A pattern book is one of the oldest tools in architecture. In the eighteenth century, builders carried slim volumes of engravings from England to the colonies, and entire neighborhoods rose from a shared set of proportions, dormers, and cornice profiles that gave American towns their visual coherence without requiring every house to be designed from scratch. Elegant in its simplicity: design well once, build many times, and let the repetition carry quality forward instead of reinventing it at every lot line.
On July 11, 2026, Congress revived the concept. The 21st Century ROAD to Housing Act became law with 358 votes in the House and 85 in the Senate, and among its provisions is a competitive grant program for communities to develop "pattern books" of preapproved housing designs. The logic tracks: if the median single-family permit in Los Angeles takes 297 days and a quarter of projects require five or more resubmissions across as many as thirteen city departments, then eliminating the design review entirely for homes that already meet code is not a marginal improvement. It is a structural one.
Several cities are already doing this. The results are extraordinary, and the limitations are immediate.
Thirteen Plans in Arkansas
In June, Rogers, Arkansas, launched its Pattern Zone Program with thirteen residential designs ranging from cottages to duplexes. Plans are free to download for eligible parcels. Each has already passed through Community Risk Reduction for building code compliance, which means a builder who selects one skips the back-and-forth that program manager Amber Long describes as "what causes more time and money." Materials can vary, but the layout stays fixed, and each plan includes a defined range of modifications before a full review is triggered.
Ohio published a statewide housing toolkit with two architect-designed plans, a narrow single-family home and a front-back duplex, engineered to fit approximately eighty percent of the state's infill lots. Overland Park, Kansas, launched a Portfolio Homes program in November 2025 with cottages and small single-family designs, waiving permit fees entirely for approved applications. LA County partnered with Canibuild, an AI-powered preconstruction platform, to create a portal for wildfire rebuild homeowners to browse preapproved plans, check zoning compliance, and visualize how a design fits their specific lot.
In every case, the reported results match the premise. Pew's analysis found that preapproved plans reduce upfront design costs by $5,000 to $10,000 per home, eliminate weeks or months of permitting timeline, and lower the barrier for smaller builders and nonprofits who cannot absorb the cost of custom architectural drawings for every project. HBI and NAHB estimate the carrying cost of a delayed single-family home at $1,333 per month. In Los Angeles, where a quarter of permits take more than 754 days, that delay cost alone can exceed $33,000 before a shovel touches dirt.
The Math That Nobody Is Doing
Thirteen plans.
Rogers has thirteen. Ohio has two. Overland Park has a handful of cottages and duplexes. Each was drawn by a licensed architect, reviewed by city staff, and calibrated to a specific subset of lots in a specific jurisdiction under a specific edition of the building code with specific zoning overlays, setback requirements, height limits, lot coverage ratios, and parking minimums.
There are more than 35,000 incorporated places in the United States, most with their own zoning ordinances. A single municipality can contain dozens of residential zoning districts, each with different dimensional standards. Lot shapes, sizes, orientations, and topographies vary enough within a single district to make a nominally compliant plan physically unbuildable on half the parcels it theoretically fits.
A pattern book with thirteen plans is a pamphlet in a library that needs thousands of volumes. Whether there is any way to write those volumes without spending the next century drawing them by hand is the only question that matters.
The Tools Exist in Pieces
They do. In Surrey, British Columbia, the city runs a digital compliance tool that checks submitted building plans against local zoning regulations using AI, returning a pass-or-fail report in minutes instead of weeks. Archistar, an Australian company, offers AI PreCheck to more than thirty cities worldwide, evaluating PDF, CAD, or BIM submissions against local codes and producing standardized compliance reports that tell the applicant exactly which rules their design violates. AutoReview.AI, developed through a decade of research at the University of Florida, has built a system that reads the entire 800-page Florida Building Code and automates the checking of site and building plans against all applicable land development and building regulations.
These tools check designs. They verify compliance. They are the back end of a pattern book system, the mechanism that confirms a plan meets code before it enters the preapproved catalog.
What they do not do is generate the designs.
Generative design tools exist too, primarily in the commercial and industrial sectors. Autodesk's platform can produce thousands of design variations from a set of constraints, optimizing for structural performance, material cost, or energy efficiency. Smaller firms are applying similar approaches to residential floor plans, generating layouts from room counts and square footage targets and adjacency requirements. In every case, the output is a design that must then be checked for code compliance by a separate system, reviewed by a licensed professional, and submitted through the conventional permitting process, which is exactly the bottleneck the pattern book is supposed to eliminate.
The Bridge Nobody Has Built
Imagine a system. It ingests a municipality's zoning code, building code adoption status, and parcel database. For each residential lot, it generates a set of code-compliant designs, checking each one against setback requirements, height limits, lot coverage, floor area ratios, parking minimums, fire separation distances, and every other dimensional standard that governs what can physically and legally be built on that specific piece of ground. What emerges is not a universal plan that works on eighty percent of lots. It is a plan that works on this lot, verified against this code, in this jurisdiction, today.
Every component of that system already exists as proven technology. AI compliance checking exists. Generative floor plan design exists. Parcel database access exists. GIS-based zoning interpretation exists. Integration is the gap. Nobody has stitched these pieces into a pipeline that takes a parcel ID as input and returns a preapproved design as output.
I can think of several reasons nobody has done it. Data is fragmented. Zoning codes are written in natural language that varies wildly between jurisdictions, and parsing them into machine-readable rules is expensive and error-prone. Liability attaches to preapproved plans in ways that make cities cautious: if a preapproved design fails in construction because the AI misread a code provision, the municipality faces exposure that a human plan reviewer's professional judgment would have caught. Architects worry, reasonably, that automating plan generation threatens the profession's economic base at the entry level where young designers earn their hours drawing exactly the kind of small residential projects that pattern books are designed to replace.
These are real obstacles. But they are not fundamental barriers, and the incentive structure created by the ROAD Act makes them look increasingly temporary.
What the ROAD Act Actually Funds
The law creates competitive grants for communities to develop pattern books. It does not specify how those pattern books must be created. An Arkansas city hiring an architect to draw thirteen plans is eligible. A California county deploying an AI system that generates five hundred parcel-specific designs is, on the plain language of the statute, equally eligible.
The law also eliminates the permanent chassis requirement for manufactured housing, which is expected to reduce building costs by $5,000 to $10,000 per unit and unlock new form factors including basements and second stories that the chassis made structurally impractical. That provision interacts with pattern books in an underappreciated way: manufactured housing has always been a pattern book product, a factory building the same design hundreds of times. Removing the chassis requirement means manufactured designs can now look and perform more like site-built homes, which means the distinction between a "manufactured home" and a "preapproved plan built on site" gets thinner with every regulatory concession.
Single-family housing starts dropped to 895,000 in June, the third straight monthly decline, with permits falling to their lowest level in ten months. Mortgage rates hit 6.55 percent, an eleven-month high, and unsold new homes sit near levels last seen in late 2007. Nobody is waiting for a perfect system. It is waiting for a faster one.
What This Means If You Are Building
If your city has a preapproved plan program, use it. Permitting savings are real, design quality is professionally reviewed, and the time compression alone can save thousands of dollars in carrying costs that no amount of design customization will offset. Ask your planning department whether they are applying for ROAD Act pattern book grants, and if they are not, ask why not.
If you are a builder working in multiple jurisdictions, the current programs will not help you much. Thirteen plans designed for Rogers, Arkansas, do not transfer to Phoenix or Portland or Raleigh, and the manual effort required to produce jurisdiction-specific pattern books means the supply of preapproved designs will grow slowly, one city at a time, one architect at a time, unless the generation process itself changes.
If you are an architect, the threat and the opportunity are the same thing. Firms that learn to work with AI compliance tools, to design at the constraint boundary where code requirements meet lot geometry meet community character, will produce pattern books that no algorithm will replicate on its own, because the algorithm does not know what a neighborhood should feel like. It knows what fits. It does not know what belongs. That distinction is the architectural contribution that pattern books have always carried, from the eighteenth-century engravings to the thirteen plans in Rogers, and no federal grant program or AI compliance engine will change it.
Congress funded the demand side. Supply exists in scattered pieces. Someone will connect them. The question is whether the result will be architecture or merely geometry.