The real cost of a stalled RFI in construction

Two and a half people. Establishing whether that is actually your number, on your own project data, starts at $15,000 and reports in six weeks.

And that is the cost when the process works. The cost when it stalls is different, and larger.

Ten days, and everything behind it

The same dataset puts the average first reply at 6.4 days and the median reply at 9.7 days. On projects running two years or longer, the median approaches ten.

An RFI is a question somebody in the field could not answer from the documents. Until it comes back, the work it concerns cannot proceed correctly. Neither can anything sequenced behind it. Ten days is not a ten-day delay on one task. It is a ten-day hole punched into a schedule that was built assuming the answer existed.

Here is the part that should be uncomfortable. Procore published its own benchmark in 2020, drawn from its customers’ projects, and found an average of 7.3 calendar days from RFI open to first response. Seven years after Navigant, on a platform built specifically to move this faster, the number sits between Navigant’s average and its median.

Buying the software did not close the gap. The software is a better place to keep the queue. The queue is still the problem.

Most RFIs are not questions

Navigant also broke down a single project’s roughly 4,000 RFIs by what they actually contained. The result is the most useful thing in the entire study.

Around 1,700 were routine correspondence, misfiled as RFIs. About 800 were duplicates or fishing expeditions. Nearly 500 were submittals and shop drawings in the wrong channel. Roughly 300 were quick answers, 200 were responses to non-conformance, and 150 were substitution requests.

Genuinely substantive issues: 350. Under nine percent. Of those, about thirty needed more than a week to answer.

That is one project, and Navigant presents it as a case study rather than a population estimate. But every contractor who reads that breakdown recognises it immediately, because the thirty questions that actually needed an engineer are buried in 3,970 pieces of traffic, and the person doing the burying is a project manager with a schedule to hold.

“We don’t do our project teams any favors by delivering a design or plan that is half-baked for them to figure out how to build.”

— Nathan Lingard, Director of Design Phase, Mortenson Construction

The RFI log is where a half-baked document meets a field crew, and it is the contractor who pays for the meeting.

The same problem, four weeks earlier

Bid week is the same failure in a different phase.

A construction CPA writing in ENR put private bid work at roughly five bids per win for general contractors, and six to ten per win on public work. Take the lower figure. Four of every five bids a firm assembles are written off.

The Bureau of Labor Statistics puts the median wage for a cost estimator in building construction at $81,490. Load that up and put it against five bids per win, and every job a contractor books is carrying the fully loaded cost of four takeoffs that went nowhere, most of it spent on takeoffs, scope-gap checking and reformatting numbers into somebody else’s bid form.

There is a story in this industry about an estimator talent gap. The federal data disagrees. BLS projects cost estimator employment to decline four percent between 2024 and 2034, with about 16,900 openings a year, nearly all of them replacing people who leave. The constraint is not a shortage of estimators.

It is what each estimator’s hours go to.

Five and a half hours a week, per person, looking for things

PlanGrid and FMI surveyed 599 construction professionals, roughly half of them at general contractors, and asked how their weeks were spent. The answers:

More than 14 hours out of a 40-hour week. Their summary: construction workers lose almost two full working days each week dealing with conflict, solving avoidable issues and searching for project information. It is a vendor-commissioned survey, and it is also the only direct measurement anybody has published.

Take just the first number and scale it. A forty-person office at $85 an hour loaded, 48 weeks: 10,560 hours, roughly $900,000 a year, spent finding information that already exists somewhere in the company.

Dodge Construction Network measured the other end of that in 2024.

Ten percent of margin. Multiply that against your own revenue and it will be a larger number than everything above it combined.

Contractors are buying the story, not the thing

Here is the gap that defines the opportunity.

AGC of America and Sage surveyed 951 contractor firms across 49 states in late 2025 and found 61% saying they use AI or plan to increase investment in it, up from 44% the year before. That is a compound question and it flatters the number, but the direction is real.

Now the money. The JPMorganChase Institute went through de-identified bank transaction data for 4.6 million small businesses and asked a harder question: which of them has ever actually paid for an AI service. Construction came in at 8.9%. Only transportation and warehousing was lower.

And of the firms that have deployed something, AGC found 45% put it into office and administrative functions first. Estimating gets 23%. Design and preconstruction gets 20%. So even the early movers are pointing it at the back office rather than at bid week and the RFI log, which is where this analysis says the money actually is. That pattern holds across every industry that runs the physical world, and construction shows it as clearly as any of them.

Donna Laquidara-Carr, Director of Industry Analytics at Dodge Construction Network, frames the window plainly: internal expertise with project data “is a competitive advantage now, but it is only a matter of time before it is a requirement.”

Where the first step goes

If most of the RFI log is traffic rather than questions, and most of bid week is assembly rather than judgment, then the first thing to fix is the drafting, not the queue.

That is a documentation problem, and it has two shapes.

Preconstruction and estimating. First-pass drawing review and scope-gap flagging, subcontractor RFPs drafted and tracked, historical bid data turned into faster and more consistent estimate narratives. The chief estimator’s team keeps the calls only they can make and stops retyping numbers into somebody else’s form. This runs alongside Bluebeam and whatever estimating platform is already in place. It is one department, measured, from $15,000.

The six weeks, concretely: a week timing how bids and RFI responses get assembled today, two weeks configuring agents against your own historical bid data and your own past responses, three weeks of the estimating desk working live bids with it, and a written comparison at the end. Nothing replaces Bluebeam or Procore, and the decision about going further comes off the numbers.

Field operations and project delivery. The heavier version: an embedded pod rebuilds the RFI pipeline against your real Procore data, so a PM receives a drafted, precedent-aware response to route rather than starting from a blank log, and the 1,700 pieces of routine correspondence stop arriving in the same channel as the thirty questions that need an engineer. That runs against live project data with a go or no-go gate before anything scales.

The full picture of what AI-native construction looks like runs from bid day to closeout, but no contractor should start there. Start with the phase where the writing is heaviest and the proof is fastest.

The number to go and find

Everything above is somebody else’s data. Here is yours.

Take your last completed project. Count the RFIs. Divide by the contract value in millions. If the answer is anywhere near 17, this article has just described roughly two and a half people on your payroll.

Then look at how many of those RFIs were actual questions.