For twenty years, the playbook was clear: take something people do manually, wrap it in a dashboard, charge per seat, and scale. SaaS ate the world because it made software accessible, predictable, and cheap to distribute. But a new thesis is emerging from the sharpest minds in venture capital, and it flips the entire model on its head. The next trillion-dollar company won't sell you a tool. It will do the work.

The Sequoia thesis

In March 2026, Sequoia Capital partner Julien Bek published an essay that has been circulating widely across Silicon Valley and beyond. The core argument is disarmingly simple: if you sell the tool, you're in a race against the model. Every time a frontier AI model improves, your product risks becoming a feature. But if you sell the work, every improvement in the model makes your service faster, cheaper, and harder to compete with. Bek uses accounting as the canonical example. A company might spend $10K a year on QuickBooks and $120K on an accountant to close the books. The next legendary company, he argues, will just close the books. This isn't a minor reframing. It's a fundamental shift in what software companies are selling. Instead of selling access to a tool that helps a professional do their job, you sell the completed outcome. The customer doesn't care about the interface, the dashboard, or the seat count. They care that the work gets done.

Intelligence versus judgement

The essay introduces a useful framework for thinking about which industries will be disrupted first. Bek draws a distinction between intelligence work and judgement work. Intelligence work follows rules. The rules might be complex, but they are rules. Translating a spec into code, debugging, medical coding, filling out insurance forms, drafting standard NDAs. These are tasks where the inputs and outputs are well-defined and the process is repeatable. Judgement work requires experience, taste, and instinct built on years of practice. Deciding which feature to build next, whether to take on technical debt, assessing cultural fit in a hire, or making a strategic recommendation to a CEO. AI has crossed the threshold where it can handle most intelligence work autonomously. The higher the ratio of intelligence to judgement in any field, the sooner autonomous AI systems will win. Software engineering got there first, which is why over half of all AI tool usage today is in code. But it's coming to every profession.

Copilots and autopilots

This intelligence-judgement spectrum maps neatly onto two product archetypes that are now competing for the same markets. Copilots sell the tool. They sit alongside professionals and make them more productive. Harvey helps lawyers draft documents faster. Rogo helps investment bankers analyze data more efficiently. The professional remains the customer, the decision-maker, and the one responsible for the output. Autopilots sell the work. They go directly to the company that needs the outcome. Instead of selling to the law firm, an autopilot sells to the company that needs the NDA drafted. Instead of selling to the insurance broker, it sells to the CFO who needs coverage. The customer is buying the completed outcome, not a tool to help someone produce it. The distinction matters enormously for market sizing. The work budget in any profession dwarfs the tool budget. For every dollar spent on software, roughly six are spent on services. Autopilots capture the work budget from day one.

The outsourcing wedge

Bek's essay lays out a compelling go-to-market playbook for autopilot companies: start where outsourcing already exists. If a task is already outsourced, it tells you three critical things. First, the company has accepted that this work can be done externally. Second, there's an existing budget line that can be substituted cleanly. Third, the buyer is already purchasing an outcome, not a tool. Replacing an outsourcing contract with an AI-native service provider is a vendor swap. Replacing headcount is a reorg. One is a procurement decision. The other is an organizational transformation. The outsourced, intelligence-heavy task is the wedge. The insourced, judgement-heavy work is the long-term opportunity.

The opportunity map

Looking at services verticals through this lens reveals staggering market sizes:

These numbers make the entire SaaS market look modest by comparison.

The broader shift

Sequoia's thesis doesn't exist in a vacuum. It's part of a larger conversation happening across the industry. Deloitte predicts that up to half of organizations will put more than 50% of their digital transformation budgets toward AI automation in 2026. Gartner estimates that by 2030, 35% of point-product SaaS tools will be replaced by AI agents or absorbed within larger agent ecosystems. IDC describes the AI agent becoming "a new enterprise SKU," purchased via marketplaces and powered by modular backend capabilities rather than monolithic SaaS platforms. Thoughtworks has articulated this as the move from Software-as-a-Service (SaaS) to Service-as-Software (SaS). The terminology flip is subtle but significant. Traditional SaaS sells tools that enable humans to solve problems. Service-as-Software sells outcomes. It's a new class of system that doesn't just enable work but automates the reasoning process itself. This reframing has real consequences for how companies are built and valued. When you sell outcomes, pricing shifts from seats and subscriptions to usage and results. The sales model changes. The competitive moat changes. Everything changes.

The convergence ahead

Perhaps the most interesting part of the Sequoia thesis is the prediction that copilots and autopilots will eventually converge. Today's judgement will become tomorrow's intelligence. As AI systems accumulate proprietary data about what good judgement looks like in their domain, the frontier between intelligence and judgement will keep shifting. A copilot that starts by helping a professional make decisions is steadily learning what those decisions look like. An autopilot that starts by handling straightforward intelligence work is steadily expanding into adjacent territory that requires more nuance. The companies that start as autopilots have a structural advantage. They're capturing the work budget, building proprietary data about outcomes, and compounding that data with every task they complete. The copilot companies face an innovator's dilemma: selling the work means cutting their own customers out of doing it.

What this means in practice

If this thesis plays out, the implications are significant for almost everyone in tech. For founders, it means the most defensible position isn't the best tool but the most reliable outcome. Every improvement in the underlying AI model should make your service better, not threaten your existence. If a new model release makes you nervous, you're probably selling the tool. For enterprises, it means rethinking procurement. Instead of buying software licenses and hiring people to use them, you may increasingly be buying completed work products. The budget line looks more like outsourcing than IT spend. For professionals, it means the premium shifts from execution to judgement. The accountant who can close the books is less valuable than the CFO who knows which books to keep. The lawyer who can draft an NDA is less valuable than the general counsel who knows which deals to walk away from. For SaaS incumbents, it's a genuine existential question. Salesforce, SAP, ServiceNow, and Workday are all building AI agents into their platforms. But the real threat isn't from each other. It's from companies that skip the platform entirely and just do the work.

The human question

There's a question that hovers beneath all of this: what happens to the people who currently do this work? The optimistic view is that AI handles the intelligence work while humans focus on the judgement that matters. The accountant stops manually reconciling entries and starts advising on strategy. The recruiter stops screening resumes and starts building relationships. The junior lawyer stops drafting boilerplate and starts learning to think like a partner. The realistic view is that the transition will be uneven and sometimes painful. Not every intelligence worker will seamlessly become a judgement worker. The accounting profession is already losing 340,000 people not because of AI but because the work isn't attractive enough to recruit replacements. In some fields, AI won't displace people so much as fill gaps that already exist. What seems clear is that the change is coming whether we're ready for it or not. The economics are too compelling to ignore. When you can deliver the same outcome at a fraction of the cost with higher consistency, the market will move.

The bottom line

The SaaS era gave us a powerful insight: software is better delivered as a service than as a product. The emerging era is taking that insight one step further. The service itself is better delivered as software. We're not at the end of SaaS. We're at the beginning of something that subsumes it. The companies that win won't be the ones with the best dashboards or the most features. They'll be the ones that quietly, reliably, do the work.

References