Feasibility is the most expensive guess in construction
Sep 2026
Hunter Cole, Solution Product Manager (Commercial)How much of your project rests on a decision you made before you had real data? Feasibility sets the ceiling on everything that follows: design, cost, program, approvals. It’s also the stage where most teams work with the least.
Sep 2026
Hunter Cole, Solution Product Manager (Commercial))
Every project makes its biggest decision at its earliest point, when the site is still an assumption on a screen. Feasibility commits capital, sets client expectations, and shapes the whole job before anyone has walked the full ground. The work has always run in this order, and the industry has built its process around it.
That order carries a cost that shows up in the numbers.
Across more than 16,000 large capital projects, fewer than half land on budget, and only 8.5% land on budget and on time. Larger projects perform more poorly against both measures. Decision makers absorb these figures as the cost of doing business, and the feasibility call is where a meaningful share of that cost gets set.
The opportunity sits at the front of the job. Firms that see the site accurately at feasibility carry that advantage through every stage that follows. Scope tightens, numbers hold up, and mobilisation brings fewer surprises. The information already exists, and it can now reach the desk early enough to change the decision.
)
Feasibility runs on the oldest data in the project
Feasibility carries more weight than almost any other call in the bid process, and it gets the least attention of any of them. Senior people run it, and the client quote is built on whatever it concludes. The number that comes out carries enormous weight and almost no additional scrutiny.
The site data underneath it gets less examination still. Someone pulls the available mapping and whatever imagery exists, then makes reasonable assumptions to close the gaps. Those assumptions travel into the number, and the number becomes the promise the firm measures itself against for the life of the job.
Teams reach these numbers with care. The front of the project has been starved of dependable information for decades, and you cannot verify what you cannot see, so you assume. Those assumptions cost nothing at the desk and a great deal on site, where they surface as rework and lost margin.
Software outpaced the data beneath it
Firms have been working on this problem for a decade, and the effort has mostly taken the shape of software. Between 2020 and 2022, investment in construction technology reached $50 billion, and adoption followed the money. The average AEC firm now manages nearly 1000 different applications, each one promising to connect teams and hold a single source of truth. Many of those firms report the opposite result: data sitting in silos, trapped in formats that will not talk to each other and left out of decision making. The promise of all that data was clarity. The reality is more logins, more exports, and every new point solution adding a seat, a subscription, and another place for the truth to fragment.
That investment pays off where the inputs are strong. Two-thirds of early adopters report higher productivity from the technology they have put in place. The tools deliver when the data feeding them is consistent and current.
Feasibility sits outside that flow. The person making the earliest and highest-stakes call is usually working from the most disconnected picture, reconciling one tool’s version of the site against another and closing the remaining gaps by hand. The stack grew around the problem without ever reaching the front of it.
An industry that buys tools faster than it builds the ground beneath them arrives here, whatever the quality of the individual tools.
)
AI inherits the data you give it
AI is already doing real work in this industry. Firms use it to make job sites safer and to run leaner operations, and the results hold up. What it returns depends on the quality of what it is given, and confidence on that point is mixed. 57% of contractors raise concerns about accuracy and reliability, and 54% flag data security and privacy.
Adoption tells a similar story. Around 63% of AEC firms remain in what one study calls “pilot purgatory,” running experiments that have yet to produce results. About 27% use AI in live decisions, and 18% measure whether it returns anything. The appetite is running ahead of the infrastructure.
The firms getting value share one habit. They fixed their base inputs before they scaled the models, so the system had something dependable to reason over. Their advantage traces back to the quality of the ground truth beneath the tool.
Applied to site data that is years out of date, AI produces answers that arrive fast and carry real confidence while resting on conditions that have since changed. Each of those answers costs a little trust, and trust decides whether teams keep using the system at all.
Purpose-built AI is structured the other way around. When the model and the imagery beneath it come from the same controlled capture program, with consistent sensors and processing on a known schedule, an output can be traced back to a dated capture of a specific site. The answer carries provenance, and provenance is what lets a team defend it when someone challenges the number.
)
What actually reduces risks at the start of the job
You hear about a site on Monday morning, and by Monday afternoon you have studied current high-resolution imagery of it, checked the topography against the setbacks, and compared how the parcel has changed over the past several years. The encroachment that the old records missed is on your screen along with the structure that went up last spring, and none of it required booking a flight or a survey crew.
Capture programs that fly on a published schedule photograph the same ground several times over before anyone has a reason to look at it, so the imagery is waiting rather than commissioned, and the processing has already turned those pictures into something you can take measurements from.
The gain shows up downstream, where you assume less and the estimate sits closer to the site as it actually stands, so that when someone questions the number you can show them what it rests on. AI fed that same verified picture produces answers that survive the same scrutiny.
What this is worth
The value of better, more comprehensive feasibility is three-fold:
- Speed. You screen far more sites in the time it used to take to assess one, and you reach a confident go or no-go while competitors are still arranging their first visit. Aspire reported a 50% increase in bid opportunities closed after moving measurement off manual methods.
- Cost. You leave the bad deals early, before they eat margin, and you cut the rework that begins as a weak assumption and arrives later as a change order. Walter P Moore reported 20% time savings through reduced rework after producing bids and designs from current imagery.
- Risk. You carry evidence into every decision, which protects both the estimate and the promise you made to the client.
Feasibility is the cheapest point in a project to remove risk and the most expensive place to carry it forward. The firms protecting their margin are the ones giving that first call better information to work with.
)
Take a real first step today
Pull up your last three feasibility studies. For each one, split what you verified from what you assumed. Then ask a simple question about every assumption that made the list: could I have seen this on day one?
If the honest answer is yes, you are not looking at a data problem. You are looking at the most valuable process in your business, waiting to be fixed.