A few weeks ago, Julien Bek, a partner at Sequoia Capital, one of Dawn Liphardt Valley’s most influential venture capital funds, posted an analysis that has been circulating among investors and startup founders alike.
The central argument is that the next big wave of value creation from AI won’t come from software but from services. Its title: Services: The New Software.
Selling the work, not the tool
The distinction may seem subtle. It is, in fact, radical. Until now, most AI startups have positioned themselves as “copilots”: they sell a tool to professionals (lawyers, accountants, engineers) who remain in charge of their work. Harvey sells to law firms. Rogo sells to investment banks. The model is AI-enhanced SaaS.
But a new generation of startups chooses, from day one, to be “autopilots.” They no longer sell the tool to the professional; they sell the outcome directly to the end client. Crosby does not sell software to law firms; it drafts the Non-Disclosure Agreement for companies that need them. WithCoverage does not sell to a broker, but directly to a chief financial officer.
In short, the autopilots capture the work budget from the outset.
Intelligence versus judgment: the boundary to watch
To understand which sectors are ripe for this disruption, Sequoia offers a simple framework: separate tasks of intelligence — translating a specification into code, classifying a medical document, filling out a regulatory form — from tasks of judgment, which rely on experience, discernment, and intuition honed by years of practice.
Software engineering has been the first profession to cross the threshold where AI can perform the bulk of the intelligent work autonomously, leaving judgment to humans. It now accounts for more than half of AI tool usage by occupation, while all other professions remain below 10%. But according to Sequoia, this movement is poised to touch every profession.
The list of markets under threat
Julien Bek sketches a map of priority sectors, categorized by their intelligence/judgment ratio and their degree of outsourcing already in place. The reasoning is that if a task is already outsourced, it means the company has accepted that this work can be done externally, that there exists a substitutable budget line, and that the buyer already purchases a result.
Replacing a contractor with an AI autopilot is merely a change of supplier. Replacing internal employees is a reorganization.
Among the markets identified:
- Insurance brokerage (US$140–200 billion) : standard commercial lines are highly standardized, the broker’s added value reducing to comparing insurers’ offers and filling out forms. Pure intelligence work.
- Accounting and auditing (US$50–80 billion outsourced in the United States alone) : the country has lost about 340,000 accountants in five years, 75% of CPAs are nearing retirement, and the pipeline of new entrants is drying up. This structural shortage pushes firms to adopt AI faster than in almost any other sector.
- Revenue-cycle management in healthcare (US$50–80 billion) : medical billing consists of translating clinical notes into roughly 70,000 standardized ICD-10 codes. The rules are complex, but they are rules. So intelligence work, not judgment.
- Managed IT services (over US$100 billion) : every SME outsources its IT: monitoring, provisioning, and alert management. Pure intelligence work that repeats identically across thousands of similar environments.
- Recruiting and staffing (over US$200 billion) : the top of the funnel — screening, matching, outreach — is pure intelligence; assessing a candidate’s cultural fit is a matter of judgment.
- Management consulting (US$300–400 billion) : a vast market, but most of the work rests on judgment. The question is whether AI can extract the intelligent components (data collection, benchmarking) to automate them, leaving strategic recommendations to humans.
The innovator’s dilemma for copilots
In 2025, the fastest-growing AI companies were copilots. In 2026, many will attempt to become autopilots. They have the advantage of product and customer knowledge. But they face the classic innovator’s dilemma: selling the work risks disconnecting their own clients from doing the work themselves. It is precisely this hesitation that opens the door to new entrants positioned from the start as autopilots.
For Sequoia, the figure that sums up everything is this: for every dollar spent on software, six are spent on services. It’s this market, six times larger than the one the SaaS revolution has already transformed, that AI is poised to attack.
