The bottleneck doesn't happen at the construction site. It happens months earlier, in fluorescent-lit municipal offices where understaffed planning departments wade through hundreds of pages of architectural drawings, checking setbacks and parking ratios against zoning codes that predate the iPhone.
AutoSitu, a Detroit-based startup fresh out of Y Combinator's Winter 2026 batch, thinks it has a fix: coordinated AI agents that can review development plans for compliance in a fraction of the time human reviewers need. The company's pitch is blunt—turn months-long approval cycles into weeks, maybe days.
Whether that promise holds up in practice remains to be seen. But the problem AutoSitu is chasing? That's indisputably real.
The Scale of the Problem
Across the country, cities are drowning in permit backlogs while developers hemorrhage cash waiting for approvals. San José sends back more than 90 percent of accessory dwelling unit applications because of missing information or compliance snags. Austin's planning department routinely quotes 12-week review times for straightforward commercial projects. In Los Angeles, a simple ADU permit can take eight months.
The costs compound quickly. Every month a project sits in regulatory limbo costs developers upward of $250,000 in carrying costs—debt service, property taxes, lost lease income. Architects and engineers bill for revision after revision as plan reviewers send back comments, often vague or contradictory, that trigger wholesale redesigns.
AutoSitu positions itself as an "AI-native workspace for municipal development review." Strip away the Silicon Valley phrasing, and what you get is a multi-agent system designed to automate the tedious work of checking whether a proposed building complies with zoning codes, fire regulations, engineering standards, and the accumulated precedent of past decisions.
The product parses architectural drawings, site plans, and project schedules into structured data. Then it runs those inputs against what the company calls a "living rules graph"—essentially a digital representation of every applicable code, updated as regulations change. When the system spots an issue—setback violations, insufficient parking, improper landscaping buffers—it flags it with citations to the specific code section. No more "see previous comments" scrawled in red pen.
Where the AI hits judgment calls it can't resolve, it escalates to human reviewers through what AutoSitu describes as "smart handoffs." The goal, at least in theory, is to let machines handle the rote compliance checking while preserving human discretion for the genuinely complex stuff.
According to the company's launch materials, the system has processed more than 1,500 plans and can complete reviews 90 percent faster than traditional workflows. It claims to reduce city review comments by 50 to 70 percent and cut resubmittal cycles from the typical two or three rounds down to one or zero.
Those are impressive numbers if they hold. The question is how many cities are actually using this in production versus pilot mode.
Two Founders, One Big Bet
Xuanshu Lin and George Zhai launched AutoSitu with backgrounds that bridge the technical and regulatory divide. Lin studied urban planning and architecture at the University of Michigan and Harvard's Graduate School of Design—credentials that include an APA Planning & Design Competition Honorable Mention. Zhai comes from the robotics and autonomy world, having studied at Georgia Tech.
Their thesis: municipal development review is fundamentally a multi-departmental coordination problem, and coordination problems are where AI agents shine. Planning checks zoning compliance. Fire reviews life safety. Engineering scrutinizes stormwater and traffic impact. Each department applies its own rule set to the same plans, creating redundant work and inconsistent outcomes—sometimes even contradictory requirements that leave applicants trapped in bureaucratic purgatory.
AutoSitu's agents are designed to orchestrate that process, triaging issues and checking compliance across departments simultaneously rather than sequentially.
The company frames its mission in the kind of ambitious language that plays well in the Y Combinator universe: "Increasing the GDP of the world" by accelerating housing and infrastructure approvals. Gustaf Alströmer, a YC partner known for championing infrastructure startups, is backing them.
It's a big swing. Maybe too big, depending on how fast they can execute.
Three Customers, Three Pitches
AutoSitu is selling to three distinct audiences, each with different motivations and pain points.
For cities, the value proposition centers on capacity. Planning departments could theoretically handle 10 times the volume without adding headcount, the company argues, while ensuring consistent, defensible reviews. There's also the political angle: elected officials facing constituent anger over housing costs can point to AI-accelerated approvals as tangible action.
Developers get a different pitch. AutoSitu highlights the carrying costs of permit delays—$250,000 or more per month on stalled projects adds up fast. The company claims its pre-submittal checks can help developers sidestep the 60 percent first-submission rejection rate it says is typical, potentially saving $100,000 to $400,000 in avoided holding costs per project.
For architects and engineering firms, the focus is on rework. Resubmission cycles cost $25,000 or more per round, AutoSitu says, and margins get eaten alive when projects ping-pong through multiple revisions. The system's marketing claims it catches 95 percent of compliance issues before submission, preserving the profitability that revision work destroys.
The company's website displays a marquee of "peer cities"—San José, Seattle, Long Beach, Boston, Coral Gables, Los Angeles, Dallas, San Mateo, Clark County, Honolulu, San Francisco. There are partner logos too: NVIDIA, architectural firm SWA. But these appear to be marketing displays rather than confirmed contracts. A search through public procurement records didn't turn up live municipal agreements, though that doesn't mean pilots or soft trials aren't underway.
A Suddenly Crowded Space
AutoSitu isn't alone in spotting the opportunity. The permit automation space has gotten noisy in the past 18 months.
San José has been piloting an AI pre-check tool called CivCheck to screen building permits before they hit human reviewers. Honolulu launched a similar CivCheck pilot in December. Austin is testing Archistar's eCheck system for zoning review. Small cities in Florida and New Hampshire have announced deployments of AutoReview.AI for site and landscape plan checks.
The incumbent civic technology platforms are moving too. Accela, which powers permitting systems in hundreds of U.S. cities, acquired ePermitHub and put AI-automated plan review on its product roadmap. OpenGov is overhauling San Francisco's permitting system as part of the PermitSF modernization—though AI features haven't been publicly detailed yet.
Then there's PermitFlow, another YC alum that raised a $54 million Series B in December and positions itself as "construction's AI workforce." It's building agents to handle permit research, submission, and inspection coordination—competing for the same pre-submission workflow AutoSitu is targeting with developers.
The question isn't whether AI will handle municipal plan review. Multiple cities are already piloting tools. The question is which approach wins: pre-check assistants that guide applicants through requirements, automated review systems that replace staff workflows, or full-stack platforms that rebuild the entire permitting infrastructure from scratch.
AutoSitu seems to be betting on the middle path—augmenting rather than replacing reviewers—but that strategy assumes cities want augmentation. Some might prefer outsourcing the whole headache.
Execution Challenges Ahead
AutoSitu operates as MirageDoodle, Inc., a corporate name that hints at a possible pivot from an earlier product idea. The company's privacy policy, last updated in January 2025, references basic security measures but doesn't claim SOC 2, CJIS, or FedRAMP compliance—certifications that larger municipal buyers typically require before signing contracts.
With a team listed between two and 10 employees on LinkedIn and headquarters in Detroit, AutoSitu is in classic early-stage mode. The company hasn't disclosed pricing. Confirmed government contracts remain elusive. LinkedIn posts mention pilots with planning departments and design firms in California and Michigan, but specifics aren't public.
That's not unusual for a startup barely out of YC. But municipal sales cycles are long, procurement processes Byzantine, and the switching costs for cities high. Building trust with risk-averse government buyers takes time—often more time than venture-backed startups have before investors start asking uncomfortable questions about traction.
The Window Might Be Open
Still, the broader market signal is unmistakable. Cities face immense political pressure to speed housing approvals, and AI plan review is one of the few automation plays that elected officials can champion without looking anti-labor. Most planning departments are so understaffed that anything promising to clear the backlog gets a hearing.
For a startup with a focused product and YC backing, the window might be open—assuming AutoSitu can navigate procurement cycles, build the trust municipal buyers demand, and move faster than better-funded competitors already in the field.
The permit backlog isn't going anywhere. Too much housing needs to get built, and every level of government is under pressure to clear obstacles. Whether AutoSitu can carve out a sustainable position in the race to fix it comes down to execution over the next 12 to 18 months.
Call it the unglamorous side of the infrastructure boom: not highways and bridges, but the permits that precede them. If AutoSitu gets it right, they might prove that the real bottleneck was always the paperwork.
