A couple in Denver offered $612,000 on a 1928 bungalow. They waived the inspection contingency because their agent told them it was the only way to win in a fourteen-offer bidding war. The sellers accepted. Thirty days later, their own inspector found the brick masonry foundation was spalling so badly that a structural engineer estimated $38,000 in repairs. Mortar joints had been deteriorating for decades, invisible behind freshly painted basement walls and a seller's disclosure that checked "No" on foundation issues. An AI tool built by a Denver inspection company could have flagged the risk in seconds, for free, before the offer was ever written. Nobody used it.
Free. Twelve seconds. Nobody used it.
That scenario is not hypothetical, and it is happening across every competitive market in the country right now.
The Numbers Behind the Surprise
InterNACHI estimates 2.1 million home inspections happen annually in the United States, a slice of the $4.9 billion building inspection market tracked by IBISWorld, and 86% of those inspections uncover problems. Roof damage leads at 19.7%, followed by electrical issues at 18.7% and window defects at 18.4%, a top-three list that has barely changed in a decade because roofs, wiring, and windows are the three systems most exposed to weather and age, and also the three systems that sellers are most likely to cover with a fresh coat of paint and a prayer. Foundation repair alone averages $5,165, with a range stretching from $2,218 to $8,111 depending on the severity and the region and the contractor's mood that Tuesday.
Run the multiplication: if 86% of 2.1 million inspections find problems, that is 1.8 million transactions where the buyer suddenly learns the house costs more than they thought. At an average renegotiation of $14,000 per deal, the total annual post-inspection price adjustment across the U.S. housing market comes to roughly $25.3 billion. Not the cost of repairing defects, but the money that changes hands after someone opens a crawl space and discovers what the listing photos conveniently framed out of every carefully angled shot.
Deals are collapsing at record rates, with Redfin reporting that 13.7% of home-sale agreements fell through in February 2026, the highest rate for any February on record, with more than 42,000 deals dying that month alone. In December 2025, the cancellation rate hit 17.6%, the worst outside the pandemic. Each collapsed deal leaves behind a trail of sunk costs: the $333 median inspection fee, the $400 appraisal, attorney hours, opportunity cost on both sides, and a seller who now has to relist a home that publicly failed to close, carrying the invisible scarlet letter of a transaction that fell apart at the finish line because something ugly surfaced too late for anyone to pretend it did not exist.
Meanwhile, Buyers Are Flying Blind on Purpose
Twenty-one percent.
That is the share of buyers who waived their inspection contingency entirely, according to the NAR REALTORS Confidence Index from September 2025, up from 17% a year earlier. In competitive markets from the Bay Area to Boise to Austin, agents routinely advise waiving inspections to make offers more attractive, a strategy that works brilliantly right up until the moment you discover the sewer line was replaced with orangeburg pipe in 1962 and has been slowly collapsing under your front yard for the past decade. Buyers comply because they want the house, and then they own whatever it has been hiding behind its drywall, under its shingles, inside its electrical panel, in the crawl space where the flashlight beam catches something that looks a lot worse than the listing suggested. Nobody who waives an inspection believes their house has a $38,000 foundation problem, which is precisely why they end up with one, staring at a structural engineer's estimate that exceeds their entire emergency fund while their agent quietly searches for the right words to say over the phone.
What Alpine Intelligence Actually Does
Alpine Building Performance, a decade-old inspection company in Denver, launched a free AI tool in March 2026 that tries to solve this problem before the offer hits the table. Co-developed by founder Andrew Sams and engineer Mason Minor, the tool is powered by ChatGPT and works like this: upload MLS listing data along with whatever supplementary documents you have: property disclosures, permit reports, neighborhood plat maps. The AI cross-references the property's construction era, materials typical of that era in that specific region, known failure modes for those materials, and regional building code history.
A pre-1940s home in Denver's Park Hill neighborhood? Flagged: 87% probability of lead-based paint, mortar deterioration in brick masonry foundations, knob-and-tube wiring remnants. A 1970s ranch in Lakewood gets flagged for polybutylene plumbing, aluminum branch wiring, and original single-pane windows with failed thermal seals. None of this requires setting foot on the property, which is the entire value proposition in a market where physical inspections happen after the contract is signed and the emotional commitment is already irreversible. It is pattern recognition applied to building vintages, and the patterns are real because builders in a given decade, in a given city, used the same materials, followed the same codes, and made the same mistakes.
Alpine is explicit that this is not a substitute for a physical inspection. "Meant to help the agent set expectations," is how they describe it. But for the 21% of buyers currently waiving inspections altogether, having any defect intelligence is better than having none.
Two Other Tools Worth Knowing About
Hosta a.i., an MIT spin-off founded by Henriette Fleischmann and Rachelle Villalon, takes a different approach. Feed it property photos and it returns precise measurements, floor plans, 3D models, bills of materials. More relevant for pre-offer diligence: it evaluates visible condition of materials, identifies damage, and flags risks like inadequate sprinkler coverage or flammable materials near heat sources. Hosta works with insurers and mortgage lenders who need fast, consistent property assessments without dispatching a human to every site.
Homegenius IQ, from homegenius Real Estate, analyzes listing photos using computer vision to identify objects, finishes, and materials, then assesses their condition and how they impact property value. It is designed for lenders and appraisers more than individual buyers, but the underlying capability, pulling condition data from photographs that are already public, points toward a future where every listing carries a machine-readable risk profile alongside the Zillow Zestimate.
The $25.3 Billion Thought Experiment
Now for the math. Suppose an AI pre-screening tool could accurately predict 70% of major defects before offers are written, and suppose those predictions reached buyers before their offers were submitted rather than after the emotional and financial commitment was already locked in. Everything changes.
Buyers who know about the foundation adjust their initial offer downward by $30,000 instead of discovering it after going under contract and entering an adversarial renegotiation that kills 13.7% of deals outright. Buyers avoid bidding on properties with dealbreaker defects entirely, saving themselves the inspection fee, the appraisal, the attorney, the emotional wreckage. Agents set realistic expectations on day one, steering their clients toward properties that match both budget and risk tolerance, instead of managing a crisis at day twenty-eight when the inspection report arrives and the buyer realizes the dream house has a nightmare lurking in the crawl space that nobody mentioned during the open house because nobody thought to look. Sellers who know their problems will be flagged might actually fix them before listing, or at least price accordingly, which is the most efficient outcome for everyone.
Those 21% of buyers who waived inspections would, for the first time, get something. Not perfect data. Not a substitute for a licensed inspector with a thermal camera and a ladder, but a data-driven flag that says: this house, built in this year, in this neighborhood, with these materials, has a statistical likelihood of these specific problems. That is more than they have now, which is nothing.
Why This Might Be Glorified Actuarial Tables
Any experienced agent in Denver already knows that pre-1940s bungalows have foundation concerns and that 1970s homes come with polybutylene plumbing. They have been telling buyers this for thirty years without needing a ChatGPT wrapper. What Alpine is really doing, the counterargument goes, is packaging local market knowledge that good agents already possess into an interface that makes it feel more rigorous than it is. An 87% probability of lead paint in a 1935 house is not a prediction. It is a lookup table. Dressed up.
Fair point, until you think about who is actually buying homes.
No independent testing validates the accuracy of any of these tools. Alpine, Hosta, homegenius: none have published third-party accuracy benchmarks against actual inspection outcomes. We are trusting marketing claims and MIT pedigrees, which is a thin foundation for a tool meant to find bad foundations, and the irony of that sentence is not lost on anyone who has watched a startup promise disruption while shipping a chatbot wrapper around a lookup table.
Regional training bias compounds the accuracy question, because Alpine was built on Colorado Front Range data. A buyer in Houston, where expansive clay soils create entirely different foundation failure modes, or in coastal Florida, where salt air and hurricane loads dominate the risk profile, gets predictions trained on the wrong geology and the wrong building stock. Generalizing from Denver to the Gulf Coast is like forecasting San Francisco weather from Phoenix data, a comparison that sounds flip until you realize that the most common foundation failure mode in Houston involves expansive Beaumont clay swelling against post-tension slabs, a failure mechanism that does not exist in the Colorado Front Range and that Alpine's training data has never encountered.
Most dangerous of all is the false confidence problem: a buyer who gets an "AI says it's fine" report may underinvest in the physical inspection or skip it entirely with even more certainty. When prediction creates complacency, the tool actively harms the person it was designed to help.
What You Should Do With This
If you are buying a home in 2026, here is the practical takeaway.
First: do not waive the inspection. The 21% of buyers doing this are absorbing catastrophic risk to win bidding wars they could win other ways: escalation clauses, flexible closing dates, seller leaseback offers. If your agent says waiving the inspection is the only path to winning the bidding war, the bidding war is not your problem, your agent is, and you should find one who can structure a competitive offer without asking you to gamble your family's largest financial commitment on the hope that nothing is rotting behind the walls.
If you are in a market where waiving feels unavoidable, use an AI pre-screening tool as a minimum due diligence layer. Alpine's tool is free, so upload the listing, read the risk flags, and factor them into your offer price before you write a number on the contract. It will not catch property-specific problems like termite damage behind a bathroom wall or a furnace that runs but should not. But it will catch era-based, region-based patterns that are statistically reliable, and knowing about them before you bid is strictly better than not knowing.
Ask your inspector whether they use AI-augmented reporting, because companies like Alpine already combine traditional boots-on-site inspection with drone imaging, infrared scanning, and AI analysis. The cost differential is small, but the information differential between an AI-augmented report and a clipboard-and-flashlight walkthrough is enormous.
For agents: run AI pre-screening on your listings before your buyers do. If the tool flags a probable $5,000 roof issue, you want to be the person who mentioned it first, not the person who said "I'm sure it's fine" three weeks before the inspection report lands on your desk.
Limitations of This Analysis
The $25.3 billion figure is an estimate built from industry averages. The 2.1 million inspection count comes from InterNACHI and covers formal inspections only, not the broader universe of pre-listing inspections, insurance inspections, or informal walkthroughs. The $14,000 average renegotiation figure comes from a single industry compilation and likely varies significantly by market, with expensive coastal markets pulling the average up and Midwest markets pulling it down. The 86% defect rate includes everything from missing outlet covers to structural failures, and the vast majority of flagged issues are minor. The 70% prediction accuracy figure in the thought experiment is speculative, chosen to be aggressive but not absurd. No AI inspection tool has published verified prediction accuracy against actual inspection outcomes.
The Redfin cancellation data and NAR waiver data come from different time periods (February 2026 and September 2025, respectively), and the gap between NAR's 6% cancellation rate and Redfin's 13.7% reflects different methodologies, not a doubling of cancellations between surveys.
Jake Kowalski covers construction technology, tools, and the machines that are changing how we build. He has opinions about inspection reports and keeps them to himself until paragraph three.