The article is written from the perspective of a business / management consultant, rather than a public policy shop perspective. In general, I think social problems move slowly, and solving them in a three year business plan isn't realistic. You'll see many agencies use a version of Mayne's Framework or Contribution Analysis to report on progress for big social problems.

It's not that they perpetuate their own raison d'être, it's that they are addressing path dependent social problems, and changing a system with embedded systemic memory within a vast number of crevices (public, private, and cultural) to hide those memories is orders of magnitude more effort than creating the system at the start.