One day, robots and artificial intelligence could eliminate millions of jobs in America, even in industries that aren’t immediately obvious. A study not long ago from two Oxford University academics found that 47% of positions are at risk of computerization over the next two decades.
Robots are likely to take jobs away from humans because they’re cheaper than humans and more efficient, but also because they’ll obviate the need for services that prop up and protect humans. Why have human resources professionals if you don’t employ human resources? Why buy insurance if automation makes accidents far less likely?
That’s why so many people think the insurance industry could suffer in the age of advanced AI–particularly the auto insurance industry.
“The truth is, if it’s a safer way of driving, it’s good for society and it’s bad for our insurance business,” Warren Buffett said recently when asked about the effect of automated vehicles (AVs) on his Geico subsidiary. “Anything that cuts accidents by 30%, 40%, 50% would be wonderful, but we would not be holding a party at our insurance company.”
In fact, some estimates are higher than that. Google forecasts that its driverless pods could eventually prevent 90% of all accidents, while at the same time taking many cars off the road (because we’ll have more sharing of cars, rather than everyone having their own vehicle). Others have speculated that premiums could be reduced 75%, especially if drivers are no longer required to get specific coverage, and liability shifts from drivers to manufacturers and technology companies.
How quickly might that happen? In a recent report, McKinsey predicted the change could start between 2023 and 2037. “Insurers might be required to shift their main goal from covering private customers from risk tied to ‘human error’ to covering [manufacturers] and mobility providers against ‘technical failure,'” it said.
The U.S. National Highway Traffic Safety Administration (NHTSA) carves auto-automation into stages, starting with “function-specific automation” (level one), which includes cruise control and automatic braking, which have been around for a while. Next comes level two, which includes adaptive cruise control, forward collision warnings, lane centering and drowsy driver detection, which are all starting to appear now. After that, we have level three, is where drivers sometimes completely cede control of vehicles in optimal conditions, and level four, where the car takes over completely at all times.
A PwC analysis predicts these level two technologies will see a 15% reduction in claims for bodily injury, a 6% decrease in collision claims and a 14% drop in property protection claims over 20 years, but it calls its projections “conservative.”
In the short term, the level of accidents may decrease, but they might become more serious and more costly, says Jeff Blecher, senior VP of strategy at Agero, which sells vehicle information systems to car-makers. He expects level 2 automation to reduce the incidence of front-on collisions, side-swipes and routine fender-benders. (The accident rate is already falling in late-model vehicles). But advanced controls could also lead to poorer driving, such that when we do have to react, we’re less good at it.
“Think about the typical driver and the amount of training they get and what happens as their skills atrophy because 95% of the time they’re actually not driving with these systems, but they only have to drive when it’s hard,” Blecher says. “The frequency might go down, but the accidents might be worse.” And because the cars contain more equipment at the points where crashes are likely, they might cost more to repair than they do today, making the overall impact to insurance claims something of a wash.
The more intriguing question is what happens when fully automated cars become adopted–something Blecher says could start happening in about 10 years. A Rand Corporation report last year speculated that drivers might cover themselves with health insurance and homeowner’s liability insurance, and not see the point of specific car insurance. After all, we don’t buy dedicated insurance for most of what we do in life, say to ride a bicycle. And more to the point, it’s likely that car thefts will become significantly less problematic than they are today. In the future, all vehicles will probably be fitted with tracking and disabling technology, so that hot-wiring your neighbor’s Mercedes will be a lot less fun.
The big question, says Blecher, is the extent to which manufacturers will actually want to have full automation. If they’re going to be held liable for crashes in the future, it’s likely they’re going to want either very advanced technology first or some kind of additional protection under the law. In the mid-’70s, when airbags started to appear, some manufacturers resisted installing them because they feared being held liable if they didn’t work. Manufacturers could do the same now with full vehicle automation. The biggest obstacles to fill AVs may not be technology or infrastructure, but rather liability concerns. It’s going to take some time to regulate how that’s going to work, especially as courts and officials have little prior evidence to work with.
Looking ahead to the middle of the century, many experts expect not only the notion of insurance to change, but also the notion of car ownership to change. Google and Uber see us sharing cars, hailing them when we need them. And, interestingly, that might be a good thing for manufacturers too, as it may help them evade liability more easily.
“By doing so, manufacturers may be able to use contract law to limit their liability and [have] better control the way their products are used,” the Rand report says.
There are so many moving parts to the future of vehicles that it’s impossible to predict. But it seems likely that greater automation and greater sharing will harm the classic insurance model and put quite a few people out of work. “The insurance companies are still going to be around, but their business models are certainly going to have to adapt,” Blecher says. That sounds like an understatement.