other end is a storage facility, where nobody cares what the units look like as long as occupancy is maximized and cost per door is low. The builder should lead.
Most commercial work, in DeVan’s read, lives center-right on that spectrum. Healthcare, data centers, multifamily for rent, industrial, logistics: all of these have measurable client outcomes that translate cleanly into program-level pricing. Built-to-rent housing, even though it is residential, lands center-right too. The places where design has to lead are smaller in volume than the industry’s process would suggest.
What the tooling gap looks like
Cost infrastructure, not better takeoff
Most of the venture money flowing into pre-con right now is going to AI takeoff. As Foundamental’s Patric Hellermann noted on the episode, takeoff is a real efficiency feature, but it sits inside a larger optimization. Quantities are an emergent property of the design. The big estimating misses are not on quantities; they are on unit costs and on scope items that got left out. AI that traces a footing faster does not change the critical path. It does compress estimator hours, but estimator hours were never the bottleneck.
What does sit on the critical path is what DeVan calls cost infrastructure. The ability to price a program parametrically, without a finished design, and to show how cost moves as the program changes. Today most GCs work either from benchmarks (comparing to past projects) or from detailed takeoff (after the design exists). The middle ground, where you can price a napkin sketch within 5-10% of what it will eventually cost, is largely unbuilt. That is a tooling problem, not an AI problem.
Layer intelligence infrastructure on top of that, capturing which decisions drive cost changes, and the conversation between owner, GC, and designer changes shape. The price becomes the anchor. The design becomes the response.
Design-led versus cost-led pre-con
| Dimension | Design-led (status quo) | Cost-led (DeVan’s argument) |
|---|---|---|
| Starting input | Schematic design from architect | Program and owner priorities |
| First pricing conversation | After DD, often months in | Within weeks, against the program |
| Typical owner-GC gap at first pricing | Often 20-50%+ over budget | Target 5-10% of agreed envelope |
| Value engineering rounds | Multiple, with scope cuts | Minimal, scope locked early |
| Best fit | Custom residential, signature buildings | Healthcare, data centers, BTR, industrial |
| Tooling required | Revit, takeoff, estimating | Parametric cost infrastructure |
Frequently asked questions
Roughly, yes. The chief architect, William F. Lamb of Shreve, Lamb & Harmon, produced the building’s drawings in about two weeks, leaning on earlier designs for the Reynolds Building in Winston-Salem and the Carew Tower in Cincinnati as a starting point. The full project, from first design to opening, took about 20 months. (Source)
By the time schematic design is wrapping up and design development is starting, the major cost drivers are already decided: building footprint, structural system, floor-to-floor heights, mechanical strategy, facade approach. Downstream trades price against those choices. You can change finishes and FF&E in later stages without moving the number much; you cannot easily change the structural grid. (Source)
Value engineering treats the symptom. By the time a VE exercise runs, the project is already over budget and the team is removing scope the owner originally wanted. A cost-led pre-con workflow tries to prevent the mismatch in the first place by pricing the program before design begins, so the envelope is agreed before any expensive drawing work starts. (Source)
Not by themselves. Takeoff is a useful efficiency feature, but quantities are not where estimates miss. The misses are on unit costs and on scope items that got left out. Faster takeoff makes the existing process less painful; it does not move the critical path. The bigger opportunity is in parametric cost models that can price a program without a finished design. (Source)
It applies most cleanly to commercial work where the client buys a measurable outcome (cost per megawatt, cost per key, cost per door). It applies less to custom single-family homes and signature buildings, where aesthetics are the product the client is paying for. Built-to-rent residential, despite being housing, falls on the commercial side. (Source)
Bent Flyvbjerg’s research on 258 major transport projects found that 9 in 10 went over budget. Rail projects averaged 45% over original estimates, and across the sample, owners paid 28% more than the original number just to get projects operational. Buildings are not as well-studied, but cost overruns are routine and largely driven by decisions made (or not made) before construction starts. (Source)
DeVan started his career as a scheduling engineer at Bechtel before founding BuildingConnected, a pre-construction bidding platform that was acquired by Autodesk for $275m in 2018. He is now CEO of Ediphi, a cloud-based pre-construction and estimating platform. His career has been spent inside the pre-con problem, which is why his read on what is broken is worth taking seriously. (Source)
Related Articles
How BuildingConnected was sold for $275m
McKinsey’s secrets to scaling construction tech
Smithsonian Magazine: When the Empire State Building Was Just an Architect’s Sketch
AIA: Defining the architect’s basic services
KPMG Foresight: Rising above (Flyvbjerg cost overrun research)
ENR: Ediphi creates platform for cloud-based precon estimates
Norwest: Ediphi, radically simplifying preconstruction bids
ProjectManager: Value engineering in construction
