AI Is Changing the Rules of Business
What Manus, DeepSeek, and Kimi reveal about a new era of growth. Where companies optimize for capability and vision, not users and revenue.I’ve spent most of my career inside companies that were built on a familiar performance management model: set OKRs, then break them down into project KPIs.For…
What Manus, DeepSeek, and Kimi reveal about a new era of growth. Where companies optimize for capability and vision, not users and revenue.I’ve spent most of my career inside companies that were built on a familiar performance management model: set OKRs, then break them down into project KPIs.For internet products, those KPIs were typically DAU, new users, engagement, and revenue. Then cascade them into team-level metrics, and eventually into individual performance. Companies became incredibly good at executing predictable improvements.In my fifteen years of internet work experience, business growth largely meant one thing: execute better than everyone else, and in doing so, gain market share.But the more I learn from founders of companies like Kimi, DeepSeek, and Manus, the more I feel we’re witnessing something much deeper than another management trend. They’re optimizing for an entirely different objective. Once the objective changes, everything else changes with it.1. The KPI changes (Manus)One comment from Manus stayed with me. Instead of asking how to maximize today’s revenue, they ask two simple questions:What is expensive today but will inevitably become cheap?Tokens.What is cheap today but will become increasingly expensive?User acquisition — and more importantly, deep understanding of user needs.In other words, don’t optimize for the resource that technology will continuously commoditize. Optimize for the asset that compounds.Their North Star metric isn’t quarterly revenue. It’s understanding users deeply and pushing intelligence so far that word of mouth becomes the growth engine.Commercial success becomes the consequence.2. What’s deeper is strategy change (Deepseek)DeepSeek made me realize something even more interesting. Its strategy almost feels irrational by traditional business standards.During the internet era, everyone chased traffic and searched for business models that could monetize it. Last year, every company wanted a chatbot. This year, everyone wants enterprise AI revenue.DeepSeek intentionally walked away from both. Instead, it open-sourced its models. It targets a profit margin of around 6x, while many closed-source AI companies may pursue returns closer to 100x.The founder’s rationale is surprisingly simple. In the internet era, we were trained to pick up every sesame seed.More users + More traffic + More features + More revenue streamsEvery small gain mattered because market dominance was the prize.But AI may be fundamentally different. If intelligence eventually accounts for as much as 10% of global GDP, then the opportunity is so large that winning may no longer require monopolizing every adjacent market.Internet companies became experts at collecting sesame seeds: more users, more clicks, more engagement, more revenue.DeepSeek seems unusually willing to step over those sesame seeds if they distract from growing the watermelon.AI frontier labs seem willing to leave many of those behind if they distract from the bigger prize. For DeepSeek, that bigger prize is AGI; everything else is simply a by-product.Picking up sesame seeds while letting go of the watermelon. Image source: Doodlewash3. And organization changes to make things happen (Kimi at Moonshot AI)If your destination is different, the company you build has to be different too. This also explains why Kimi feels so different.While industry books and even the World’s AI Conference last week are still fixated on the original AI Scaling Law — more parameters + more data + more compute = a smarter model — a relatively small team like Moonshot AI is already rewriting the equation.With the release of models like Kimi K3 ranking among the top global performers, Kimi proves that the frontier has shifted. The game is no longer just about brute-forcing pre-training; it’s about scaling Reinforcement Learning. They aren’t just building bigger models; they are teaching intelligence how to reason, adapt, and self-correct in real time.The company has around 300 people, and most employees know one another. The five co-founders each work directly with roughly forty to fifty people, compressing decision-making chains as much as possible: “No titles to optimize for, no OKRs, no clocking in.”Employees are hired not simply for domain expertise, but for their taste, curiosity, learning velocity, and ability to generalize.The management system isn’t held together by metrics. It’s held together by vision.4. How top talents are retained is changedThis is probably the part that resonates with me the most. We’re entering an era where exceptional AI talent can realistically build their own company.The cost of creating software has collapsed. Starting a company is no longer the difficult part. So the real question becomes:Why would extraordinary people stay?When exceptional people can increasingly build their own companies, why would they stay in yours?Traditional organizations answer that question with compensation, promotions, and titles.The AI-native organizations that fascinate me increasingly answer it with vision.When people genuinely believe in the same vision, they stay.DeepSeek has expressed an even more fundamental belief: its most important asset is not today’s users, revenue, or market share. It is the stability of the team itself.As the founder put it, “Our greatest core interest is maintaining the stability of the team. This is our biggest core interest — perhaps even our only core interest. As long as I can keep the team stable, I believe we will eventually achieve AGI.”This is why DeepSeek employees rarely work overtime. They don’t waste energy on performative tasks just to satisfy short-term market hype. Protecting the team from burnout and keeping them focused on true breakthroughs is the strategy.Their management structure balances top-down execution with bottom-up autonomy:Top-Down (Formal Initiatives): When launching major milestones, like DeepSeek V4, the entire company aligns and allocates the work. However, these top-down mandates account for less than half of an employee’s time.Bottom-Up (Autonomous Exploration): Employees spend over 50% of their time pursuing self-directed research, exploring whatever problems they believe are most valuable, backed by the company’s full support and compute power. (Think about that for a second. That completely blows Google’s famous “20% time” rule out of the water!)For marketers, note that the user funnel has also changed.From the user’s perspective, this explains why the AI market feels so fluid today.The biggest change I’ve personally experienced is that switching costs have collapsed.AI models are improving at a speed we have rarely seen before, from pre-training scaling to reinforcement learning to agent-based systems. As a user, I no longer build the same kind of loyalty around a single chatbot or application.Instead, I find myself constantly switching between different AI tools. One chatbot may be better for reasoning. Another may be better for research. Another may be better integrated into a specific workflow. An AI knowledge hub may become more useful today, while an agent tool may become the preferred choice tomorrow.The deciding factor is no longer simply familiarity or habit. It is capability.This makes me reflect on one of the most fundamental assumptions from the internet era: that user acquisition is the ultimate growth engine. While user acquisition is still important, in an AI world where capabilities evolve so quickly, acquiring users may only be the beginning. The harder question becomes whether a company can continuously create enough value that users choose to stay as better alternatives emerge.Perhaps the AI era requires us to rethink what loyalty means. Users may no longer stay because they have invested years in a product. They may stay because the company continues to push the boundaries of what is possible.It’s an incredibly inspiring time to see how businesses evolve, and I can’t wait to see what comes next.This story is published on Generative AI. Connect with us on LinkedIn and follow Zeniteq to stay in the loop with the latest AI stories.Subscribe to our newsletter and YouTube channel to stay updated with the latest news and updates on generative AI. Let’s shape the future of AI together!AI Is Changing the Rules of Business was originally published in Generative AI on Medium, where people are continuing the conversation by highlighting and responding to this story.Source: Generative AI Pub — Published — Category: Image AI