How AI Spotting Apps Could Help Cities Fix Potholes Before You Hit Them
A San Francisco company called Samsara just launched a new system that uses artificial intelligence to find and map potholes and damaged roads across America. The company announced Chicago as its first major customer, with other cities signing on as well.
Here's how it works: Samsara already has sensors in thousands of delivery trucks and commercial vehicles across the country. When these trucks drive over potholes or rough road surfaces, their sensors feel the jolt. The AI learns to recognize what those jolts mean—is it a small pothole or a big one? Is the road deteriorating fast or slow? The system then alerts city workers about the worst problems first, so they can fix them before accidents happen or potholes get worse.
Why Cities Care About This
Potholes and broken roads cost American cities more than $750 million a year to repair. Right now, most cities rely on residents calling in complaints through systems like 311, or they send workers out to look for damage by hand. Both approaches are slow. By the time a worker gets to a pothole, it's often gotten worse and more expensive to fix.
Samsara's approach is different. The company doesn't need to install new equipment on city streets. It uses the delivery trucks that are already out there every day. Billions of data points from these vehicles create an invisible monitoring system that covers almost all major U.S. roads. A city gets this coverage instantly, without spending millions on its own sensors.
The system can pinpoint exactly where a pothole is and guess how serious it is. City maintenance teams can then plan repairs based on urgency rather than waiting for the next complaint call.
How This Actually Works in a City
Chicago maintains about 4,100 miles of roads. Right now, keeping track of all that damage and fixing it all is expensive and slow. With Samsara's system, the city gets alerts on a dashboard showing which roads need attention most. Workers can then decide what to fix first based on real data instead of guessing.
The system works better in areas where a lot of Samsara-equipped trucks are already driving. A city with heavy delivery and commercial traffic will get more coverage than a rural area.
The broader context here involves a shift in how cities think about maintenance. Twenty or thirty years ago, power companies moved from fixing broken equipment only after it failed to checking on it regularly before things broke. Samsara is applying the same idea to roads, but using trucks instead of special sensors.
Two Other Tools for Cities
Samsara is also launching two related systems. One helps garbage collection services plan better routes and schedules to save fuel and labor costs. Another helps public transit agencies understand passenger flow and adjust bus or train routes based on demand.
All three systems use the same basic approach: they take data from Samsara's existing vehicle sensors and turn it into software that cities can use. This allows the company to create new services quickly without building new hardware.
What Could Go Wrong
There are real challenges to work through. A fully loaded delivery truck will register different impacts than an empty one hitting the same pothole, so the AI has to learn the difference. Bad weather complicates things too. Snow, ice, and water can hide real road damage or create false alarms. The system needs to tell the difference between permanent damage and temporary wetness.
Many cities also use old computer systems for work orders that don't easily talk to new software. Getting Samsara's system to work smoothly with existing city processes could be harder than the technology itself.
What This Means
If Samsara's system actually saves cities money by catching problems early and preventing big repairs, other cities will probably adopt similar tools. The real question is whether it reduces emergency repairs and extends how long roads last. That's what will convince cities with tight budgets to pay for it.
For now, this approach sits between two extremes: it's cheaper than building purpose-built sensor networks, but smarter than waiting for complaints. Whether it delivers on that promise will determine how widely it spreads.
