Chief Changes

Vivek Kundra led bold tech reforms, but his legacy is more about management than machines.

When President Obama named then-District of Columbia Chief Technology Officer Vivek Kundra to be the nation's first chief information officer in 2009, he cited Kundra's "depth of experience in the technology arena" and said the new CIO would "work to ensure that we are using the spirit of American innovation and the power of technology to improve performance."

When Kundra announced he would leave his post in August to take a fellowship at Harvard University, government and industry officials praised his "bold vision," "forward thinking" and "revolutionary" reforms that resulted in billions of dollars in information technology savings.

What happened during the intervening 20 months at agency IT offices, though, was less a technological revolution than a shift in paradigms, a new set of organizational and management practices and an increase in accountability aimed at keeping projects on track, on time and on budget.

"I think it's 100 percent organizational," Veteran Affairs Department CIO Roger Baker says of the changes he's instituted during two years at the helm.

"The technology is pretty straightforward," Baker says. "Especially for an organization like VA, we don't try to be bleeding edge or frankly even leading edge. We just try to use modern technology. The major thing for government is that what we do tends to be huge in scale compared to the private sector, so all the things that can cause a program to go off the rails are magnified. Inadequate requirements won't cause a little problem here, they'll cause a huge problem."

The main question State Department CIO Susan Swart asks when looking at IT projects isn't what technology they're using, she says, but whether that technology is the right fit for the problem it's supposed to address.

"If you're looking at what we spend the preponderance of our IT budget on and improving how those things meet the needs they were intended to," Swart says, "it's not so much looking at the top layer of technology. It's looking at the process and how we're implementing that technology."

The extent to which the changes Kundra, Swart, Baker and other agency CIOs have instituted in the past two years have focused on management rather than technology is underscored by a slate of best practices case studies the federal Chief Information Officers Council began publishing on its website in April. Eight of those 10 best practices primers highlight changes in deadline structures, improved interagency communication or some other management issue rather than a purely technical fix to an IT problem.

The Social Security Administration case study, for example, details a drive to raise the certification standards for the agency's IT project managers, and the Office of Personnel Management study describes imposing a monthly self-assessment review on all IT projects.

Even Kundra's 25-Point Implementation Plan to Reform Federal Information Technology Management, published in December 2010-the master document of his tenure as federal IT chief-is much more about changing the way government interacts with IT than about the technology itself.

The plan begins on the technical side: shutting down a third of the government's 2,100 data centers by 2015 and making large, versatile computer clouds the de facto first choice for housing new government IT initiatives.

More than 20 of the 25 initiatives, though, are nontechnical: formal career paths and training for IT program managers and acquisition professionals; a program team approach to IT projects, which creates stronger links between IT builders and their agency constituents; and, perhaps most ambitious, working with Congress to open up the IT budgeting process so the government can more quickly and easily benefit from innovation.

In the days before the 25-point plan and other management reforms, major government IT projects tended to chug along for years without a full review. Programs were restructured as they went, adding pieces to fill new needs or to incorporate the latest technology, agency CIOs say.

Many of those projects came off well in the end, they say, but others lost direction. Sometimes a massive and multilayered project would get held up for months because one small but vital component was delayed. Other times, the vision of the IT staff building the new project would fall out of sync with what agency leaders wanted.

"Technology can do so many things that scope creep is easy," Swart says. "I think sometimes there's this sense that if technology can do it then technology should do it. It's easy to stray away from the KISS principle: Keep it simple, stupid. People say, 'well, a tweak here and we can add this report and do that' and it makes it easy for IT projects to stray away from their initial intent."

The remedies that Kundra and agency chief information officers came up with focused on heightened management and oversight, generally scaling back the scope of technology projects rather than ramping them up, and building stronger communication with vendors so more IT work could be done by the private sector.

The most far-reaching management fix was Kundra's TechStat review process, a series of adapt-or-perish presentations by managers of over-budget and delayed projects to the CIO and agency leaders. In the first two years, those reviews led officials to cancel four major projects outright and drastically scale back 11 others, saving the government about $3 billion, according to Kundra.

The 25-point plan requires agencies to set up internal TechStat sessions and as of March, most reported having held at least one such session. Just as important, agency CIOs began ensuring projects that weren't canceled stayed on track, by what Kundra calls "chunking" them down into periods of six months, or even less, and demanding some level of functionality with each deadline.

"This way, if you have to stop investing in it or if it doesn't prove to be feasible for whatever reason, you [still] have something delivered that works well," Swart says. "In every phase you have to get some functionality out of it, not just 'in three years we'll have the whole kit and caboodle.' "

NEXT STORY: Unplugging Data Centers