It Takes More to Audit Federal Agencies and Programs Than DOGE Knows

Lambert here: Excellent interview with Wired’s Vittoria Elliott, worth reading in full. Hard to excerpt!

Terry Gerton: Victoria, you spoke to actual government auditors who wouldn’t go on the record about what they’re seeing. What did you find out?

Vittoria Elliott: I think something that’s really important to understand is DOGE has come in and said that what they’re doing, going into agencies, cutting their contracts, accessing sensitive payment or IT systems, is that fundamentally what they are doing is a quote-unquote an audit. And I think a lot of times when DOGE talks about what they were doing, they make it seem like we don’t already have systems to do these things within the federal government. And I think that’s sort of a hallmark sometimes of the tech industry generally. … [W]e already had cabs before Uber, we already had hotels before Airbnb and so just because that’s the case doesn’t again mean that there aren’t improvements to be made to industries or systems. But I think similarly DOGE has come into the federal government and sort of pretended like systems don’t exist to do these things where they do. And there are actually entire systems set up to be able to audit the federal government.

Every agency has an inspector general that can run an audit, that can investigate these things either based on complaints or congressional investigations that order them to do so. And there are trained people who know how to go through these systems and do forensic audits. And so I spoke to two auditors who have worked on technical and financial audits, meaning that they know how to go through the minutia of an agency or program spending. They know how to go into their systems and they know these processes should work. And what they told me was, first off, an audit normally takes six to 18 months and that’s not auditing an entire agency. Often that’s auditing a single program and that’s because, first off, you have to interview a lot of people who are involved in that program to really understand how it works. Then you’ve got to get into the nitty-gritty of their books, of their systems to really figure out what’s going on. And then you write up a full report where you sort of cite all these things and make actual recommendations about how these things can be fixed and that gets submitted often for public consumption.

So I think when we’re talking about efficiency here, we really need to ask the question efficient for who and efficient towards what end and in what way because the thing is, all these people who work for the government, who are losing their jobs, they’re saying, ‘My job being eliminated won’t make things more efficient.’ Cutting people who work, for instance, in the call centers for Social Security, that’s not going to make that system more efficient for an elderly user who maybe doesn’t feel comfortable using an online portal to access their benefits. That makes it deeply inefficient for them, but it is quote-unquote efficient to have less people that you need to manage, have less people that you to pay if you’re thinking about it from this purely business perspective, which a lot of these people have. And I think AI is part of that. The big push of AI is that it makes things more efficient. It makes these systems more efficient, it means you don’t have to have as many people on staff, but efficient for whom? Yeah, that’s maybe more efficient for someone who is a manager, someone who is like DOGE, who maybe they don’t want to oversee a ton of people at an agency or whatever, but is that efficient for the American taxpayer? Is that efficient for the American citizen who’s going to have to interact with that system?

But because those systems don’t talk to each other, it can feel really inefficient that you have to put in the same information multiple times from multiple agencies. When DOGE talks about it as if we just made all these data sets talk to other, it would just be so much more efficient and effective. Well, yeah, but it would also be such a higher risk for so many people, and again, would make everybody more vulnerable to having their information compromised.