Airbnb writes 60% of its code with AI. Klarna ran a function on it, then learned its limits. Gamma hit $100M with 50 people. Here's what the best founders actually do.
Most people use AI to save ten minutes. A handful of founders use it to compress ten hires. That gap is the whole story.
Ask the average professional how they use AI and you will hear the same short list. They draft emails. They tidy up a Slack message. They rewrite a paragraph that was fine to begin with. Useful, sure. But it is the productivity equivalent of buying a Ferrari to sit in the driveway.
The founders pulling away from the pack are not using AI as a fancy autocomplete. They are using it as infrastructure: to write their software, run a function, hold the line on headcount, sharpen their decisions, and redraw the shape of the company itself. None of this is theoretical. Every example below is on the public record, with names and numbers attached.
Here are the five moves, and what each one is really teaching you.
1. They ship the product itself, not just the to-do list
The most expensive thing a founder used to need was an engineering team. AI is quietly dissolving that constraint.
On Airbnb's first-quarter 2026 earnings call, CEO Brian Chesky told investors that nearly 60% of the code his engineers produce is now written by AI, which he estimated to be roughly twice the industry average. His point was not the percentage. It was the consequence: the company is shipping more features and iterating faster with the same people.
Airbnb is not an outlier; it is the median catching up to the frontier. On Shopify's earnings call days earlier, president Harley Finkelstein put the company's AI-generated share of code at around 50%. Google has publicly pegged its own figure near 75%. The work that once justified a twenty-person team can now be pushed forward by a small group directing AI under supervision.
The lesson: The bottleneck has moved from who can write the code to who knows what to build. Clarity of intent is now the scarce skill. If you have a sharp picture of the product and a willingness to direct rather than do, you can build something this year that would have required a funded engineering org five years ago.
2. They put AI on the front line, then learn exactly where it cannot stand
In February 2024, Klarna and OpenAI announced a number that ricocheted through every boardroom in fintech. Klarna's AI assistant had handled 2.3 million conversations in its first month, two-thirds of the company's customer service chats, doing work the company estimated as equivalent to 700 full-time agents. Resolution time dropped from 11 minutes to under two. By Klarna's Q3 2025 reporting, that figure had climbed to roughly 853 agents and about $60 million in annual savings.
Now the part most write-ups skip, and the part that actually matters.
By May 2025, co-founder and CEO Sebastian Siemiatkowski publicly walked it back. The company had cut its human support too aggressively, conceding that a cost-first push had delivered "lower quality," and began rehiring people for premium and complex cases under what he called an "Uber-type" model. Klarna did not pull AI off the high-volume tier. It put humans back on the hard edge: disputes, empathy, the nuanced cases where a wrong answer costs you a customer for life.
The lesson: The winning move is not AI instead of people. It is AI on the front line of high-volume, low-ambiguity work, with humans reserved for the moments that require judgment and trust. The founders who got burned automated a task without redesigning the system that task lived inside. Deploy AI where the cost of a mistake is low and the volume is high. Keep a human on everything else.
3. They run lean on purpose, because revenue per head is the new scoreboard
For two decades, a startup's growth was measured by how fast it could hire. That instinct is now a liability.
Consider Gamma, the AI presentation platform. In a $68 million round announced in November 2025, led by Andreessen Horowitz, the company confirmed it had crossed $100 million in annual recurring revenue at a $2.1 billion valuation. It had been profitable for over two years. The team size: roughly 50 people. That is about $2 million of revenue per employee, several times the efficiency of a traditional software company at the same scale, with only around $23 million in initial funding burned to get there.
This is the early shape of a thesis Sam Altman floated back in 2024, in conversation with Reddit co-founder Alexis Ohanian. Altman described a private bet among tech CEOs over who would build the first one-person, billion-dollar company, the kind of business that he said "would have been unimaginable without AI and now will happen." It is not science fiction. Instagram had just 13 employees when it sold for a billion dollars in 2012. AI simply lowers the floor on how few people a serious company requires.
The lesson: Headcount is a cost, not a trophy. Before you hire, ask whether the role is a judgment role or a volume role. Volume roles increasingly belong to AI systems you supervise. The founders winning today protect revenue per employee the way an earlier generation protected gross margin.
4. They treat AI as a research-and-decision engine, not an answer machine
Most people ask AI for an answer. The strongest operators use it to think.
Gong's 2026 research, built on an analysis of 7.1 million sales opportunities across more than 3,600 companies, found that revenue teams regularly using AI generate 77% more revenue per representative than those that do not, and that organizations embedding AI into their core go-to-market are 65% more likely to lift their win rates. Gong co-founder and CEO Amit Bendov framed the mechanism precisely: "Humans are making the decision, but they're largely assisted." AI as a second opinion on the intuition and guesswork that used to govern strategy.
The same pattern shows up in how new ventures are built. McKinsey's 2026 work on AI venture building documented a company that automated its upper-funnel research and outreach with agentic AI and saw outbound volume rise 25-fold, with click-through rates more than doubling against the old human-only process. Others are using AI to spin up "synthetic personas" from real interview transcripts and call notes, an always-available voice of the customer to pressure-test messaging before a dollar of budget is committed.
The lesson: Point AI at the expensive, slow parts of building a company: market research, customer discovery, competitive intelligence, prioritization. Use it to front-load the learning that used to arrive too late. You still make the call. You just make it with far better evidence and far less delay.
5. They rebuild the org around the work, starting with their own job
The final move is the one founders find hardest, because it implicates them personally.
On that same Airbnb earnings call, Chesky said the company has no room for "pure people managers," nor for thousands of hands-off managers who supervise work they cannot do. Design and engineering leaders, he said, are back to writing code themselves or directing AI tools to do it. Coinbase ran the same play from a different angle: CEO Brian Armstrong announced the company was flattening its structure to a maximum of five layers below the top.
This is the structural truth underneath every example above. McKinsey describes the new operating model bluntly: humans orchestrate, supervise, and intervene, while agents execute the research, analysis, and coordination. The org chart stops being a hierarchy of people managing people and becomes a small core of judgment directing a large layer of capability.
The lesson: The role of the founder is shifting from doing the work to orchestrating the system that does the work. That demands fewer layers, more hands-on leadership, and a willingness to stay close enough to the work to direct AI well. The leaders who insist on managing from a distance are the ones being managed out.
The thread that ties all five together
Read those five moves again and notice what they share. Klarna put humans back on the cases that require trust. Gong's data says humans still make the decision. Airbnb keeps engineers in oversight of the code. Gamma kept taste and judgment at the center of a tiny team. In every single case, AI did not replace the founder's judgment. It amplified the founders who had it, and it exposed the ones who did not.
That is the real line between writing emails and building companies. One uses AI to do the same job slightly faster. The other uses it to do things that were structurally impossible a year ago: ship a product without an engineering army, run a function at a fraction of the cost, stay lethally lean, decide with better evidence, and lead a flatter, faster organization.
The tools are now available to almost everyone. The advantage goes to the founders who stop using them as a shortcut and start using them as leverage.
Sources: Airbnb Q1 2026 earnings call (via TechCrunch and Quartz); Shopify Q1 2026 earnings call; Klarna and OpenAI press releases, CNBC, Bloomberg, and Fast Company; Gamma Series B announcement (Business Wire, November 2025) and Sacra; Sam Altman in conversation with Alexis Ohanian (2024); Gong 2026 State of Revenue report (via VentureBeat); McKinsey, "How to build businesses faster and better with AI" (2026); Coinbase via CEO Brian Armstrong. All figures reflect public statements as reported at the time of writing.
