An Information Week article says:

The manner in which the federal government funds some state IT projects is at odds with the way states are implementing common IT architectures back home, state CIOs complained to a congressional panel on Tuesday. Specifically, the IT systems bolster state-run social-service programs, including food stamps, child welfare, child-support enforcement, and Medicaid.

We continually fight this problem.  The ADP process has good intentions: ensure money is spent on what it was appropriated for.  On the other hand, as Aldona says:

"It's clear that the [ADP] process strongly discourages using federal program funds to create common IT infrastructure," Kentucky CIO Aldona Valicenti said in testimony prepared for the House Subcommittee on Technology and Procurement Policy. "Although it frequently costs less to the original program than creating a separate, standalone IT system, it's precisely because it would benefit other programs that it's often termed unallowable."

We fight this by being willing to engage the Federal program officers early and gain permission for what we're doing.  Governor Leavitt sees it as a "federalism" issue and we have a fairly good track record of engaging the Feds on it.  Still it takes time and effort, so you only do it when it really matters.

Another way we combat it is through architecture.  A good example is our eREP project.  We administer billions of dollars of federal benefits every year.  The IT system that manages this is being replaced (old crufty mainframe code).  The system is essentially a CRM system with lots of specialized rules to determine eligibility and calculate specific benefits.  Every federal welfare program is funded separately, so theoretically, we should build 6 or 7 eligibility systems.  Instead, a core CRM system and rules engine will serve them all, with individual modules built for the individual programs.  To satisfy the Feds, we have to be able to track and correctly allocate the costs and get some relief on ADP, but its workable.  For a system that will ultimately cost $50 million or so, its worth it.