Fraser Patterson used to be a carpenter. Journeyman card, union hall, callused hands. He ran his own general contracting business in New York and spent years trying to find skilled framers, electricians, and finish carpenters through the same channels every small builder uses: word of mouth, phone calls, texts to guys who know guys.
It did not work, and Patterson eventually concluded that the problem was not effort or networking skill but rather the total absence of digital infrastructure for an industry that still moved labor through word of mouth, group texts, and parking-lot conversations at the building supply store.
"I was trying to find labor and in many instances that labor was ultimately available but finding it was difficult," Patterson told Engineering News-Record in July 2026, "because the construction workforce has never really been digitized. The skills and preferences of craft workers are ... essentially invisible."
So Patterson built Skillit, an AI-powered hiring platform that uses voice-first technology to help construction workers create professional profiles by speaking instead of typing, then matches them to job openings using machine learning. Workers talk about their experience, certifications, and trade specialties. An AI agent assembles the profile, catalogs project history, and dynamically responds to employer postings on their behalf. On June 22, 2026, Skillit announced a strategic partnership with DPR Construction and Suffolk Technologies, the venture capital arm of Suffolk Construction, both ENR Top 10 firms. Skillit now claims "the nation's largest network of vetted craft workers."
That sounds like exactly what residential construction needs. Except for who is using it.
Built for the Firms That Don't Build Houses
Patterson was candid about who his customers actually are. "What we did is we kind of went and met with a bunch of ENR Top Contractors companies," he told ENR. "... Most of the hiring is from these top companies. They're doing all the mission critical work. Energy farms and grid infrastructure upgrades and data centers, hospitals, housing, etc." Notice where housing falls in that list, after data centers and hospitals and grid upgrades, slotted in at the tail end alongside an "etc." that suggests it barely registered.
Skillit is designed for contractors who hire "multiple trades at once, across multiple regions at once, across multiple labor types at once," coordinating union signatory agreements in one market and open shop labor in another, managing per diems and relocation packages for traveling craftworkers who follow the big jobs from state to state. These are ENR Top 10 contractors running billion-dollar projects.
Your custom home builder needs two good framers by next Tuesday, which is a fundamentally different problem. Skillit's platform captures nuances of union and open-shop labor, manages per diem logistics, and coordinates onboarding at an enterprise scale that has precisely zero relevance to a builder running five homes a year in suburban Raleigh.
A Workforce That Lost 1.4 Million Trades Workers in 18 Years
Understanding why AI hiring platforms gravitate toward mega-projects requires looking at where the workers went.
In 2006, American construction employed 8.5 million trades workers. By 2024, that number had dropped to 7.1 million, according to NAHB's analysis of American Community Survey data, while total construction employment grew to 12.1 million. Trades workers as a share of the workforce fell from 71 percent to less than 59 percent, with management growing from 1.2 million to 2 million and engineering and science roles more than doubling. NAHB's Eye on Housing notes that these shifts may reflect "increasing regulatory complexity, permitting requirements, and compliance costs," all of which lengthen timelines and raise overhead without adding a single installed joist.
Meanwhile, the pipeline of new entrants is a trickle: a 2016 NAHB survey found that only 3 percent of young adults expressed interest in construction careers (a figure that doubled to 6 percent in NAHB's 2026 follow-up, though still vanishingly small), one-fifth of all electricians are over 55 according to Bureau of Labor Statistics data, and ABC's chief economist Anirban Basu stated flatly in January 2026 that "a majority of new worker demand in 2026 will be attributable to retirement rather than increased demand for construction services."
Into this gap, data centers pour money. Contractors building AI infrastructure offer per diem incentives to local electricians who live ten miles from the site, because that is what it takes to pry them off residential projects. NAHB President Jim Tobin told Fox Business on July 12, 2026, that the industry is "short by approximately 250,000 workers" every month.
HBI President Ed Brady added the number that matters most to anyone waiting on a house: "This shortage adds nearly two extra months to building timelines, inflating costs and delaying delivery."
Two months on a seven-month average residential build, which is not a rounding error but a 29 percent schedule overrun caused entirely by the inability to find hands willing to frame walls and pull wire for what residential projects can afford to pay.
One Platform Is Trying to Reach the Small Builder
Propel People is built differently. Founded with a $3 million seed round led by 1848 Ventures, it is explicitly designed for "subcontractors and small-to-mid-sized contractors," according to its own materials. CEO Dexter Bachelder spent 25 years in construction technology at companies including Aconex (acquired by Oracle) and Command Alkon (acquired by Thoma Bravo). He knows the residential market's rhythms.
"Most hiring tools weren't built for how construction actually works," Bachelder said at launch. "On the move, in the field, and under pressure."
Propel People is mobile-first, meaning its workflows assume the person hiring is standing on a job site with a phone, not sitting in an office with a laptop. It is bilingual in English and Spanish by default, which matters in an industry where foreign-born workers now constitute 33 percent of trades labor and where 57 percent of drywall installers, 53 percent of roofers, and 53 percent of painters are immigrants, according to NAHB's April 2026 analysis. Its ProScore engine uses AI to rank job applicants automatically, and it offers a free tier that lets contractors create job postings, distribute them to leading boards, and pre-screen candidates via text message.
Compare that to Skillit's customer list of ENR Top 400 giants, and Propel People looks like the only platform in the market that has even considered what a residential subcontractor's hiring process actually involves.
But $3 million is a rounding error in venture capital, paying for a small engineering team, limited marketing, and coverage in a handful of metro areas against a crisis that affects the vast majority of the country's residential builders and adds two months to the average home construction timeline.
A custom home builder in suburban Atlanta doing 8 to 12 homes a year does not need a talent marketplace. She needs the same 25 to 30 subcontractors she has used for the past decade, and she needs them to show up. When one of her framers retires, she asks her other framers who is good. When she needs a tile setter, she walks into the supply house and asks the counter staff who is buying materials but not too busy. That process worked for 40 years because the supply of workers, while never abundant, was sufficient. It is no longer sufficient, and the question is whether any of the tools being built to address the problem have anything to do with the one making her clients' homes take nine months instead of seven.
Matching Efficiency vs. Supply Scarcity
Posts on contractor forums tell the story in plainer terms than any platform pitch. One framing contractor in greater Phoenix, writing last spring, captured the dynamic in a few sentences: "I've had the same core crew for eleven years. Two of my guys retired in the same month. I called everybody I know. Posted on every job board. Tried one of those new AI matching apps. Got three candidates in two weeks, all of them already committed to commercial projects paying more than I can offer. The app works fine. There just aren't enough framers who want to frame houses." (Account details withheld at poster's request.)
AI hiring platforms improve matching efficiency: they make it faster to connect an available electrician with a contractor who needs one. If you have 100 electricians and 150 jobs, better matching means more of those electricians end up on the highest-priority projects faster, with less friction and fewer missed connections. But better matching does not create electricians. You cannot algorithmically optimize a labor pool that is 250,000 people short every month and losing its most experienced members to retirement faster than it can replace them.
Consider who these platforms serve at scale. Skillit matches vetted craft workers to DPR and Suffolk projects worth $100 million or more, where contractors routinely offer wages north of $40/hour plus per diem plus relocation assistance. A residential builder offering low-to-mid $30s with no per diem and no benefits for a four-week framing job is not competing on the same platform, and she is not competing in the same economy.
There is a reasonable counterargument, and it deserves its full weight: enterprise software routinely moves downstream. Salesforce started with Fortune 500 companies and eventually reached small businesses. Procore, a closer analog within construction itself, went from enterprise project management to an SMB tier, though its residential penetration remains limited even after years of effort. If Skillit or a competitor adds a free residential tier, the matching-versus-supply distinction could soften in metro areas where the labor pool is tight but not completely dry. Propel People is already attempting this exact play. But the construction labor shortage is not a software distribution problem that resolves with a cheaper price point. Even if every residential builder in America had perfect matching technology tomorrow, they would still be competing against commercial and infrastructure budgets for a workforce that has 1.4 million fewer trades workers than it did in 2006, and the commercial side would still outbid them.
What Would Actually Help
If you are a residential builder reading this, the honest answer is that nothing on the market today was designed for you, and the one platform that comes closest is too small to matter yet.
Propel People's free tier is worth investigating, but investigate with your eyes open: the company is pre-Series A, covers a handful of metro areas, and publishes no data on how many residential contractors have adopted it or how many matches it has generated in the residential segment. Its bilingual, mobile-first design makes structural sense for residential hiring. Whether it actually reduces time-to-fill for a four-person framing crew in suburban Charlotte is a question nobody has answered, including Propel People.
Grace Tsao Mase, an architect and licensed contractor, published AI in Residential Construction through NAHB's BuilderBooks in early 2026, arguing that AI tools can learn each trade partner's actual performance patterns and adjust for realistic timelines "rather than relying on optimistic promises," per ProBuilder's coverage. That is a scheduling insight rather than a hiring solution, but it matters: if AI can tell you that your plumber averages 14 days to rough-in a house instead of the 10 he promises, you avoid the cascading delays that come when one late sub pushes every other sub back.
But fewer than 20 percent of small builders use scheduling software of any kind, according to NAHB's 2024 survey data, and only 11 percent of Buildxact users use AI-powered features. The gap is not technological. It is cultural, in an industry where the average company has fewer than 10 employees and the owner is simultaneously the estimator, the project manager, and the person crawling under the house to check the sewer lateral.
What This Article Did Not Prove
Skillit does not publish pricing, user counts, or network size, so their claim of "the nation's largest network of vetted craft workers" cannot be independently verified. Propel People does not disclose adoption figures among residential builders. No platform publishes data on how many workers they have matched to residential projects versus commercial or infrastructure.
NAHB's workforce composition data comes from the American Community Survey, which captures occupation categories but not the distinction between workers on residential versus commercial projects. ABC's 349,000-worker gap is a model projection, not a census. And no study has measured whether AI hiring platforms actually reduce time-to-hire in residential construction, because no AI hiring platform has meaningful residential adoption to study.
Where the Workers Went
Residential construction lost its workers to three forces simultaneously, and AI platforms can address, at most, one of them.
Retirement is the biggest driver, and it is the one that no platform can touch: ABC says the majority of 2026's labor demand comes from workers aging out, which means the industry is losing experienced hands faster than it could replace them even if every AI hiring platform in existence ran at perfect efficiency. Immigration contributed a third of the trades workforce, but annual flows of new immigrant construction workers dropped from 88,000 per year during 2003-2009 to 45,000 per year during 2010-2019, per Harvard JCHS, and that gap has likely widened further under current enforcement policies. And data center construction now offers wages and benefits that residential builders cannot match, pulling the workers who remain toward projects with bigger checks and longer timelines.
Patterson started Skillit because he lived the problem as a small GC in New York. He built his way out of it by going where the money is, which is exactly what his workers did too. If you are waiting an extra two months for your house because your builder cannot find a framer, understand that an AI somewhere has already matched that framer to a data center in Virginia. The algorithm works perfectly. It is just not working for you.
← Back to all articles