ExpertsFMScott Kupor on Building a16z and Rebuilding the Federal Workforce
The a16z Show · November 9, 2026
5 Selected Lessons:
case · 2 framework · 1 priority · 1 reasoning_pattern · 1
Right people, right objectives, right incentives
Source
The a16z Show — Scott Kupor on Building a16z and Rebuilding the Federal Workforce (November 9, 2026) · at 0:00 ▸ · at 26:42 ▸
When this applies
when diagnosing why any organization is underperforming, regardless of sector
The takeaway
Organizational health reduces to three questions in order: do you have the right people, the right objectives, and incentives that connect behavior to those objectives with real consequences for over- and under-performance.
Why it works
Organizations are just people aligned around objectives; if incentives don't reward behavior that serves the objectives, people will optimize for something else, and performance drifts independent of talent.
One level up
Incentive design, not stated mission, determines actual organizational behavior.
Where it stops holding
This stops working when external constraints (legal, political, risk tolerance) make it impossible to actually implement differentiated incentives even after correctly diagnosing the problem.
Detail
hierarchy:
- Right people in the right seats
- Right objectives the organization is aligned around
- Right incentives that reward achievement of objectives and correct/remove failure to achieve them
rationale: It doesn't matter if you're a government organization or a private organization or a not-for-profit — they're all the same: a collection of people lined up around objectives, and the deciding factor is whether incentives actually track those objectives.
In their words
Look, organizations are pretty simple. It doesn't matter if you're a government organization or a private organization or a not-for-profit. They're all the same, which is all an organization is, is a collection of people who are just lined around a similar set of objectives.
Disaggregate the 'John Doerr' job into specialized network functions
Source
The a16z Show — Scott Kupor on Building a16z and Rebuilding the Federal Workforce (November 9, 2026) · at 5:22 ▸
When this applies
when a single generalist expert relationship doesn't scale across a growing set of domains needed to serve a client
The takeaway
When a single-point-of-contact role can't scale with growing domain complexity, disaggregate the role into specialist functions surrounding (not replacing) the core relationship.
Why it works
A single person's relationship network is finite and non-scalable past a certain point; splitting the function into specialized roles lets expertise deepen in each domain while the core relationship stays intact.
One level up
Scale comes from institutionalizing expertise that used to live in one irreplaceable person.
Where it stops holding
This stops working when the domains being served aren't numerous or specialized enough to justify dedicated headcount, or when clients value the single trusted relationship over many specialists.
Detail
procedure:
- Identify the generalist role whose breadth is now the bottleneck (e.g., the all-purpose venture GP or talent agent)
- Keep the core relationship (GP-to-founder, agent-to-actor) intact as the entry point
- Hire functional domain experts (talent, BD, government relations, marketing) who do nothing else but that function across all clients
- Pitch the client on access to the whole team/table, not just the individual ('reverse pitch')
benchmarks: {"ovitz_caa_market_share":"~90% in 1990"}
rationale: Software ate the world, so the domain knowledge needed per deal kept expanding beyond what one GP could master; specialization became necessary to keep up.
In their words
The fundamental theory that Mark and Ben had was: let's hire GPs who are functional domain experts in the particular areas of investing they want, and then let's surround them with this failings of somebody who all they do all day long is manage CFO relationships or CEO relationships or VPs of engineering, somebody who all they do does government relations or business development or other stuff like that.
Diagnosing a 'do nothing' culture via misaligned risk-avoidance incentives
Source
The a16z Show — Scott Kupor on Building a16z and Rebuilding the Federal Workforce (November 9, 2026) · at 26:42 ▸ · at 52:30 ▸
When this applies
when an organization's actual (unstated) incentive is to avoid blame rather than to produce good outcomes, leading to institutional paralysis
The takeaway
Asymmetric risk incentives — large penalties for visible failure, no reward for invisible success or innovation — produce institutional stagnation regardless of the people's talent.
Why it works
When downside (public/political blame) is punished severely and upside (innovation) is never rewarded, the dominant strategy for any rational employee is to do nothing new, because doing nothing is the only reliably blame-free action.
One level up
You get the behavior your incentive structure punishes and rewards, not the behavior your mission statement asks for.
Where it stops holding
This stops working when the domain genuinely has low risk tolerance for correctness (e.g., life-or-death systems), where risk-aversion is actually the right incentive.
Detail
approach: Identify what people are actually optimizing for (e.g., 'keep you out of the news') versus what you want them to optimize for (good outcomes); introduce 'permissioned risk' sized to the actual stakes of the domain.
signals:
- staff explicitly state their job is to avoid controversy, not deliver results
- zero tolerance for failure paired with zero reward for upside
- the organization hasn't changed a broken process in 50 years despite it being publicly known to be broken
rationale: Venture capital tolerates risk because uncapped upside can cover losses; most government functions have neither the uncapped upside potential of VC nor genuinely life-or-death stakes, so near-zero risk tolerance is miscalibrated to the actual consequences.
In their words
So, part of the incentive structure that I think we have to put in place and we're trying to put in place is I've been using this term permissioned risk, which is deliberately different from the permissionless risk that I think exists in Silicon Valley, which is we got to have this conversation and we got to be willing to let people take some risk.
The Skype buyout gamble
Source
The a16z Show — Scott Kupor on Building a16z and Rebuilding the Federal Workforce (November 9, 2026) · at 12:14 ▸ · at 15:00 ▸
When this applies
when an early, high-conviction, against-the-grain bet either validates or destroys a new venture's credibility right out of the gate
The takeaway
A fast, visible early win can disproportionately accelerate growth because momentum begets momentum in reputation-driven markets.
Why it works
Early high-profile wins create narrative momentum that compounds — success attracts more capital and talent faster than the underlying skill alone would, because perception and reality reinforce each other in trust-based businesses.
One level up
In businesses governed by trust and perception, speed-to-proof-point matters as much as being right.
Where it stops holding
This stops working when the contrarian bet fails publicly, which can permanently brand a new entrant as reckless rather than visionary.
Detail
situation: a16z had just raised its first fund promising early-stage venture investing, then pursued a contrarian buyout of Skype (part of eBay) despite an unresolved IP dispute with original founders that scared off other bidders.
decision: Partner with Silver Lake and Michael Dell, work collaboratively with the disputing founders to moderate the IP risk, and close the buyout.
outcome: Sold to Microsoft 18 months later for 4x the purchase price, which changed the firm's narrative and let them raise their second fund well ahead of schedule.
lesson: An early, bold, correctly-executed bet that defies market skepticism can compress years of reputation-building into months.
In their words
But, anyways, everyone's like, look, this is crazy. But the beauty of it, literally 18 later, we sold to Microsoft for four times what we paid like 18 months before.
The Boyers, PA limestone mine paper retirement process
Source
The a16z Show — Scott Kupor on Building a16z and Rebuilding the Federal Workforce (November 9, 2026) · at 46:40 ▸ · at 50:15 ▸ · at 52:30 ▸
When this applies
when a visibly absurd, decades-old broken process survives for generations because fixing it carries political/career risk while leaving it broken carries none
The takeaway
Visibly absurd, long-known inefficiencies that nonetheless persist are strong evidence of a misaligned risk incentive, not of the problem being hard to solve technically.
Why it works
The asymmetric incentive (risk of blame for trying and failing vs. zero risk for leaving a known-broken process alone) preserves even obviously absurd processes indefinitely, because no individual actor's personal calculus favors fixing it.
One level up
The persistence of an obviously fixable problem over decades is itself diagnostic data about an organization's incentive structure, not about its difficulty.
Where it stops holding
This stops working when the inefficiency persists for genuinely technical or resource reasons rather than incentive reasons (e.g., true cost prohibitiveness).
Detail
situation: Federal retirees mailed paper applications that traveled through HR, payroll, and finally an underground mine where 800 people manually calculated and processed benefits, taking 3-12+ months, with articles describing this absurdity dating back to the 1970s-80s.
decision: Elon Musk and Joe Gebbia (National Design Studio) began digitizing the intake process; Kupor continued it, ending paper submissions entirely by June 30 and digitizing 400 million existing paper records.
outcome: Processing time cut by over 50% immediately, with a goal of 90% reduction and a 15-day turnaround commitment for electronic applications.
lesson: Decades-old, publicly known absurd processes often persist purely because no individual has career incentive to risk fixing them, not because they're hard to fix.
In their words
I didn't like discover this, and Elon, to his credit, didn't discover this. You can go back literally to the 1970s and 1980s, and you will see articles written about how fucking ridiculous this mine is. Nobody has ever tried to solve this because, God forbid, some congressional person is going to say, Oh, you screwed it up, or you didn't do this, you didn't do that.