From Managing People to Managing AI

In a recent talk, Kai-Fu Lee posed a critical question: What will determine competitive advantage for companies in the future? The underlying trend he identified is profound: most work will shift from being organized around people to being organized around AI agents—what we might call “super-employees.” This mirrors the core logic behind our course AI Leadership.

The managerial mindset must evolve accordingly.

In the AI era, resources expand beyond people to include AI tools—but the fundamental leadership capabilities remain unchanged: setting clear goals, deeply understanding the strengths and limitations of your collaborators (human or AI), and designing effective collaboration workflows.

What does change is the rhythm and structure of teams: with AI, pace accelerates, boundaries blur, and teams shrink—becoming flatter and more autonomous.

Key shifts in management perspective and capability:

Define success with precision: Just as with people, managing AI starts with crystal-clear definitions of success. “Improve retention” is too vague; “increase 30-day revisit rate by 5%” gives AI actionable direction—and better outputs.
Resources now include AI tools: Your toolkit isn’t just headcount—it’s a portfolio of AI agents, each with distinct capabilities. Whether used proactively or reactively, they’re part of your productive capacity.
“Human + AI” remains the dominant operating model: Almost all high-value scenarios still require tight, intentional human–AI collaboration—not replacement.
Flatter, role-blended teams: AI enables individuals to grow multidimensionally. Small teams can own end-to-end outcomes, reducing dependency on traditional roles—e.g., many functions no longer require dedicated product managers.
De-emphasize job titles; emphasize outcome ownership: The future favors generalists who “get things done”—which demands rapid learning agility from everyone.
Prioritize data literacy: Use data not just to report, but to diagnose root causes and spot emerging opportunities.
Accelerate feedback loops: Shift from monthly reviews to weekly—or even daily—iterations. Aim for continuous 1% improvement.
Use AI to compress scale and deepen accountability: Smaller, end-to-end teams mean faster decisions, tighter ownership, and higher velocity.

Sora 2’s Ambition

I spent time today deeply exploring Sora 2—and the results were stunning. This iteration marks a major leap:

Higher physical fidelity: It renders complex physics—buoyancy, gymnastics—with remarkable accuracy. More impressively, it simulates physically plausible failures, not just ideal outcomes.
True multimodal integration: Seamless fusion of vision and audio—generating multilingual dialogue where lip movement, tone, and timing align perfectly for immersive storytelling.
A TikTok-like AI video social platform: Launched alongside Sora 2, its interface dramatically lowers creative barriers. I hit my 100-video quota within hours—by comparison, TikTok’s creation flow still feels comparatively “high-friction.”
Cameo feature: One-time identity verification lets users insert their own (or friends’) likeness and voice into AI-generated videos—personalization at unprecedented speed and scale.
This reveals Sora 2’s real ambition: Not just a tool, but an attention ecosystem. Its goal is to capture, retain, and monetize user attention—not through features alone, but through network effects and behavioral stickiness.
For social content, convenience + believability + unpredictability beats technical perfection: A lightly flawed but emotionally resonant, AI-generated story often outperforms a technically flawless but sterile video.
AI-generated “junk content” is a real threat—and OpenAI acknowledges it.
Sora 2 may trigger the creative industry’s inflection point: Rather than compete solely on model benchmarks in enterprise or pro markets, OpenAI is betting on C-side adoption and viral growth—building defensibility via usage, not specs.
The hyper-personalized entertainment era is imminent: Media will shift toward real-time, on-demand generation—e.g., films dynamically casting actors whose appearance, voice, and mannerisms match your aesthetic preferences.
The barrier to creating photorealistic synthetic video has collapsed, fueling urgent demand for next-gen detection and digital provenance tools—like C2PA metadata standards.
For individuals: Shift learning focus—from mastering specific software to cultivating aesthetic judgment, narrative design, and directing AI systems. Your creative vision—not technical execution—will be your differentiator.
Critical thinking becomes non-negotiable: With synthetic content flooding feeds, adopt a default stance: assume all digital media is synthetic until verified. Before sharing, habitually check sources and look for provenance markers.

The Truth About Attention

For individuals, productivity hinges on stable, sustained, and precise attention—not busyness.
Real productivity isn’t about doing more—it’s about consistently directing attention toward what compounds: output (creation), learning, and health—and protecting that flow from interruption.
A common misconception: attention isn’t just “seeing.” It operates across three layers—perception → attention → deep focus—and only the last yields meaningful increment.
Why does dopamine make us busy but unproductive? Because dopamine fuels the “wanting” circuit—not satisfaction, but craving—reinforced by novelty and surprise. The antidote? Design dopamine rewards into valuable output, not distraction.
Apps don’t just “grab” attention—they dilute your personal productive capacity.
Attention is the prerequisite resource for all higher-order work: without sustained attention, there’s no learning, no insight, no design, no writing—no closed loop.
Attention is scarce, non-storable, and zero-sum: interrupt it, and it resets to zero. In this sense, attention is more valuable than time.
The foundation of attention training is values alignment. Example daily priority order: sleep > movement > rest > learning > work; interpersonal priority: self > family > friends > colleagues > world. Most people invert both.
Attention determines “energy density” per unit time. One hour of deep focus delivers orders-of-magnitude more knowledge gain and output quality than the same hour fragmented across tasks.
Crucially: structured emptiness matters more than fullness. Masters deliberately schedule “blank space”—not as idle time, but as cognitive reset and incubation.
Attention-switching cost is wildly underestimated. So resist packing your day with “important” items—fewer high-stakes commitments enable deeper focus.
For managers: Productivity = attention orchestration. Where your team’s attention flows—shaped by goals, plans, methods, incentives, and training—defines your collective output.
Sustaining attention on the right things—and holding it long enough—is what unlocks multiplicative returns from tools, collaboration, and intelligence.
Yet identifying “the right thing” is among the hardest challenges—rooted in your cognition, values, and judgment.
When attention becomes your primary capital, productivity follows—not the other way around.
For me, the ultimate aim of attention management isn’t efficiency—it’s freedom of choice.

The Four Levels of Sales

Sales operates across four ascending levels—from lowest to highest:
Order → Customer → Market → Team

Rephrased as guardrails:
→ Don’t sacrifice a customer to close one order.
→ Don’t sacrifice a market to serve one customer.
→ Don’t sacrifice your team to win one market.

“Sacrifice” here means compromising a higher-order objective for short-term gain.

The core practice: Always operate from the level above your current role.

Examples:

As an individual seller: Don’t optimize for the order—optimize for long-term customer trust and lifetime value.
As a sales leader: Don’t fixate on customers—focus on capturing sustainable market share and building scalable motion.
As head of sales: Don’t stop at market share—invest relentlessly in building a team that wins repeatedly, even when you’re not in the room.

Tools, AI, and the Power of Slowness

I was mapping out a business model—and casually asked AI to build two small internal tools. They worked surprisingly well.

Later, chatting with a friend on WeChat, we marveled at how effortlessly AI writes code today.

We joked: If only we’d had this a decade ago—we’d be unstoppable.

Then I paused. That “if only” is likely an illusion of time.

We assume tools define ceilings. But the deeper truth is this: Tools don’t empower people—the prepared person empowers the tool.

Readiness depends on cognition, existing resources, attention allocation, and sheer commitment.

Ten years ago, even with AI, if our mental models hadn’t evolved—if our foundations were weak, if we didn’t prioritize the work—we’d have missed the opportunity anyway.

Growth takes patience. You don’t rush to the tool—you grow into it.

This season of reflection reaffirmed something quiet but vital: Patience isn’t just virtue. It’s submission to the natural rhythm of development.

And those so-called “foundational skills”—learning, execution, thinking, communication, physical stamina, emotional regulation, focus—they’re not background noise. They’re the bedrock. More essential than we ever imagined.