The FBI spent more than $1 billion over a decade on failed attempts to build a modern digital case-management system, ultimately succeeding only after bringing the project in-house with a fraction of the original budget, according to a case study published by the Niskanen Center. The analysis, written by public interest technologist Mark Lerner, chronicles how the bureau's missteps on information technology contributed to security failures before the September 11 attacks and continued wasting taxpayer money for years afterward. The report argues that rigid project management practices, lack of in-house technical talent, and misaligned oversight allowed contractors to burn through hundreds of millions of dollars while delivering nothing usable.
The bureau's troubles began in 1995 when it rolled out the Automated Case Support system, which was already outdated on launch day. The report describes how this "greenscreen" application required dozens of complex keyboard commands for simple tasks, forcing agents to avoid using it or delegate work to junior staff who would then print paper records as the official source of truth. By 2001, the FBI launched a three-part program called Trilogy with a $380 million budget, later expanded to $458 million—about $850 million in today's dollars. The third part of Trilogy, called Virtual Case File, was contracted to Science Applications International Corporation but was cancelled in 2005 after four years and over $100 million spent, with software so poor the bureau scrapped it entirely. A replacement effort called Sentinel began in 2005 with a budget of $425 million, eventually growing to $451 million. When contractor Lockheed Martin delivered an unusable version missing basic features like autosave and deleting hours of users' work, an independent audit estimated that continuing with existing practices would cost an additional $351 million and wouldn't finish until 2016. By the time the FBI pulled the project from Lockheed Martin in September 2010, the agency had spent $405 million on Sentinel alone—bringing total losses from both failed programs to roughly $884 million when adjusted for inflation.
The technical shortcomings had catastrophic consequences beyond wasted money. The 9/11 Commission found that the FBI possessed data on people connected to the attacks entering the country and engaging with flight schools, but the information was isolated across different offices, unavailable to those who could have used it. Immediately after the attacks, FBI agents in Florida had to send suspected terrorist records to Washington by overnight shipping because the bureau lacked software capabilities for transmitting data electronically. At a Senate oversight hearing, then-Senator Charles Schumer noted that an FBI official "had to personally deliver [testimony] on a disk," and described the bureau's database warehouse as "like Medusa, with over 40 databases with separate functions operating out of the same body but totally disconnected from one another." A whistleblower at SAIC tried raising concerns about quality and cybersecurity practices, claiming the company's attitude was that "it's other people's money, so they'll burn it every which-way they want to," but the FBI removed him from the project.
Success came only when the FBI abandoned its reliance on major contractors and restrictive management processes. An unofficial group of bureau staff with technical skills built their own prototype nicknamed "Lavender" using modern development practices—small teams, direct user engagement, and iterative building. After seeing the demonstration, the FBI brought Sentinel in-house in September 2010, reducing the team from hundreds to around 40 people and adopting agile methods that prioritized working software over documentation and rigid plans. The report identifies several root causes of the earlier failures: waterfall-style requirements documents running hundreds of pages that were finalized before any code was written, change-control processes requiring paperwork and meetings for minor updates, and oversight bodies measuring success through documentation rather than actual value to users. Both VCF and initial Sentinel programs generated over 20,000 pages of documentation for Sentinel's first phase alone, creating massive administrative burdens without preventing failure. The turning point came from hiring leaders with private-sector technical experience and empowering in-house federal employees with skills beyond just writing code—including product management and human-centered design.
The in-house team delivered the complete Sentinel system on July 1, 2012, for just $20 million—roughly 2 percent of what had already been spent—after confidently rejecting Lockheed Martin's proposal to finish the work in three years for $194 million. Users gave the final system a satisfaction rating of 8.5 out of 10 and immediately adopted it bureau-wide, a stark contrast to earlier versions that agents actively avoided. The report argues that modernizing government infrastructure requires shifting how agencies fund, build, and buy software—treating systems as enduring products with steady teams rather than time-bound projects driven by arbitrary deadlines. It notes that oversight bodies' focus on procedural compliance and adherence to predefined plans created perverse incentives, measuring proxies like earned value management reports instead of whether users' experiences actually improved. After more than a decade of failures under heavy congressional scrutiny, the FBI's ultimate success came from capability-based budgeting, in-house technical talent, and the flexibility to shift priorities based on real feedback—a model the report suggests has broad application across public and private sectors for organizations trying to overcome operational barriers and deliver promised products.