The bureau called Fulgham to see if he could find a cheaper way to get things moving again. “I secretly always wanted to work for the FBI or the CIA,” he told me. “So when they called me, particularly with this huge, hairy problem, it was like getting offered my dream job.”
First, though, Fulgham needed to convince the bureau that his approach was the right one. Fulgham’s management style, he explained, drew its inspiration from examples like NUMMI. In the previous two decades, as NUMMI’s success had become better known, executives in other industries had started adapting the Toyota Production System philosophy to other industries. In 2001, a group of computer programmers had gathered at a ski lodge in Utah to write a set of principles, called the “Manifesto for Agile Software Development,” that adapted Toyota’s methods and lean manufacturing to how software was created. The Agile methodology, as it came to be known, emphasized collaboration, frequent testing, rapid iteration, and pushing decision making to whoever was closest to a problem. It quickly revolutionized software development and now is the standard methodology among many tech firms.
Among filmmakers, the “Pixar method” was modeled specifically on Toyota’s management techniques and became famous for empowering low-level animators to make critical choices. When Pixar’s leadership was asked to take over Disney Animation in 2008, executives introduced themselves with what became known as “the Toyota Speech,” “in which I described the car company’s commitment to empowering its employees and letting people on the assembly line make decisions when they encountered problems,” Pixar cofounder Ed Catmull later wrote. “I stressed that no one at Disney needed to wait for permission to come up with solutions. What is the point of hiring smart people, we asked, if you don’t empower them to fix what’s broken?”
In hospitals, the distribution of authority to nurses and others who are not physicians is referred to as “lean healthcare.” It is a management philosophy and a “culture in which anyone can, and indeed must, ‘stop the line,’ or stop the care process if they feel something is not right,” the chairman of one lean hospital, the Virginia Mason Medical Center, wrote in 2005.
These approaches emerged in different industries, but they and other adaptations of lean manufacturing shared key attributes. Each was dedicated to devolving decision making to the person closest to a problem. They all encouraged collaboration by allowing teams to self-manage and self-organize. They emphatically insisted on a culture of commitment and trust.
Fulgham argued that the FBI’s technology efforts could succeed only if the bureau embraced a similar approach. FBI officials had to commit to distributing critical decision-making power to people on the ground, he said, such as lowly software engineers or junior field agents. This approach was a significant shift because previously, bureau executives—distrustful of one another and consumed by internal power struggles—had designed new technology systems by first outlining thousands of specifications each piece of software needed to satisfy. Committees filled hundreds of pages with rules for how databases ought to function. Any major change required approvals from numerous officials. The system was so dysfunctional that software development teams would sometimes spend months building a program, only to be told it was canceled when they were done. And the results were often dysfunctional as well. When Fulgham asked for a demonstration of the Sentinel work done thus far, for instance, an engineer led him to a computer monitor and invited him to input some keywords, such as a criminal’s alias and an address associated with a crime.
“In fifteen minutes, we’ll have a report of previous cases linked to that address and name,” the engineer said.
“The people I’m going to report to carry guns, and you want me to tell them it will take the computer fifteen minutes to provide help?” said Fulgham.
A 2010 inspector general’s report had said it would take another six years and $396 million to get Sentinel working. Fulgham told the bureau’s director that if they gave him the authority to distribute control, he would cut the number of people needed from more than four hundred to just thirty employees and deliver Sentinel for $20 million in a bit over a year. Soon Fulgham and a team of software engineers and FBI agents were holed up in the basement of the bureau’s Washington, D.C., headquarters. The only rules, Fulgham told them, were that everyone had to make suggestions, anyone could declare a time-out if they thought a project was moving in the wrong direction, and the person closest to a problem had primary responsibility for figuring out how to solve it.
The main problem with Sentinel, Fulgham believed, was that the bureau—like many big organizations—had tried to plan everything in advance. But creating great software requires flexibility. Problems pop up unexpectedly and breakthroughs are unpredictable. The truth was, no one knew exactly how FBI agents would use Sentinel once it was functional, or how it would need to change as crime-fighting techniques evolved. So instead of meticulously predesigning each interface and system—instead of trying to control from above—they needed to make Sentinel into a tool that could adapt to agents’ needs. And the only way to do that, Fulgham was convinced, was if developers were unfettered themselves.
Fulgham’s team started by coming up with more than one thousand scenarios in which Sentinel could be useful, everything from inputting victims’ statements to tracking evidence to interfacing with FBI databases that looked for patterns among clues. Then they started working backward to figure out what kind of software should accommodate each need. Every morning, the team conducted a “stand-up”—meetings where everyone stood to encourage brevity—and recounted the previous day’s work and what they hoped to accomplish over the next twenty-four hours. Whoever was closest to a particular problem or a piece of code was considered the expert on that topic, but any programmer or agent, no matter their rank, was free to make suggestions. In one case, a programmer and a field agent, after brainstorming, suggested that they model part of Sentinel on TurboTax, the popular financial software that reduced thousands of pages of complicated tax laws into a series of basic questions. “The idea was basically ‘Investigations and Justice for Dummies,’?” said Fulgham. “It was absolutely brilliant.”
Under the old system, getting approval for that proposal would have taken upward of six months and required dozens of memos, each carefully scrubbed of any mention of TurboTax or any indications that programmers intended to simplify federal procedures. No one would have wanted an enterprising lawyer or journalist getting their hands on something that used plain English to explain how the system worked. Under Fulgham, though, none of that bureaucracy existed. The programmer and agent mentioned the idea on a Monday, had a prototype ready by Wednesday, and everyone agreed to use the approach going forward on Friday. “It was like government on steroids,” Fulgham said.