By switching to agile development, the FBI says it can now finish its troubled Sentinel project for $20 million and within 12 months--despite having already spent nearly 90 percent of its $451 million budget, and despite also being two years behind schedule and having completed, at best, two of the four phases of Sentinel's development.

Oh, and it'll do it despite a planned reduction in contract employees from 220 to 40, and a reduction in bureau employees working on Sentinel from 30 to 12--in total, a 67 percent reduction of personnel from current levels.

Does the FBI know something about agile development that the rest of us don't? Agile development is often a good idea, especially when it becomes apparent that all the front loading of the waterfall method is terminally bogging a project down.

But agile development isn't magic. It can't create a case management system with all the planned functionality out of thin air, of which a pretty good definition might be the atmosphere around a project that's had 67 percent of its workforce cut.

The FBI's new plan to rescue Sentinel admittedly involves paring back of some planned functionality, apparently mainly a planned migration of data from the Automated Case Support system into Sentinel. But, in order to arrive at that $20 million figure, are there other functions the FBI has decided to no longer support?

Or maybe there's an alternative explanation. Might the FBI simply have stumbled on a $20 million commercial-off-the-shelf system that will work perfectly?

The FBI's completion estimates look especially wondrous when compared to other estimates for completing Sentinel, which involve figures like at least $351 million and six more years (that's what MITRE says, according to the Justice Department inspector general) or $194 million and about 28 more months (Lockheed Martin's figure).

I certainly hope the FBI shares the unknown ingredient behind its newest cost, schedule and personnel estimates. Other federal agencies might want to adopt whatever it is, too--that is, unless the FBI's estimates are really just a sign of surrender. In which case $20 million can buy a lot of white flags. - Dave