In auto insurance, claims often move at the pace of the slowest step – and there are plenty of steps. From the moment a policyholder reports a loss to the final settlement, the process is riddled with repetitive tasks and partial information, driving up costs for both insurers and customers.
Mike Nannizzi (pictured left), a board member at UVeye, a US-based developer of automated vehicle scanning systems, said one of the root issues is the insurance industry’s reliance on uniform processes that leave little room for innovation.
“Insurance companies tend to like to have one process for everything,” Nannizzi said. “That makes it hard to try new things, because by definition, new things are different than old ones.”
When a claim is filed, the first inspection often happens at a designated location that can only conduct a visual review. In most cases, Nannizzi said, these inspections are intentionally non-invasive to keep the vehicle drivable afterward. The assessor will take photos and prepare a report – but rarely dismantle anything to see the full scope of damage.
Because of this, the initial estimate is “almost always wrong,” Nannizzi said. The true extent of damage is only revealed when the vehicle goes to a repair facility for a second inspection, where parts are removed and a more accurate assessment is made.
When new damage is discovered during repairs, the original estimate is adjusted through a process known as a supplement. While supplements are standard in the industry, they add delays to repair timelines and create additional administrative work for insurers, shops, and claimants.
“This first part is just to get a context of how bad the incident is,” Nannizzi said. “But it starts the whole thing over again.”
Nannizzi noted that inefficiencies in claims handling are often compounded by the absence of consistent, pre-loss vehicle condition records. Unlike programs that track driving behavior, insurers rarely require regular physical inspections before a claim arises.
“If I’m insuring a car, I want to know the condition of the car when I start insuring it, because I don’t want to be responsible for prior damage,” he said. Without that baseline, even minor accidents can trigger disputes about whether damage was pre-existing or caused by the incident in question.
These disputes not only slow down claim resolution but can also lead to overpayments when unrelated damage is mistakenly included in repair estimates.
Nannizzi pointed out that a vehicle’s overall condition matters just as much as the immediate damage in determining claim outcomes. For example, a small dent might be enough to total an older truck with a rusted-out bed, but it would barely register for a newer, well-maintained vehicle. Without a complete picture at the outset, adjusters may misjudge whether repairs are economically viable.
Nannizzi emphasized that claims inefficiencies aren’t limited to personal auto insurance. The journey a vehicle takes from the assembly line to the end consumer involves multiple transfers – from manufacturer to transport truck, onto a ship, back onto land transport, and finally to a dealership.
At each stage, different parties assume responsibility for the vehicle, and each typically carries its own insurance coverage. In overland transit, this often falls under inland marine insurance, a category that covers goods in motion outside of traditional ocean shipping.
When damage is discovered during these transitions, determining liability can be challenging. “If a vehicle is moving from truck to truck and there’s damage, whose fault is it? Nobody knows,” Nannizzi said. In the absence of clear accountability, insurers for each party may be drawn into disputes, prolonging settlement timelines and increasing costs.
For many claims, the most definitive assessment only happens once a repair shop dismantles the vehicle. Until then, adjusters are working with partial information, which can slow decisions and create a cycle of revisions.
He noted that much of the process remains “overly manual,” involving multiple handoffs and a reliance on in-person rechecks when initial estimates fall short. This, he warns, can stretch the timeline.
“That estimate process could take five days,” Nannizzi said.
Photography presents another bottleneck. Adjusters often work with a limited number of images they can attach to a claim file. If a reviewer at the insurance company later needs a different angle, the adjuster may have to track down the policyholder and arrange a return visit – a delay that adds frustration for all parties.
Because many adjusters are independent contractors paid per assignment, they may photograph multiple vehicles in a single day and process estimates in batches later. Working from memory and a small set of images increases the risk of missed details, forcing further follow-up and prolonging settlement, Nannizzi said.
These constraints not only slow the claims process but can undermine accuracy and transparency, leaving both insurers and policyholders waiting longer for resolution.
While inefficiencies in claims handling are well documented, some technology providers are attempting to tackle these issues. One example is UVeye, whose AI-driven vehicle scanning systems are designed to capture exterior, interior and underbody details that are often missed during traditional inspections.
Yaron Saghiv (pictured right), the company’s chief marketing officer, explained that the process is meant to replicate a thorough technician walk-around – but in seconds, and with uniform lighting, angles, and resolution. The system can detect dents, scratches, glass damage, corrosion, leaks, tire wear, and other visible issues, marking them automatically in a digital record.
Saghiv stressed that UVeye’s focus is on what can be seen and measured from the outside: “As long as it’s visual, it’s exterior, and to a very fine detail… that’s what we indicate.” The company, he said, is adding optional sensors for thermal readings, engine sound recording, and other diagnostics.
He added that UVeye’s use of AI dates back to its early work in vehicle security inspections, when it applied computer vision, machine learning, and deep learning models to identify anomalies in undercarriages. Today, those same capabilities drive its damage detection, while generative AI is being tested for automotive merchandising applications.
However, this technology is not yet available in Canada, and any potential benefits to Canadian insurers remain hypothetical for now. Nannizzi noted that industry-wide adoption depends heavily on scale and market penetration. “Insurance companies like this – but they’ll say, ‘Call me when they’re everywhere,’” he said.
