In a demonstration last week, the bus operator drove the bus as usual until he reached his intended point on the Forth Road bridge in Edinburgh. A soft ding alerts passengers that the driver, Steven Matthew, has engaged autopilot.

That’s when he carefully kept his hands hovering over the steering wheel—still ready to wrest control of the bus from the computerized driver in the event of an accident.

From a few feet away, Matthew still looks like he’s driving a bus. Only on closer inspection can it become clear that his arms are not moving with the steering wheel.

“I think the technology is great,” 47-year-old operations executive Matthew told NBC News. “It stays in the lane and brakes when it senses other traffic. The only thing you might have to worry about is other drivers not knowing what they’re going to do.”

Even with Matthew’s complete trust in the technology, his persistent presence behind the wheel is reassuring as the bus races across the massive bridge at 50 mph. Experts say the bus may need a Matthew, not to protect humans from technological flaws, but to support technology from human driver and pedestrian flaws.

“The biggest hurdle for self-driving cars is dealing with people, especially in urban environments, where people make their own decisions,” said Ram Murthy, 48, a professor at the University of Edinburgh’s School of Informatics. Directly involved in transit projects.

Human drivers are always “relaxing the rules a little bit so they can get through and work with each other,” he said. Murthy added that if there were only self-driving cars on the roads, the technology would work almost flawlessly and the number of motor vehicle accidents and fatalities would plummet.

But even when human frailties lead to technical errors, there are reasons for public skepticism.

Last year, the National Highway Traffic Safety Administration reported that “nearly 400 accidents involving vehicles equipped with partially automated driver assistance systems occurred in a 10-month period, including 273 tesla,” according to the Associated Press.

NHTSA cautioned that the data should not be used to compare the safety of different automakers because the data is not “weighted by the number of vehicles from each manufacturer that use the system or how many miles those vehicles drive.”

Earlier this year, regulators urged Tesla to recall more than 363,000 vehicles equipped with its “Full Self-Driving system” because the system doesn’t always obey traffic safety rules and can lead to crashes. Tesla disputed the regulator’s decision, even though the company agreed to the recall.

Stagecoach officials and researchers at Fusion Processing, the company that innovated the CAVStar self-driving system, acknowledged that even if the bus routes are real, the so-called CAVForth bus project is still just an experiment. It is part of a project part-funded by the UK government’s Center for Connected and Autonomous Vehicles, which also involves the local transport department and two universities.

Organizers hope that the technology will eventually reduce human error, thereby reducing road accidents and fatalities. By reducing the need for human drivers, organizers hope to reduce costs and thus make transit systems more accessible to smaller towns and cities that currently cannot afford to provide them.

Each bus is packed with about 20 sensors, cameras and radar, as well as a sophisticated satellite-linked global positioning system.

Every 30 minutes, a bus will traverse the 14-mile route, which Fusion says includes “a series of complex traffic maneuvers such as roundabouts, traffic lights and ‘weaving’ highway changes.”

Jim Hutchinson, Fusion’s chief executive, said prudent passengers should remember the advantage computers have over human drivers: Autopilot doesn’t have to check blind spots and be free from distractions. The sensor never blinks.

Fusion and Stagecoach officials said that while the service is still experimental, the buses are only hitting the road after extensive testing: The buses have been developed over a decade and have more than 1.1 million miles of testing.

If the trial is successful, organizers hope to roll out similar technology to four other UK cities by the end of the year. The various companies and government agencies behind the project hope that wider adoption will spur regulatory and legal changes that could eventually lead to truly “driverless” bus routes.

“We still understand that we need to make sure that, you know, the public supports us in this matter,” Hutchinson said. “So I think there’s still work to be done on that. But you know, the technology is ready now.”