Most conversations about AI and founders focus on tools.
Which models to use. What to automate. Whether to build or buy.
I think the more interesting question is what happens when intelligence and execution stop being scarce.
If almost every ambitious founder can access extraordinary analytical and executional leverage, what actually separates the companies that matter from everyone else?
I asked Reid Hoffman, co-founder of LinkedIn, Manas AI, and Inflection AI and partner at Greylock, three questions about exactly that.
Not about tools.
Not about specific technologies.
About the fundamental shifts in what it takes to build something that matters.
His responses are presented below, followed by my takeaways on what they mean for founders.
1. You’ve backed some of the best founders alive. When you imagine the founder who builds the most important company of the next decade, what does their relationship with AI agents look like that’s fundamentally different from how today’s best founders work?
REID HOFFMAN
Every generation of great founders is native to something. The best internet founders never had an “internet strategy” per se, they just couldn’t imagine the business any other way. People like to think everything changes about heuristics in different trend waves, but that’s not necessarily true.
I’d look for founders who see everything in an AI native way. The founders who build the most important companies of the next decade won’t have an “agent strategy”. Their first reaction to almost any problem will be, “Can my agents fix this?” And their second will be, “What would I have to give them so they could?” They assume their most capable colleagues are models, and they build the company on that foundation.
The second difference is that they natively understand, and sprint to build, learning loops. I’ve said for years that the founder with the faster OODA loop (observe, orient, decide, act) wins and the other one dies. The game has become building similar loops for the agentic infrastructure inside your organization. An AI-native company is really a set of feedback loops (product, distribution, analytics, spend) hill-climbing against an objective function 24 by 7. The founder’s job is to constantly close as many feedback loops as possible as quickly as possible, and then add leverage.
Third, they’re better allocators. That has always been the job—founders have always allocated capital and people—but an AI founder keeps three new ledgers. They allocate:
People: which decisions stay with humans and which get handed to the machine, and where the few people they do hire sit in the loop.
Context: what an agent knows determines what it can do, so context gets designed as carefully as an org chart.
Compute: A frontier model that costs the equivalent of $10,000 an hour belongs on the problem that deserves it, smaller models on smaller problems.
And lastly, none of this exempts you from the old rules. If anything, they matter more. A good product with great distribution has always beaten a great product with poor distribution. That continues to be true. So the AI founder also has a very strong thesis on network effects and distribution: why customers won’t leave, why they bring others, and why being first to scale (which is different than first to market) still decides winners.
PETAR DIMOV
What stands out here is that AI-native does not simply mean using more AI.
It means redesigning the company around faster learning, sharper allocation, and tighter feedback loops.
The technology may change the speed of execution, but the founder still has to decide where that speed should be pointed.
2. AI is making intelligence and execution increasingly abundant. If intelligence stops being scarce, what do you think becomes scarce enough to become the defining advantage of the best founders?
REID HOFFMAN
Fifteen years ago I said something that got quoted more than I expected, that social networks do best when they tap into one of the seven deadly sins. It was an attempt at being a bit cheeky, but the underlying point was that founders—especially in the consumer game—needed to have a deep understanding of the human beings they’re trying to serve.
Technology succeeds when it’s built on an accurate model of human beings—what people actually want, as opposed to what they say they want.
When a model can generate thousands of somewhat plausible products before breakfast, someone is going to need to figure out which one will actually build habits around. I don’t think that is a compute problem. Human needs, aspirations, and dreams are all older than the tech trend of the year. Our job as technologists and founders is to understand the gap between where humans are and where they want to go, and build technology that bridges the divide. That gets them to a meaningfully better place, and puts society on the right track.
Somewhat related, EQ is also scarce. A company is a set of relationships before it’s a set of systems, operating in a larger network of systems. When you’re starting out, you still have to get a co-founder to say yes, people to leave safe jobs and bet on you, customers to sign contracts, and investors to give you massive sums of money. In an AI world, the team is small and the leverage is enormous. So, every one of those relationships matters way more. And EQ runs inward as well: you’re the absorber of everyone else’s stress, with no one above you to vent to. Reading people, and knowing how to manage yourself, becomes a very important job.
So the short way to put it, one of the scarcest skills is understanding networks, human behavior, and people.
PETAR DIMOV
This is where the conversation shifts from capability to judgment.
If technology can produce more options, more products, and more execution at lower cost, the harder problem becomes understanding which of those possibilities actually matter to people.
That makes human behavior less like a soft skill and more like a core part of product strategy.
3. You’ve watched competitive advantages come and go across multiple technology cycles. What’s one thing founders treat as a durable advantage right now that you think is far more fragile than they realise?
REID HOFFMAN
Right now, particularly as founders are more and more AI native, they need to figure out how to build trust with consumers and the market. It’s imperative for all companies to both communicate and demonstrate trustworthiness, particularly in AI where consumers are skeptical and, even more so, afraid of the effects on their jobs, livelihoods, and communities.
This could change. Companies need to demonstrate trustworthiness now, but that need becomes less important if the market decides that whatever norms are in place are enough to move forward.
Trust and speed don’t trade off the way many founders assume. Founders must take smart risks, make room for innovation along the way, and not treat cultivating trust as avoidance for launching, scaling, and innovating.
In general, people overweigh what could go wrong and underestimate what could go right, when they could be launching, gathering feedback and learning from it, iterating on it, and creating the conditions for everyone else in the market.
PETAR DIMOV
The interesting part is that Reid doesn’t frame trust as the opposite of speed.
He frames it as something founders have to build while moving quickly.
In a market where skepticism can slow adoption, the ability to earn confidence without slowing innovation becomes part of the job.
The Thread Connecting All Three
PETAR DIMOV
Reading Reid’s three answers together, something becomes clear that isn’t obvious in any single answer on its own.
In every generation of technology, the raw capability eventually becomes more accessible.
The internet gave everyone a website.
Mobile gave everyone a camera and a storefront.
AI is giving founders access to capabilities that previously required entire teams: research, coding, analysis, content, and execution.
The interesting question is what happens after access is democratized.
The founders who create important companies are not necessarily the ones who access the capability first.
They are the ones who know what to point it at.
That is the thread running through all three of Reid’s answers.
Build faster learning loops.
Understand human behavior.
Allocate people, context, and compute.
Build distribution.
Develop relationships.
Earn trust.
None of these become less important because AI becomes more capable.
In some cases, they become more important precisely because the technology is becoming more capable.
These are not secondary considerations in building a company.
In an AI-native world, they become central to the job itself.
They determine what the intelligence gets pointed at, what the company chooses to build, and whether anyone trusts what it creates.
The founders who understand this aren’t thinking about AI as a capability to adopt.
They’re thinking about it as an amplifier.
And asking themselves very carefully what, exactly, they’re amplifying.
Links
Follow Reid
LinkedIn · Substack · Instagram · X · YouTube
Reid Hoffman
Follow Petar
LinkedIn · Substack · Instagram · X · YouTube
Continue Exploring the Frontier
If this piece resonated, you may want to go deeper.
This article is part of our 22nd Century Brief collection, where we explore ideas, frameworks, and strategies alongside the leaders shaping the next era of entrepreneurship and investment for the 22nd Century.
You can also explore our main topic categories to discover more insights across entrepreneurship, venture capital, fundraising, company building, and frontier technologies.
