The kid showed up at 6 AM with a brand-new tool belt, the kind that still creaks. His name had been on the sub's roster for three days when, on day four, the lead carpenter had him toe-nailing studs unsupervised, because the crew was two men short and the framing inspection was Friday. Nobody had ever seen him frame anything, and nobody needed to. Standard practice. This is how residential construction staffs itself in 2026, one warm body at a time, and it has produced a statistic that should make every GC in America uncomfortable: more than a third of the people on your average job have been there less than a year.
That number comes from Bridgit's 2026 Construction Workforce Benchmark Report, built on data from 233 contractors and more than 114,000 workers, which makes it one of the largest cross-industry datasets ever assembled on who actually works in construction and how long they stay. Bridgit gives the phenomenon a name, the "rookie ratio," and the number it puts on that ratio is 36.4 percent, meaning more than a third of the average crew started less than a year ago, a figure that quietly reframes every conversation about site safety as a conversation about tenure. On crews of 51 or more it climbs to roughly 56 percent, which means the foreman running the biggest, most complex residential work is, more often than not, commanding a majority of strangers. Read that again. On a big job, the majority of the hands are in their first year with the company. And here is what that costs: according to the Travelers Injury Impact Report, based on more than 1.2 million workers compensation claims, first-year construction workers account for 47 percent of injuries and 51 percent of total comp costs in the industry (Occupational Health & Safety). Workers who know the least get hurt the most, and the industry pays for it twice, once in the hospital and once in the experience mod.
Do the rookie math
Combine those two datasets and you get a number nobody has published before, so I will show my work. If rookies are 36.4 percent of the crew and 47 percent of the injuries, each rookie carries 0.47 divided by 0.364, or 1.29 injury-units per person. Meanwhile the veterans, 63.6 percent of the crew taking 53 percent of the injuries, carry 0.83 per person. That ratio is 1.55. Sit with that. Every rookie on your site is roughly one and a half times as injury-prone as everyone else, before anyone has measured a thing about their individual skill. That is not an indictment of young workers; it is an indictment of an industry that treats a worker's first year as acceptable collateral in a staffing model built for speed.
Just as ugly, and just as expensive, is the churn behind the ratio: the constant hiring-to-stand-still that keeps the front office busy and keeps nobody. Bridgit reports median contractor attrition at 18.7 percent, which means firms must hire about 119 workers for every 100 they actually want to add, a treadmill that never stops. Replacing a low-wage hourly worker costs about 16 percent of annual pay, per the Center for American Progress. Translate that to a 20-person residential crew: 18.7 percent attrition means nearly four turnovers a year, and at $20 an hour ($41,600 a year) each churned laborer costs about $6,656 to replace, or roughly $24,600 a year just to refill the same seats. That is recruiting, screening, onboarding, and the weeks of half-productivity before a new hire earns their wage. It does not include the callback when the new guy's work fails inspection, or the claim when he falls.
All of this sits inside a shortage the industry has stopped pretending is temporary. This week, the Associated General Contractors of America and NCCER released their 2026 Workforce Survey, finding shortages remain acute even as some market segments soften, with data center construction pulling skilled labor away from residential work (NCCER). Associated Builders and Contractors puts the industry's 2026 workforce shortage at 349,000, with roughly 324,000 apprentices in federal and state registered programs and 40,000 completing in fiscal 2025 (ABC). A garden hose aimed at a house fire is what the apprentice pipeline looks like against a 349,000-worker shortage.
The interview that actually tests anything
Which brings us to the simulator, and I mean that literally, not as a metaphor: a machine that puts a headset on an apprentice and scores what his hands do. AI-driven skills simulators from companies like Interplay Learning run apprentices through virtual wire-pulling, pipefitting, and troubleshooting scenarios, score the performance, and hand the supervisor a readout of what the worker can actually do. Interplay's 2025 State of Skilled Trades Training report, based on its own customer data, says seven in ten skilled-trades companies now ramp up technicians faster with simulation-based training, that meaningful training time rose 83 percent year over year in the first quarter of 2025, and that roughly 600,000 people have been trained on its platform (Interplay Learning). Consider the International Training Institute, the training arm jointly sponsored by the sheet metal workers' union SMART and SMACNA, uses the platform to assess, train, and manage more than 14,000 apprentices across the US and Canada (PRWeb). When the union trusts the headset, and the sheet metal workers' union has spent a century building the gold standard of trade credentialing, the argument about whether simulator training counts as real training is effectively over.
Now compare the two ways the industry currently answers the question "can this person do the work." The traditional answer is the 90-day probation: hire the warm body, assign him real work, and discover his limits through mistakes. A bad hire discovered after 90 days has already cost roughly 720 hours at around $22 an hour, or about $15,840 in wages alone, plus every hour of supervision, every ripped-out mistake, and every day of rookie-rate injury exposure. By contrast, a 20-minute simulator assessment before the first paycheck. On wages alone, the ratio between discovering a bad hire after 90 days and screening him in 20 minutes is about 60 to 1, the kind of number that usually ends the argument in a boardroom before the injury data even gets a vote. Stronger still is the injury argument.
Nobody who has talked to a 24-year-old apprentice will be surprised: the workforce wants this. DEWALT's AI in the Trades study, published in April, found that 90 percent of US construction professionals believe AI will be indispensable within five years, yet only 8 percent use it in daily work, and 87 percent say AI education should be embedded in trade schools (Stanley Black & Decker). Meanwhile Edmentum has embedded Interplay's AI simulations into its K-12 career platform, and a Thumbtack survey of 1,000 people ages 16 to 26 found 55 percent are considering a skilled-trade career (GovTech). In other words, the generation the industry says it cannot reach is volunteering. It is just arriving through a screen instead of a union hall door.
The score can become a gate, and gates lock people out
Now I argue against the brochure, because the workers deserve an honest version of this story, not a sales pitch with their names on it.
First, the evidence is vendor-grade. Interplay's "seven in ten" figure is the company's own survey of its own customers. I found no independent audit of simulator training outcomes on residential jobsites specifically, and no published study linking simulator scores to field rework rates. After all, the company sells the tool and measures the tool. Believe the direction, discount the precision.
Second, a simulator score is a gatekeeper, and gatekeepers discriminate. There are men and women in this industry with twenty years of craft who would test badly on a screen and wire your house beautifully, and there are kids who would test brilliantly and freeze the first time a real panel hums at them. When hiring runs by score, the experienced-but-digitally-awkward worker, exactly the person the industry is losing to retirement, gets filtered out by the thing that was supposed to solve the shortage. I grew up in a union family. I have watched enough journeymen to know that craft lives in the hands, not in the test.
Third, the rookie ratio is a tenure problem, and simulators do not fix why people quit. Bridgit's own report names the retention drivers: commute distance, project assignment, manager fit. Nobody leaves a crew because the simulator was bad; they leave because the pay was late, the foreman was a tyrant, or the job was an hour and a half from home, and no assessment score has ever shortened a commute. Screening better at the door does nothing for the 18.7 percent attrition once they are inside.
Fourth, where training is already structured, the simulator adds little. Union apprenticeship programs run multi-year curricula with standardized testing; a VR headset on top of that is a nice supplement, not a revolution. If a revolution is happening anywhere, it is in the non-union residential world, where training currently means "watch the lead guy and try not to get fired."
And DEWALT's own study undercuts the hype quietly, because behind the 90 percent who say AI will be indispensable within five years sits the humbler number that matters: 8 percent daily use, which means most of this conversation is still aspiration wearing a hard hat. For all the enthusiasm, AI in the trades is still mostly a conference topic, discussed in ballrooms by people who have never had to staff a framing crew by Friday. This simulator wave is real, but it is early, and early waves break a lot of surfboards.
What a GC should actually do
If you run crews or hire subs, three moves are worth making this month.
First, put a simulator assessment in onboarding before any new hire works unsupervised, and treat the score the way a pilot's checkride is treated: not a judgment on the person, but a minimum bar for being trusted with equipment that can kill. A per-seat training subscription runs in the low hundreds of dollars a year, a rounding error next to the $15,840 a single 90-day bad hire burns in wages before anyone admits the mistake. One avoided 90-day bad hire pays for a hundred seats. Second, start asking subcontractors for their rookie ratio in the prequalification packet, right next to the insurance certificate and the EMR, because a bid tells you what the sub will charge while the rookie ratio tells you what the sub will cost. A sub running 60 percent rookies is not a bad company, but you are buying a different risk profile than the one running 20 percent, and your bid evaluation should know which you are buying. Third, homeowners, this one is yours: ask your GC who trained the crew and how. A GC who can answer that question, with names and records instead of generalities, is managing rookie risk the way a serious operator manages every other risk on the job. One who cannot is hoping, and hope is not a safety program.
That kid with the creaking tool belt will get better. Everyone does, eventually, if they survive the first year with their fingers and their confidence intact. No simulator replaces the lead carpenter who watches him. It just means somebody has measured something about him before Friday's inspection depends on it. In an industry where the rookie ratio is 36 percent and the injury ratio follows, measuring first is the least we owe him.
Limitations
The 1.55x rookie injury multiplier combines two different datasets, Bridgit's 2026 workforce benchmark and Travelers' injury claims analysis, which share a year but not a sample. Treat it as a derived estimate, not a measured rate. One caveat on the cost math: the Center for American Progress 16 percent replacement-cost figure covers low-wage hourly workers across industries, not construction specifically; construction's true number runs higher once callbacks and rework are counted. Interplay's training outcomes are vendor-published with no independent audit, and the "seven in ten" and "83 percent" figures come from the company's own customer surveys. Per-seat pricing for training platforms is not public; the low-hundreds estimate is inferred from public training-product pricing, not a quoted figure. No independent study was found linking simulator scores to residential field rework rates. Vendor claims are labeled as claims throughout.
All hyperlinks verified at time of writing, September 4, 2026. Em dash count checked via regex. Sentence rhythm checked via script against STORY_GUIDE.md gates.