Brian Chesky stared at the wreckage of his company in 2020. Bookings at Airbnb had collapsed 80% in weeks. The advice he once followed — hire talented executives and step back — had left him disconnected from the product that defined his vision. So he changed course. He cut a quarter of the staff. He reorganized around functions instead of business units. And he started making decisions on hiring, firing and product direction several layers down the org chart.

That shift, later labeled founder mode by Paul Graham in a 2024 essay, has moved from Silicon Valley debate to boardroom imperative. As artificial intelligence compresses information flows and shrinks the value of traditional management layers, leaders who stay close to the work gain an edge. Those who don’t risk watching nimble competitors — or even small teams armed with agents — pass them by.

The Origins of a Leadership Reckoning

Chesky told a Y Combinator audience that conventional management nearly destroyed Airbnb. Professional managers treated parts of the organization as black boxes. They optimized locally. The founder, with context spanning the entire company from its earliest days, saw connections others missed. Graham’s essay crystallized the contrast: manager mode relies on delegation through direct reports who shield the CEO from details. Founder mode rejects that. It normalizes skip-level meetings, values individual contribution over title, and keeps the leader in the details that matter.

Examples abound. Steve Jobs obsessed over product pixels. Elon Musk walks factory floors and reads code. Jensen Huang maintains intense involvement at Nvidia. These aren’t quirks. They reflect a style that scales differently when the product is software or, increasingly, AI systems.

Fast forward to 2026. AI has accelerated the logic. Jack Dorsey slashed 4,000 jobs at Block in February, arguing that hierarchy once existed to move information a single person could not track. Now AI handles routing. Block’s stock rose on the news. Grey Journal reported that Dorsey explicitly tied the cuts to AI’s ability to replace layers of coordination.

At Google, Sundar Pichai reported that 75% of new code is AI-generated. Airbnb itself claims 60%. These figures aren’t just productivity gains. They signal a smaller, faster organization where the founder or CEO can once again touch the core work. Chesky has said founder-mode companies stand a better chance of surviving the AI age because proximity to both the technology and the people using it beats distant oversight. A Times of India article from last year captured his view directly: “Founder-mode companies are more likely to thrive or even survive the age of AI.”

But survival demands more than nostalgia for startup days. AI collapses the information advantage of size. A handful of engineers with powerful models can match the output of much larger teams. That forces a choice. Bulk up with layers of managers who translate strategy into execution. Or flatten, empower hybrid manager-makers, and keep decision rights closer to the top.

Satya Nadella flattened Microsoft’s hierarchy in 2025, brought in talent with direct access to his office, and focused on infrastructure competition. Sergey Brin returned to hands-on coding and AI agent work at Google amid a “code red.” These moves echo Chesky’s pivot. They reject the idea that scale requires professional detachment.

CEOs Who Own the Agenda See Different Results

Research backs the pattern. Bain & Company found that companies stuck in pilot mode for AI share a common trait: the CEO treats the technology as someone else’s problem. Firms where CEOs spend 15% to 25% of their time on AI see faster adoption and measurable business impact. The consultancy outlines four stages: understanding the shift, running scattered tests, evolving the business, and leading full transformation. Most organizations linger in stage two. Bain’s analysis concludes that scaling requires CEO ownership of platforms, talent, and priority use cases rather than siloed experiments.

McKinsey reached similar conclusions in late September. CEOs fall into three camps: technology supporters who hand AI to CIOs, productivity champions chasing cost cuts, and visionaries who treat AI as central to competitive advantage and personal responsibility. The share of visionaries is rising. These leaders focus on three non-delegable tasks — raising strategic ambition, rearchitecting the organization, and resetting culture. McKinsey’s report notes that only the CEO can make those calls without dilution.

Sal Khan at Khan Academy offers a case study in the visionary style. He began not with a roadmap but with questions about how students learn. Internal hackathons and experiments produced Khanmigo, an AI tutor now used globally. The approach blended learning science with technology instead of bolting AI onto existing processes.

Walmart’s leadership has centralized AI capabilities while pushing operating units closer to customers. Integrations with tools like Google’s Gemini aim at more personalized experiences. The message is consistent: AI succeeds when business leaders drive it, not when IT owns the project.

Yet risks remain. Founder mode can slide into micromanagement without judgment. Andy Grove’s concept of task-relevant maturity offers a guardrail — dive deep where it counts, step back where others have earned autonomy. Pure people managers who only run one-on-ones may not survive. The new model favors hybrids who still touch the work.

Some companies push further. Cursor’s small team at a multibillion-dollar valuation relies on spontaneous engineer projects rather than rigid planning. Tencent dismantled its standalone AI lab and fused the work into product teams. These moves treat AI as infrastructure that reshapes the org chart itself, not a feature added to old workflows.

Paul Graham warned that manager mode treats subtrees of the org chart as black boxes. In the AI era that approach looks obsolete. Real-time context from models, asynchronous updates, and agentic systems let one mind comprehend far more than before. The successful CEO will use those tools to stay close to customers, code, and culture — not to abdicate.

Critics call it romantic. Large enterprises cannot run like 50-person startups, they say. But the data increasingly disagrees. Smaller, founder-led or founder-minded organizations move faster on AI initiatives. They avoid the coordination tax that once justified bureaucracy. And they retain the taste that separates good products from great ones.

Chesky still reviews details personally. He co-manages a core group of about 50 people on hiring, promotion and performance. It is labor intensive. He calls it necessary. In an age when AI can generate most new code and surface insights once buried in middle management reports, that hands-on presence may separate companies that adapt from those that merely automate yesterday’s processes.

The debate Graham sparked in 2024 has evolved. What began as a critique of professional management has become a playbook for competing with AI-native upstarts. Boards now ask whether their CEO operates in founder mode. Investors differentiate between serious adopters and dabblers. The companies that thrive won’t be those with the most pilots. They will be those whose leaders refuse to step back at the moment technology makes closeness more powerful than ever.

And the rest? They may find that size, once a moat, has become a burden.