With the rise of driverless vehicles, automakers are transforming into tech companies. And there is one key behind the AI and machine learning that fuels driverless technology: Big data. The data comes from sources such as Tesla’s real-world miles, Toyota’s simulations of autonomous driving, and Ford’s synthesis of internal data in developing autonomous technology.

But while much of this research is relatively new, there is one automaker that has been gathering data for decades: General Motors (GM). With OnStar, GM was a pioneer in connected vehicles. For an insider’s view of the history of OnStar, TechRepublic visited its command center in Detroit.

Launched in February 1996 at the Chicago Auto Show, OnStar is a highly innovative space. “We file for a patent about once every six days,” said Terry Inch, executive director at Global Connected Customer Experience (GCCX) Operations. “We’ve filed for more patents than any other part of GM. A lot of them are a combination of technology and business processes.”

OnStar vehicle diagnostics is an example of that, Inch said.

“What you see is the evolution of the technology and the exponential increase in number of interactions and transactions that we have with the customers,” said Inch, pointing to the history of OnStar. By the end of its first year, there were a little over 3,173 interactions in a year. Today, there are 5 million live conversations with customers in North America every month.

In 2015, mobile interactions surpassed live interactions, for the first time ever. And by the end of 2016, Inch said, mobile interactions will surpass live interactions by three or four times.

Up until about four or five years ago, Inch said, OnStar was “basically the air traffic controller between the vehicle that was in distress, or the person inside the vehicle that was in distress, and the 911 fire and police.”

Recently, OnStar became accredited at the same level of 911 advisors. “If there’s an injury or an emergency in your vehicle, they can go through and actually start medical support over the phone,” Inch said. In two out of the last three years, he said, OnStar won an award for emergency medical dispatch of the year.

So far in 2016, Inch said, the service has helped deliver three babies in vehicles. And sometimes, advisors have to get creative–when an OnStar user suffered an asthma attack, the advisor asked the driver to use their horn to respond, honking once for yes, twice for no.

Customers can use OnStar even if the issue is unrelated to a problem with the car, or with the driver–like carjacking scenarios. “People will actually call from their house because somebody’s breaking into their car,” said Inch. They can also press the button inside the car if they find someone inside the vehicle.

SEE: Job description: Big data modeler (Tech Pro Research)

OnStar also answers about 100,000 “Good Samaritan” calls each month, in which users report drunk drivers or other activity on the road, like seeing a biker hit by a car.

While companies like BMW, Mercedes, and Lexus have services similar to OnStar, “nobody’s democratized the technology and the connectivity like we have,” Inch said. These other automakers, he said, have it built-in and included in the sticker price at a relatively high markup.

The inside of OnStar’s Command Center looks like something from NASA. There are huge interactive maps with blinking lights, updated every two minutes; excel documents full of figures; and employees at desks monitoring all the activity. They’re also monitoring weather activity nationwide, to see if there are any states of emergency.

“If, all of a sudden, you had severe flooding or a tornado come through, we can actually tie that back to counties that have declared an emergency and then put it on the advisor desktop,” Inch said. “So that when you press in from your county, the first thing we do is deliver the service you request, then we let you know that you’re in a county that has an emergency warning for weather right now. Are you in a safe spot? Can we do anything else for you? The advisor knows if you’re pressing in from a county that has an emergency.”

But beyond time of day, year, and weather, there is only one factor that truly predicts how many calls come in: The number of miles driven.

“The number of vehicles on the road at any given time, and the miles being driven by people, is a direct correlation to the calls. It’s just an incredible amount of information and business intelligence that we can translate from that itself,” Inch said.

The blinking lights differentiate between calls that come in from people who need directions and whether something like an airbag went off, or an automatic crash response.

Today, Inch said, they’ll do 180,000 live interactions.

“We’re constantly monitoring all the different lines of business to make sure the number of calls that we forecast are coming in during the time period that advisors are logged into the system and taking the calls that they’re supposed to be taking,” said Inch.

“We’re also managing the servers and data centers,” he said, and “somebody’s monitoring Verizon and AT&T and all their switches and what’s happening too.”

Over 20 years, Inch said, there have been a billion live transactions. What does that mean? Tight-knit relationships with cities, counties, and states. “You know how to get ahold of all of them,” he said. “We have a database of people and have relationships with all these people.”

These relationships are essential to the success of OnStar. It has priority lines at about 75% of the 7,000 911 centers nationwide, Inch said. That means if a call comes in from an OnStar driver, “they take it right away because they know that it’s a real emergency.”

“In the early days,” Inch said, “probably seven out of 10 of the calls were false-positives; people that just hit the wrong button by mistake.”

When OnStar launched in Europe, “the EU has something called eCall mandate,” said Inch, “which means if you’ve been in a crash, you have to contact a fire or police. They wanted it just to be machine to machine. We were able to come back with 20 years of data and say, ‘Let us explain why you don’t want to do it that way,’ because most of these things will be false-positives.”

You need a person in the mix, he said, otherwise “you’re going to get 10 calls in an already cash-strapped city or region or state, and you’re going to be overwhelmed. You don’t know what you’re asking for.”

“We actually were able then to go back to the head of the EU and say, ‘Okay, we will allow the call come into you guys, have a live person triage it, and determine whether it needs to get passed on or not.”

Everything GM has learned from OnStar, Inch said, now applies to the new world of connected vehicles. “What we’re doing with our ride-share Maven, with Lyft, with autonomous driving,” he said. “Those are all things we’ll have to apply these learnings to, and apply our relationships with those local fire and 911 communities.”

Also see