What does AI for startups actually mean?

It means building your company for a revolution in computation, not just adding AI features. At Sequoia's AI Ascent, the partners framed AI as different from the internet, cloud, and mobile: those changed how information is distributed, while AI changes how it is processed. For founders the practical test of where we are is simple, in their words: if you can dispatch an agent to do a job and it recovers from failure and persists until the job is done, that already feels like AGI. So AI for startups means designing the product, the team, and the workflows around agents that do real work, not bolting a chatbot onto an old process.

How do startups build a moat in AI?

By wrapping around the customer, not the capability. Sequoia's MAD framework (moats, affordance, diffusion) starts from a counterintuitive point: in a revolution of computation the technology you build on changes daily, but your customers change slowly, so the durable advantage is how tightly you wrap yourself around them. a16z's Benedict Evans makes the same point about where value accrues: not in the foundation models themselves, but further up the stack in the applications built around real customer problems. The moat is whether you deliver compounding advantages to a specific customer.

How should a startup adopt AI safely?

Aggressively, but with guardrails first: scope what agents can access, keep secrets out of prompts, log what they do, and keep a human approving anything irreversible. The mindset from operators like Brex's Pedro Franceschi is to focus on what only you can do and hand the rest to agents, and Nvidia's Jensen Huang puts it as you will not lose your job to an AI, but to someone who uses AI. The reason guardrails come first is that the same autonomy that makes agents powerful makes them a new attack surface, including autonomous AI-powered attackers.

AI for startups: the founder's guide

Q: What is the biggest AI opportunity for founders right now?

Services. Sequoia's partners call it the first wave that is both software and services, and Sonya Huang's line is that services is the new software. Some of the biggest companies of the next decade will not be software businesses at all; they will be law firms, insurance carriers, and accounting practices rebuilt from scratch with AI doing most of the work. Legal services in the US alone is a 400 billion dollar market, the same size as all of software. The opening is the diffusion gap: capabilities are being created far faster than the average enterprise adopts them, and that gap is where application founders win.

Q: Do AI-native startups still need people?

Yes, but the work changes. The durable human trait is adaptability, and as the Sequoia keynote closed, AI can do the work but only the human connection gives a reason to care. YC's framing is that AI changes the unit of production to humans in concert with agents, with memory, evals, and a customer loop. Teams get smaller and a handful of people ship what used to take hundreds, but a founder still owns the vision, the taste, and the calls that matter.

The biggest opportunity in software is no longer software. At Sequoia's AI Ascent 2026, the firm's partners opened with a reframe every founder should sit with: AI is not another revolution in how information is distributed, the way the internet, the cloud, and mobile were. It is a revolution in how information is processed. "AI is a revolution in computation," as Pat Grady put it. That sounds like semantics. It is not. It changes the shape of the wave, and for a founder it changes what AI for startups even means.

This is a guide to AI for startups: what the shift means, where it is heading, and how founders should build for it. It draws on the people building the frontier, the Sequoia keynote, Nvidia's Jensen Huang, OpenAI's Greg Brockman, a16z, YC's Garry Tan and Pete Koomen, and links down to the practical playbooks on this site.

AI is a revolution in computation, and this is AGI

By a test founders can act on, AGI has functionally arrived: if you can dispatch an agent to do a job and it can recover from failure and persist until the job is done, that feels pretty much like AGI. The Sequoia partners are careful, they are venture capitalists, not about to propose a technical definition, but commercially, by that test, it is here. The last few years gave us faster horses, applications that made you 10 to 40 percent more productive. Now we are getting cars, applications that make you 10 to 40 times more productive and change the nature of the work itself.

The prize is bigger than software has ever been, because for the first time a wave is both software and services. Legal services in the US alone is a 400 billion dollar market, the same size as all of software, and that is one vertical in one country. On the BG2 podcast, Gavin Baker and the panel size the compute underneath it, an AI capex cycle running into the hundreds of billions in revenue within a couple of years. The exact figure is not the point. The point is that the technology foundation moves every day, "no lead is safe," and when capabilities arrive this fast, anybody can win.

The cognitive revolution: where this is heading

More than 99 percent of physical work on Earth is now done by machines, and cognition is on the same curve. Sequoia's Konstantine Buhler draws the parallel: for most of history almost all physical work was done by muscle, human or animal, until the Industrial Revolution changed it. He projects that in the near future 99.9 percent of cognition on the planet will be done by machines, and that the cognitive revolution will be a lot like the Industrial Revolution, just much bigger and much faster.

It is easy to read that as mystical. Nvidia's Jensen Huang, in a separate Sequoia talk, pushes back: AI is not mysterious. It is computer and software that can be understood and improved. That is the founder mindset to keep. Three of Buhler's images are worth holding onto:

Skills become disposable. Aluminum was once so precious it capped the Washington Monument and sat on display at Tiffany's; within decades, electrolysis made it cheap enough to wrap a sandwich and throw away. Aluminum is intelligence, and AI is the electrolysis. PhD-level skills that took decades to earn become things you invoke and discard.
Alien design. When machines do the cognition, the outputs stop being intuitive to us, the way an AI-evolved antenna looks nothing like one a human would draw. Founders have to stay open to results that do not look like what a person would build.
New sciences. We are in the tinkering phase, the way engineers tinkered with combustion for a century before Carnot formalized thermodynamics. A science of AI that fundamental is coming, and it will be taught in high schools.

The disruption is not abstract. Anthropic's Dario Amodei has warned that AI could eliminate half of all entry-level white-collar jobs within one to five years. The flip side, as guests on Harry Stebbings' 20VC argue, is that AI lets a handful of people build companies that reach billion-dollar valuations with a fraction of the headcount. Both are true at once, which is exactly why founders should be building.

Agents are the unlock

If 2022 to 2024 was chat and 2024 to 2025 was reasoning, 2026 is agents. Sequoia's Sonya Huang defines an agent simply: a system that perceives its environment, chooses actions, and progresses autonomously toward a goal. Three things had to mature, and all three crossed the line around the turn of the year: the models (the brain, now able to sustain a complex task for hours, not minutes), the tools (the arms and legs, decades of software built for humans, now wielded by agents), and the harness (the persistence to stay on task).

The founder consequence is leverage. Huang's line is that whatever you could imagine building over the next 100 years is now possible in 100 days. On stage at Sequoia, Andrej Karpathy drew the distinction to internalize: vibe coding raises the floor so anyone can build, while agentic engineering keeps the quality bar of professional software as you go faster. In a Stanford CS 153 lecture, Garry Tan put a number on it: people using AI coding agents are 10 to 100 times more productive than engineers using Cursor and chat. The practical version, running agents like a team, is in how to build software with AI agents.

What "AI for startups" means: build an AI-native company

AI for startups is not a feature you add. It is the shape of the company. Garry Tan and Diana Hu describe the AI-native company as a closed loop: agents read the artifacts the company produces and feed that context back into the work, so the company self-heals. The org redraws around a few clear roles plus a new one, the AI-native founder who lives at the edge of the tooling. Brex CEO Pedro Franceschi, whom YC calls the most AI-pilled CEO they know, gives the leadership version: the CEO has to be the chief AI officer, not someone who delegates it to an engineering or product team. The plumbing underneath, a shared context store and an internal tool registry, is what YC's Pete Koomen calls internal AI infrastructure for startups: using AI as the building layer for everything, not as a copilot bolted on the side.

For where founders are applying this today, the YC startups using AI by function, from sales to customer support to coding, are the live examples to study.

Services is the new software

The line Sonya Huang repeats, borrowed from Pat Grady, is that services is the new software. Some of the biggest companies of the next decade will not be software businesses at all; they will be services companies, law firms, insurance carriers, accounting practices, rebuilt from scratch with AI doing most of the work. We broke that model down in how to build a service-as-software company: you sell the outcome, not the tool, the process is the product, and the human is the interface that lets the work scale non-linearly. YC's pattern for the founders who win here is three attributes: domain fluency, model fluency, and operational rigor. Getting it to market is its own discipline: the sales motion you pick, top-down or bottom-up, shapes your pricing, hiring, and roadmap.

How to win on top of the models: the MAD playbook

If you build on top of the labs, Sequoia's advice is to get MAD: moats, affordance, and diffusion.

Moats. In a revolution of computation the capabilities you build on change daily, but your customers change slowly. So wrap yourself around the customer. As Greylock frames it, the moat is whether you are building the solution that delivers compounding advantages to that customer.
Affordance. A tool with affordance does not need to be explained. Claude Code is powerful, but hand a terminal to the average Fortune 500 employee and watch how far they get. The opportunity is to build the path of least resistance to the outcome your specific customer needs.
Diffusion gap. Capabilities are created far faster than they diffuse into the market. Every day the labs move faster than the average enterprise, that gap, and your opportunity, gets bigger.

It lines up with where a16z's Benedict Evans says value accrues: not in the foundation models themselves, but further up the stack, in the applications wrapped around real customer problems. The model is becoming the commodity; the company is everything you build around it, and the teams that win build so each new model makes them stronger, not obsolete.

Adopt aggressively, but stay safe

Adopt AI aggressively, then wrap it in guardrails. Brex's Pedro Franceschi advises focusing on the things only you can do and handing the rest to agents; Nvidia's Jensen Huang's version is that you will not lose your job to an AI, but you might lose it to someone who uses AI. The aggressive-adoption playbook is in why a startup CEO went all in on AI.

As a CISSP, here is the part I will not let founders skip. The same autonomy that makes agents powerful makes them a new attack surface. Sequoia's session with XBOW was about exactly this: autonomous, AI-powered attackers, and the warning that we need every possible defense against them. Sonya Huang's "dark factories," taking human review out of the loop entirely, are possible only with good enough guardrails and good enough engineering. Build the guardrails first: scope what agents can access, keep secrets out of prompts, log what they do, and keep a human on anything irreversible. Anthropic's way of shipping fast without cutting safety is a useful model, broken down in Anthropic: lessons for founders. And the frontier OpenAI's Greg Brockman named at Sequoia is that human attention, not raw model capability, is becoming the bottleneck, which is a design problem founders get to solve.

What stays human

Konstantine Buhler closed the keynote with a line from Protagoras, 2,500 years old: man is the measure of all things. Nothing has value in a vacuum, not aluminum, not art, not intelligence; it has value because of the human experience around it. His conclusion, and ours: AI can do the work, and AI will do the work, but only the human connection can give you a reason to care. The durable human trait is adaptability. Garry Tan's version is that AI changes the unit of production to humans in concert with agents, with memory, evals, and a customer loop. The company is still yours to point.

What to do this week

Pick the one workflow where AI changes the math 10x, not 10 percent, and rebuild it agent-first.
Decide what only you, the founder, can do. Hand the rest to agents and own the spec and the review.
Map your customer, not the capabilities. Your moat is how tightly you wrap around them; the tech underneath will change.
Stand up the guardrails before you scale the agents: access scope, secrets, logging, and a human on every irreversible call.
Find your diffusion gap, the place where the labs are already capable but your market has not adopted. That is the opening.

The cars have arrived, and no lead is safe. That cuts both ways: the incumbents' leads are not safe either. What a time to build.

If you want this as a system rather than a stack of talks, that is what we teach in the AI Operating System for Startups.

Sources

This is AGI: Sequoia AI Ascent 2026 keynote, the primary source for this guide, with Sequoia partners Pat Grady, Sonya Huang, and Konstantine Buhler.
Nvidia's Jensen Huang at Sequoia: "AI is Not Mysterious" and "you'll lose a job to someone who uses AI".
Sequoia talks: Andrej Karpathy on agentic engineering, OpenAI's Greg Brockman on human attention, and XBOW on autonomous AI attackers.
Garry Tan and Diana Hu, Stanford CS 153 on the AI-native company; Inside YC's AI Playbook with Pete Koomen; The Most AI-Pilled CEO We Know with Pedro Franceschi; How to Build an AI-Native Services Company.
a16z: The Economics of AI Usage with Benedict Evans and The New Rule for Picking AI Winners; Greylock on moats and AI agents; BG2 with Gavin Baker on the AI capex cycle; 20VC on AI and jobs; and Bloomberg's Inside Anthropic, where Dario Amodei warns AI could remove half of entry-level white-collar jobs.