Getting the right people in the right seats

In June, we celebrated the launch of Launchpad with a fireside chat with Scott Kupor, Director of the US Office of Personnel Management.

Scott leads the agency responsible for managing the federal civilian workforce, a role he started last July after nearly three decades in the private sector. Before entering government, he was managing partner at Andreessen Horowitz, the venture capital firm he helped start in 2009, and a former chairman of the National Venture Capital Association. He also writes a Substack, Secrets of OPM, which I highly recommend (at least to fellow government capacity nerds).

Our conversation, excerpted here, came a few days after two major presidential directives on AI talent: an executive order expanding the government’s AI talent strategy and a national security memorandum establishing an AI National Security Strategic Reserve for non-governmental talent. OPM was given central tasks in both, and Scott offered an early look at how the strategic reserve might take shape.

Scott is unusually well-placed to talk about the private-to-public move, having made it himself and now leading the agency trying to widen the path for others.

We cover:

• What government can learn from venture capital about recruiting and using talent well

• Why the federal hiring pitch needs to change

• How the new US Tech Force program is going, six months in

• How to build a high-performance culture inside a workforce of two million

• An early look at the Trump administration’s idea of a strategic AI talent reserve, and where he thinks AI is headed

Remco Zwetsloot: You spent a lot of your career at Andreessen Horowitz, where figuring out what makes great talent and great ideas is core to the work. What from that experience has carried over to how you approach your role at OPM? And more generally, what are some of the lessons government can learn from venture capital and business?

Scott Kupor: I think there are a couple of things. Let’s start with the most obvious one, which I still think is the most foundational. In venture capital, if you talk to any venture capitalist, almost everything we’re trying to do when evaluating opportunities is to evaluate founders. The reality is that every organization, whether it’s a startup or a for-profit, a government entity, a nonprofit, or even an academic institution, are just collections of individuals. And individuals do one thing very well, which is respond to incentives.

I think that is very easily translatable to government. Anything you want to do inside of any organization is fundamentally a question of: do we have the right people sitting in the right seats, and have you created the right set of objectives, and incentives that line up with those objectives? Almost all organizational problems can basically be boiled down to that.

But some things are different, unsurprisingly. In the venture capital business, we take crazy amounts of risk. We look for things where we’ve got capped downside and completely asymmetric upside. And we can afford to have a terrible track record. What happens in our industry is we’re wrong half the time. You can afford that because, if you get it right, there’s some 10 to 20 percent of your investments that go crazy.

Government is obviously very different from that. But one of the things I’ve noticed in trying to bring this into government, at OPM as well as more broadly, is there has to be some ability to think about upside capture and downside risk. And at least my observation, which is probably not surprising to anyone who has worked in government, is that we’ve created such a compliance and oversight culture, whether that’s the GAO or your OIG. When I first got there, a bunch of people told me, “Our job is to protect you from getting embarrassed in some congressional hearing.” Which I appreciate, that is very nice, but I told them that’s not the objective.

So one of the things I think we have to do is be able to take measured risk inside the government. We can’t take crazy amounts of risk, and obviously the risk tolerance will depend on what the downside is. What we have to do, and this goes back to the incentives, is create a permissioned culture of reasonable, measured risk-taking in government, with some balance against what the upside capture is.

What’s pervasive across government, from what I’ve seen, is that we’ve created a culture that essentially makes that almost impossible to do. So a huge portion of what I think we can bring from the private sector, in a government-focused way, is a way to think about making decisions where we don’t assume the answer is “no” unless it survives every conceivable review, but where there’s some modicum of this making sense relative to what we think the upside capture is.

So the risk-averse culture wasn’t a huge surprise. On the flip side, coming up on a year, have there been things that have been surprising to you about working in government?

Scott Kupor: Let me start on the very positive side, and again, this is probably not surprising to people in government: the vast majority of people are good, qualified people who are here because they really, really care about the mission. Just like any other organization, there’s always some group of people trying to do the minimum they can, but that’s a very small minority, and frankly that’s true in any organization. So the quality of the people and the dedication to the concept of public service, for those of you who’ve been here, that’s probably obvious. As an outsider who spent 35 years in the private sector, I never really had exposure to it, so that was a positive surprise to me.

The downside surprise, again probably only to me, is the difficulty of going from the conception of an idea to actual execution, and actually getting things through. Some of that is rightful: we have to think about whether an idea is a good thing, and people’s definitions of what’s good varies.But the other thing is just the complexity of the organizations within each agency, which I don’t think you realize until you get here.

There are 300,000 hiring managers inside the government. So I can have a conversation with my political counterpart at an agency, and we can all agree that something is a good idea to do, but the amount of telephone that has to happen inside the government for things to actually get translated to the level where there’s execution is amazing.

And the fidelity in that telephone chain is not great. And I don’t mean that in a bad way. People are not literally trying to garble things as they go down the chain—it’s just that the number of hops on the chain is so great that there’s massive packet loss, for those of you who know the networking term, in every handoff.

There are 300,000 hiring managers inside the government. So I can have a conversation with my political counterpart at an agency, and we can all agree that something is a good idea to do, but the amount of telephone that has to happen inside the government for things to actually get translated to the level where there’s execution is amazing.

I was at a meeting today over at the White House and we were talking about TechForce, and someone said, “Explain it to me again.” I walked out of the meeting thinking, I think I’ve had this same conversation with this guy 50 times. And this is not a dig at him. It’s just that we forget that the stuff that’s important to us isn’t necessarily important for everybody around us. Then you multiply that against 300,000 hiring managers, 2 million civilian employees, and 250-plus agencies—the risk of things being lost in translation is just massive. From the outside, I would have believed that once we generally agree on what the things are and agree we’re going to do them, the execution kind of happens.

People in this room, and others we advise, are often considering coming from the private sector into public service, at least to try it out for a year or two. As someone who has gone through that yourself, what do you tell people who are on the fence about making the jump, whether it’s for something like TechForce or something more permanent?

Scott Kupor: You’re probably not surprised I’d say this, but look: 100 percent do it, basically. And I think what you said is really important and I want to highlight it: we have to have a culture where we don’t expect people to make 10-, 20-, 30-, or 40-year commitments. Maybe there was a time in my generation where that made sense. My sense is that for anybody under the age of 40 today, that concept just doesn’t make sense. People don’t think in commitments of that length.

When I first got here, one of the things that troubled me, talking about surprises, was the extent to which we have used the concept of permanent or lifetime employment as an inducement for people to come here. In fact, one of my managers, literally on day one, came up to me and said, “You guys are ruining things.” I said, tell me more. And she said, “Our whole value proposition with people has been: come to the federal government and you have essentially lifetime employment.” I said, all right, well, you’re not going to like me any better when you hear the answers I have. First of all, it’s just a lie, and we should be transparent: there’s no such thing as permanent employment. It doesn’t exist in government, and it doesn’t exist anywhere else. The world is, unfortunately, a fragile and changing place.

But more importantly, that’s just a horrible marketing pitch. I don’t know anybody for whom that’s an attractive pitch. The pitch has to change. The pitch has to be: come here for a period of time, work on the great, complex, challenging problems we have, and by the way, in doing so you are furthering your career, whether you decide you want to stay in government, go to the private sector, or go to a nonprofit or something else. Somewhere along the line we lost that thread. We have to find a message that’s attractive to the people we’re actually trying to recruit.

Come here for a period of time, work on the great, complex, challenging problems, and in doing so further your career, whether you decide you want to stay in government, go to the private sector, or go to a nonprofit, or something else.

So those are some of the big things we’re trying to do, whether it’s TechForce or more broadly: how do we help people understand that you can have a big impact without committing your entire career to this? We have to meet people where they are, and historically I think we haven’t. And to me that’s everything from things like TechForce to finding ways to intermingle public and private industry more broadly, making it easier for people to go back and forth, and knocking down barriers, whether in terms of compensation, deferred compensation, or all those other things. That’s a lot of the message we’re trying to push.

It’s been about half a year since the US Tech Force launched. What’s the level of interest you’ve seen, and how has the process been of converting that interest into people actually starting jobs and working on projects in agencies?

Scott Kupor: The interest continues to be great. The program, for those of you who are interested, is intended to be a two-year program where we’re trying to bring a thousand early-career engineers into government. We’re trying to do a couple of things. One, there’s an enormous amount of modernization work that has to be done, and a huge portion of that, of course, is having the talent inside government to do it. That’s not to say we don’t have any talent in government, but I think we can augment the talent in a very different way.

There’s an enormous amount of modernization work that has to be done, and a huge portion of that is having the talent inside government to do it.

The other thing TechForce is trying to do, more broadly, is figure out a message to people, particularly early-career. Roughly seven percent of the federal workforce is under the age of 30. Across the entire US workforce, it’s about 22 percent. So by a factor of about three to one, the federal government is underindexed on early career. On the other side, 44 percent of our population is over the age of 50, compared to a little over a third in the general US workforce. So even if we did nothing else from a tech perspective, we have this pending demographic challenge, which is that close to half of our federal workforce will retire soon, and we don’t have a plan to bend the curve.

So in addition to TechForce solving a technical gap problem, we’re also trying to use it as a showcase: can we successfully start to bend the curve on this demographic trend? And we’re expanding that into a broader early-career conversation on the policy side as well.

The short answer to your question is that it’s going very well. We have a little over 200 people now actually working in the program. My frustration with it, which is always the frustration, is that I wish it went faster. The translation from “we agree this is a good idea, we need more TechForce engineers” to “okay, have you actually made an offer to somebody,” there’s still a big telephone gap between agreed and done. So a lot of what our team is doing is literally at the tactical level: agency X knows they want people, they have the budget, they have the headcount, so what is causing them not to move? Sometimes it’s literally that one person among the 300,000 hiring managers either didn’t get the memo or has a different understanding of what they need to do. There’s all kinds of things—”you name it, I’ve heard it,” is how I’d describe it. So we’ve had to be very deliberate about where we put our resources.

A huge portion went into front-end recruiting and getting people through the pipeline, and now, quite frankly, those resources need to go into back-end execution to get all the rest of it filled. But we’re going to get there. The talent we’re seeing is phenomenal, but getting from here to there takes some time.

As you well know, with the federal government you have to get people hired, but once they’re there, you also have to set them up to actually do great work on the inside. What are some of the key ingredients for success there? And from your position at OPM, running a program like TechForce without being in the agencies where people are actually going to go, what are some of the tools you have, or things you tell agencies, to make that happen?

Scott Kupor: Objective number one is getting the right talent in the right seats. TechForce is part of that, and more generally having a hiring process that’s better and more efficient across government. Objective number two, which goes directly to your question, is: once people get here, what’s the environment they walk into, and does it actually attract and retain the kinds of people we want? The idea there is to create a high-performance culture inside the government. A huge part of what OPM does is policymaking, in addition to being a talent organization.

Objective number one is getting the right talent in the right seats.

What does a high-performance culture mean? It means you have to have accountability on the low end, and historically in government we’ve done a really poor job of that. Everybody gets ranked one through five every year in most agencies, five being you’re awesome and one being you’re not. Today, 99.7 percent of employees in the federal government get a three, four, or five, and 70 percent of people get a four or five. That leaves 0.3 percent getting a one or two, which would mean you’re not succeeding. I wouldn’t prescribe what that number should be, but in a population of two million people you might expect a somewhat more normalized distribution.

So what are we doing? Number one, from a policy perspective, we have policies in place to cap the percentage of people who can be at the top, in the fours and fives. We didn’t put a target on the low end, but we are making a bunch of policy changes to make it easier for managers to put people on plans, and to not have to take 12 to 24 months going through a very comprehensive and expensive process to remove people, while still giving them due process.

Number two, more generally: how do we make sure that merit matters not only in the hiring process but in actual performance? To do that, we’re trying to eliminate anything that’s a proxy for merit. “Do you have a college degree” is an interesting question, but having a college degree itself doesn’t tell me a whole lot about you. If you’re applying to be an engineer in TechForce and you happen to have been a Russian literature major in college, great, you have a college degree, but what’s the applicability of that knowledge to the job? And maybe Russian literature majors are also great coders, but we can assess that by assessing their coding skills rather than looking at the degree. Tenure has also been used as a proxy for skills in federal hiring; almost all jobs have had tenure requirements.

On the front end, on TechForce, we’re hiring people at the GS-14 level, which is unheard of for early-career people, because we basically rewrote all the job descriptions to say that college degrees and tenure no longer determine your leveling in government. What determines it is your performance on a skills-based coding test and a subject-matter-expert interview, so we can level people appropriately for what they bring.

And once they’re here: we just put out a notice of proposed rulemaking to eliminate time-in-grade, which is also a tenure-based proxy. For most jobs in the federal government, if I hire you at the GS-5 level and you knock the cover off the ball in terms of your performance, I can’t promote you for another 12 months, and once I promote you, you have to sit at that level for another 12 months before the next one. We’re eliminating that, basically to say: I don’t really care how long you’ve been here; if you’re performing at the next level, we need to move you up.

These are all things we’re trying to do to create an environment where, as we bring in people who are incredibly good and really aggressive about their career growth, we don’t have a system that spins them out the other end because they get frustrated every time they try to get a merit-based promotion, or recognition, or awards commensurate with the skills they bring to the job.

It was a big week in AI policy last week, first with an executive order expanding government AI talent strategy and then an AI national security memorandum that included a new concept of an AI National Security Strategic Reserve. OPM was given central tasks in both. Can you say a little bit about what the strategic reserve program may look like, as far as we know, and how it might work?

Scott Kupor: I’ll tell you what I think it will look like, because we are just starting to figure out what it’s going to look like. “Strategic reserve” throws people off a little bit, because you usually think of something like the petroleum reserve: usually it’s commodities, not people. But in this case the strategic reserve is people.

The basic idea is that we’re going to work with the intelligence community and the security apparatus to make sure we have a pool of talent in the private sector, and potentially also in the academic community, that we can tap as subject-matter experts when things come up or when we need something. It could be as simple as literally having somebody on speed dial, so that when the next thing comes up, everybody knows what to do. Or it could be: hey, we have this really cool project, can we get somebody to come in as an SGE, a special government employee, for six or nine months.

We’re still working through it, but that’s the basic idea, aligned with trying to more actively make sure the government taps into the private sector in a way that’s conducive and constructive. And I want it to go the other way too. One of the things I would love is to have more people from the public sector also spend some time in the private sector. For example, 97% of the Senior Executive Service today have been in government consecutively for more than 12 years. There’s nothing wrong with that, by the way, but imagine if, on the way to becoming an SES, we had a program where you could go spend six, nine, twelve months in the private sector and then come back to your job. The whole idea is getting a different experience: what it’s like to operate in the private sector, how they use technology, whatever the case may be. We have to do these things both ways.

To make this work, though, there are a bunch of other hurdles we’ve had to get over. A lot of people in the private sector have stock-based compensation, for example. We were lucky, working with the general counsel’s office and the DOJ, to get a legal opinion that says: if you are at OpenAI, for example, and you decide to take a tour of duty in the federal government, you don’t have to forfeit the vesting of your stock while you’re in government. There are small things like that which, when you start to dig into these problems, as you know well from the work you do, can become hurdles to this kind of public-private exchange.

So as part of standing up this program, we want to clear all those things, so that hopefully we have a reserve full of first-round draft picks and we can actually use them.

Zooming out a little bit for a final question on AI’s potential in general. On your Substack you wrote a post called “This Time’s Different“ about how AI can change government functioning and service delivery. People often reach for different historical analogies when they talk about AI, as with technology generally, and what the future will look like: electricity, the steam engine, the internet. Other people say it’s entirely unprecedented and we don’t know what’s coming. What are some of the analogies you reach for, and where do you think all of this is going?

Scott Kupor: I’ve been a technology person my whole career, so not surprisingly I’m a fan of technology generally. In that post I reference the work of Carlota Perez, a professor who has studied major technological revolutions. And look at what is factually true: it has always been the case that technological revolutions have created massive surpluses and economic growth, and ultimately net job creation. The process along the way can be disruptive, no question about that. But if you get to the other side of these things, what she calls the deployment phase, these technologies have been unabashedly positive from a societal perspective. You might argue otherwise if you think we should all still be farm workers, but that’s not really my position.

What’s also true, as you said, is that AI may be different. AI is certainly unprecedented. But the way I’m trying to frame the conversation, particularly inside the government, is this. Every time we talk about doing something in government, and this is true of any organization, the response is: great, if we had unlimited resources, of course we’d be able to do all those things. I have never once sat in a budget review where a manager came back to me and said, you know what, I don’t need those extra resources. The reality is we still live in that world today, and personally I don’t want to live in a world where the answer is always 20 percent more expense for 20 percent more work.

So what I’m really trying to get people to think about is: what are all the things we’re not doing today that we could do to dramatically increase the level of service the government provides? And I can tell you, we are far from perfect today. If we didn’t have people doing things they don’t need to do, because AI lets us increase the efficiency of our call center support, for example, we would be massively better off, and we wouldn’t be talking about reducing headcount; we’d be talking about delivering more output per person.

So I’m as optimistic as I possibly could be. Yes, as in any technological revolution, there are some jobs that will change, definitely. But I think the opportunity, and I think the job of government, is to look at the opportunity for people to utilize the technology, adapt their skills, and find ways to improve their lives. The last thing you want to do when you’re encountering a technological revolution is to try to turn back the clock.

The change will happen. The question is whether you allow people to embrace it, give them the tools, and make sure they have what they need to adapt to this new environment. That, I think, is at least the right frame.