Every two weeks, the team demonstrated their work for a broad audience of high-ranking officials who provided feedback. The bureau’s director had forbidden anyone from micromanaging or making demands. At most, division heads could offer suggestions, each of which was cataloged and evaluated by whoever was closest to that piece of code. Gradually, the Sentinel team became bolder and more ambitious, not just building systems for record keeping, but linking Sentinel to tools that identified trends and threats and made comparisons among cases. By the time they were done, Sentinel was at the core of a system so powerful it could look across millions of investigations simultaneously and pick out patterns agents had missed. The software went live sixteen months after Fulgham took over. “Deployment of the Sentinel application in July 2012 represented a pivotal moment for the FBI,” the agency later wrote. Sentinel, in its first month alone, was used by more than thirty thousand agents. Since then, it has been credited with helping solve thousands of crimes.
At NUMMI, decentralizing decision making helped inspire a workforce. At the FBI, it played a different role. Lean management and agile methods helped fuel the ambitions and innovation of junior programmers who had been beaten down by bureaucracy. It emboldened them to come up with solutions no one had considered before. It convinced people to swing for the fences because they knew they wouldn’t be punished if they missed the ball now and then.
“The effect of Sentinel on the FBI has been dramatic,” wrote Jeff Sutherland, one of the authors of the Agile Manifesto, in a study of Sentinel’s development published in 2014. “The ability to communicate and share information has fundamentally changed what the bureau is capable of.”
More important, the way that Sentinel succeeded served as a source of inspiration to the agency and its leadership. “The Sentinel experience taught us a lot about how much potential can be unlocked when you give people more authority,” Jeff Johnson, the bureau’s current chief technology officer, told me. “We saw how much more passionate people can become. You look at some of our recent cases—the kidnapping in North Carolina, hostage rescue situations, terrorism investigations—and in situations like that, we’ve learned it’s critical that agents feel like they can make independent decisions.
“But empowering people in an agency this size is really hard,” Johnson said. “That was one of the problems before 9/11—people didn’t feel rewarded for independent thinking. Then you look at things like the development of Sentinel, and you see how much is possible.”
IV.
After the agents working on the Frank Janssen kidnapping case inputted the data they had collected into Sentinel, the software and the databases connected to it began looking for patterns and leads. The agents had entered the cellphone numbers the bureau had collected, the addresses investigators had visited, and the aliases the kidnappers had used in intercepted calls. Others inputted the names of people who had visited Kelvin Melton in prison, license plates caught on cameras around Janssen’s home, and credit card transactions from inside the stores where the burner phones were bought. Every detail was fed into Sentinel in the hopes a connection would emerge.
Eventually, the agency’s databases discovered a coincidence: The phone that had sent photos of Frank Janssen to his wife had also made a call to Austell, Georgia, a small city outside Atlanta. The FBI’s computers had looked through millions of records from other cases, and had found a link to Austell from another case.
In March 2013, a year earlier, a confidential informant had given the bureau the address of an apartment in Austell that he said criminals used as a safe house. That same informant, in a different conversation, had also mentioned an imprisoned gang leader who had “put a hit on the female District Attorney who prosecuted him,” a reference the FBI believed was to Kelvin Melton, the man who allegedly planned the Janssen kidnapping.
At the time of those conversations, no one inside the FBI knew what the informant was talking about. Janssen wouldn’t be kidnapped until a year later. And since then, no one had given the conversation a second thought. The agents who had interviewed the informant weren’t even part of the team looking for Janssen.
But now, the computers connected to Sentinel found a link: A confidential informant had mentioned someone who fit the description of Kelvin Melton, who had allegedly planned the kidnapping. That informant had also mentioned an apartment in Austell—an apartment that, Sentinel had just revealed, might have received a phone call from one of the kidnappers’ phones.
Someone needed to visit that apartment.
The problem was, this was just one of dozens of leads investigators were chasing. There were former associates of Melton’s to track down, prison visitors to scrutinize, former girlfriends who might be involved. There were too many potential leads, in fact, for agents to pursue them all. The FBI needed to prioritize, and it wasn’t clear that chasing a clue from a year-old conversation was the best use of time.
In recent years, however, as the success of Sentinel had attracted more notice within the bureau, officials had become increasingly committed to using lean and agile techniques throughout the agency. Commanders and field agents had embraced the philosophy that the person closest to a question should be empowered to answer it. FBI director Robert Mueller had launched a series of initiatives—the Strategy Management System, the Leadership Development Program, Strategic Execution Teams—that were designed to spark, as he told Congress in 2013, “a paradigm shift in the FBI’s cultural mindset.” One particular focus was encouraging junior agents to make independent decisions about which leads they should pursue, rather than waiting for orders from superiors. Any agent could chase a clue if they thought something was being overlooked. It was a law enforcement version of pulling the andon cord. “It’s a critical shift,” said Johnson, the FBI chief technology officer. “The people closest to the investigation have to be empowered to make choices about how they spend their time.” Sentinel wasn’t the only influence behind this change, but it accelerated the adoption of an agile philosophy inside the bureau. “The FBI’s basic mindset is agile now,” Fulgham told me. “Sentinel’s success solidified that.”
The investigators on the Janssen case had dozens of leads to choose from. But junior agents were encouraged to make decisions for themselves. So two young investigators decided to visit the apartment the confidential informant had mentioned over a year ago.
When they arrived at the apartment, they learned it was occupied by a woman named Tianna Brooks. She wasn’t home, but her two young children were there, unsupervised. The agents called child protective services, and once the kids had been collected by social workers, the agents began canvassing neighbors, asking where Brooks had gone. No one knew, but one person said Brooks had been visited by two men staying nearby. The agents found those men and questioned them. They said they didn’t know anything about Brooks or any kidnappings.
At 11:33 P.M., a call came in to one of the many phones the FBI had linked to the kidnappers and, as a result, were under surveillance.
“They got my kids!” a woman’s voice said.
The agents in Austell were told about the call, and began questioning their suspects more forcefully. The agents pointed out that the two suspects had recently visited Tianna Brooks. Now, the FBI had intercepted a telephone call of a panicked woman—possibly Brooks herself—saying the FBI had her children.
In other words, the two suspects had recently visited someone who may be linked to a kidnapping.
Was there anything else they wanted to say?
One of them mentioned an apartment in Atlanta.