Services as Software: Why the Biggest Investment Thesis of 2026 Is About Labor, Not Tools

For thirty years, software companies sold tools. In 2026, the most valuable ones are starting to sell the work itself.

That single shift is the most important reframing in venture capital right now, and it changes how investors should think about market size, defensibility, and where the next generation of enormous companies will come from. It has a name that has spread rapidly through the investor community: services as software. And the thesis behind it is bold enough that it is worth stating in its strongest form.

In March 2026, Sequoia Capital published an essay arguing that the next trillion-dollar company will not sell software tools. It will sell the actual work, powered by AI, delivered as a seamless service, and disguised as a traditional services business. Foundation Capital has gone further, calling services as software the default frame for B2B investing in 2026.

If you are an investor and this thesis is not yet shaping how you evaluate B2B companies, this post explains why it should be, where the opportunity actually is, and what separates the companies that will capture it from the ones that will not.

The Distinction That Determines Everything

The entire thesis rests on one distinction that every investor needs to internalize. It is the difference between a copilot and an autopilot.

A copilot sells a tool to a professional who remains responsible for the output. An autopilot sells the output directly. The example that has become standard in the venture community makes it concrete: Harvey sells legal AI to law firms, so the lawyer still does the work with better tools. Crosby sells the reviewed contract to the company that needs it reviewed, so the work itself is the product.

That distinction is not a matter of product design. It determines which budget the company competes for, and that determines everything about its potential scale.

Copilots fight for software budgets. Autopilots fight for labor budgets. And in most professions, labor budgets are orders of magnitude larger than software budgets. When a company stops selling a tool to a professional and starts selling the completed work that professional used to produce, it stops competing with other software vendors and starts competing with employees, contractors, and outsourcing firms.

That is the whole game. The unit of sale moves from the seat to the outcome, and the moment it does, the addressable market changes by an order of magnitude.

The Math That Re-Rates the Entire Sector

The reason this thesis has captured so much investor attention is the market sizing implication, and it is genuinely dramatic.

For decades, the services economy was unreachable for software companies. Professional services, healthcare administration, compliance, consulting, business process outsourcing, and support are collectively measured in the high teens to mid-twenties of trillions of dollars annually. Software could never touch that pool, because services require cognition, judgment, iteration, and contextual awareness. Software could store and retrieve, but it could not decide and execute.

Generative AI changed that constraint. And the market math follows directly. One widely shared formulation expresses the new addressable market as the traditional software market plus the automatable portion of the labor budget. If even 25 to 40% of high-wage cognitive tasks become substitutable over the next decade, the addressable market for software shifts from hundreds of billions of dollars to multiple trillions.

This is not incremental feature expansion. It is a structural re-rating of the entire sector.

The crucial nuance, and the reason this is not science fiction, is that it does not require replacing entire departments or eliminating every human. It requires shifting the marginal dollar of spend. Enterprises do not need to fire their teams to generate enormous budget migration. They only need to automate meaningful slices of repetitive, high-volume cognitive work. When a CFO evaluates payroll against software for a specific, repetitive workflow, the substitution calculus is changing quickly, and each shifted workflow moves spending from the labor line to the software line.

For an investor, the implication is direct. The companies selling outcomes are addressing a market that is ten to twenty times larger than the one traditional SaaS companies compete in. That difference in ceiling is the difference between a good outcome and a generational one.

Why This Is a Threat to Incumbents and an Opening for Startups

The most interesting strategic dynamic in this thesis is that the companies best positioned to capture it are structurally reluctant to try. That reluctance is the entire opportunity for early-stage investors.

In 2025, the fastest-growing AI companies were copilots. In 2026, many of them will try to become autopilots. They have the product and the customer knowledge to do it. But they also face a genuine innovator's dilemma, because selling the work means cutting their own customers out of doing that work.

Consider the position of a company selling AI tools to a professional services firm. Its customers are the professionals whose labor the autopilot model would replace. To move from copilot to autopilot, that company would have to build a product that makes its own paying customers obsolete. That is an extraordinarily difficult thing for an incumbent to do, and most will hesitate at exactly the moment decisiveness is required.

That hesitation is the opening for pure-play autopilots built from scratch to sell outcomes. A startup with no existing customer relationships to protect can go directly after the labor budget without the conflict that paralyzes the incumbent. The examples are already emerging: companies selling the drafted contract to the company that needs it rather than to the outside counsel, or selling the completed insurance placement to the CFO rather than to the broker. The customer buys the outcome directly, and the intermediary is disintermediated.

The established SaaS giants are inadvertently accelerating this. As the largest enterprise software companies push their own agentic offerings, they are legitimizing the category and making enterprises more willing to bet on faster-moving startups. The incumbents are educating the market that startups will then capture.

For investors, this is a familiar and attractive pattern. A large incumbent is structurally constrained from pursuing an enormous opportunity, and nimble startups without that constraint can move directly into it. That is where outsized early-stage returns have always come from.

The Last Mile Problem Every Investor Must Underwrite

A disciplined investor cannot evaluate this thesis without understanding its central risk, which is the gap between a demo and a deployment.

The services as software model lives or dies on reliability, and the reliability bar is brutal. As Foundation Capital put it, you can get to 80% of the way there with 20% of the effort, which is enough to close a pilot. But production demands 99% or more, and that last stretch can take 100 times more work.

This matters more for autopilots than for almost any other kind of company, because of what they are selling. When you sell a tool, an occasional error is the user's problem to catch, because the human remains responsible for the output. When you sell the completed work, an error is your liability. An autopilot that drafts a contract with a mistake, or completes a compliance task incorrectly, has not shipped a flawed feature. It has delivered a defective work product that the customer trusted it to get right.

This is why the last mile problem is the central diligence question for any services as software investment. The reason a raw model cannot simply plug in and deliver is that real enterprises are messy ecosystems of legacy systems, fragmented data, security requirements, and governance complexity. Getting from a model that works in a demo to a system that reliably delivers completed work inside that messy reality is where most of these companies will succeed or fail.

The practical diligence questions follow directly. What is the company's accuracy in real production environments, not in controlled demonstrations? How does the system handle the cases it gets wrong, and is that failure contained and recoverable? What evidence exists that the company has crossed from pilot reliability to production reliability in its specific vertical? A founder who has genuinely solved the last mile in one workflow has something far more valuable than a founder with an impressive demo across many.

How AI Pricing Follows the Outcome

The shift from selling tools to selling work forces a corresponding shift in how these companies charge, and investors need to understand it because it changes how revenue quality should be assessed.

The traditional SaaS model prices by the seat. That model breaks in the services as software world, and it breaks in a specific and revealing way. If your AI reduces the number of people a customer needs, then a seat-based model means your revenue shrinks at exactly the moment your product delivers the most value. The better your autopilot works, the fewer seats your customer needs, the less they pay you. That is a broken incentive.

The model that fits the thesis prices by the work performed and the outcome delivered, rather than by the number of users with access. Value is tied to the work the AI does and the results it can prove. For an investor, this is actually a positive signal when it appears, because outcome-based pricing aligns the company's revenue with the value it genuinely creates and captures a share of the large labor budget rather than the small software budget.

When evaluating a services as software company, the pricing model is a tell. A company still selling seats for what is fundamentally outcome-delivering software has not fully made the transition and is leaving both value and defensibility on the table. A company pricing by outcomes is positioned to grow its revenue in direct proportion to the value it delivers.

Where the Global Opportunity Is Hiding

Here is the dimension of this thesis that most of the venture commentary, written largely from within the US market, tends to miss entirely.

The services economy is not concentrated in Silicon Valley. It is everywhere that work gets done. Healthcare administration, compliance, customer support, back-office processing, and professional services exist in every economy on earth, and in many markets they represent an even larger share of economic activity than they do in the US.

This means the services as software opportunity is fundamentally global, and in some respects the opportunity outside the US is larger. Labor-intensive service industries in emerging markets are enormous, often less digitized, and frequently underserved by the US-focused autopilot startups chasing US labor budgets. A founder building an autopilot for a specific high-volume cognitive workflow in an emerging market is going after a labor budget that no Silicon Valley company is prioritizing, in a market where the cost structure and the competitive dynamics both favor them.

There is a second, subtler advantage. Services as software companies, by their nature, generate real external revenue from real customers who pay for completed work. In a market where investors are rightly worried about circular financing and companies whose revenue loops back to their own investors, an autopilot selling genuine work product to genuine external customers is the definition of a fundamentally sound business. These companies have real revenue by construction, because their entire model is getting paid for work delivered.

For investors, the combination is compelling: a thesis with a ten to twenty times larger addressable market, companies with inherently real external revenue, and a global opportunity set that the US-concentrated venture market is largely ignoring outside its own borders.

How SeedScope Positions You for the Services as Software Wave

The services as software opportunity is global, generates real external revenue, and is being pursued most aggressively in the US while the equally large opportunities in other markets go comparatively unnoticed. Capturing those opportunities requires access to founders building autopilots in the markets where the venture consensus is not yet looking.

That is exactly what SeedScope provides. With active founders across 30+ countries, filterable by stage, sector, and geography, SeedScope gives investors structured access to the companies applying AI to real, high-volume cognitive workflows in markets with enormous, underserved labor budgets. The AI-powered valuation benchmarking lets you evaluate and price these opportunities against global comparables, which is essential when you are assessing a company going after a labor budget in a market you do not operate in daily.

The next generation of services as software companies will not all be built in San Francisco. The labor budgets they are targeting exist worldwide, and the founders best positioned to capture specific regional and vertical workflows are building in markets that most investor deal flow does not reach. SeedScope is built to help you find them.

The Bottom Line

The defining B2B investment thesis of 2026 is not about better software tools. It is about software that does the work, competing for labor budgets that dwarf the software budgets that defined the last era of SaaS.

The math is dramatic, potentially a ten to twenty times expansion of the addressable market. The opportunity is real, because incumbents are structurally constrained from pursuing it and startups are not. And the central risk is specific and underwritable: the last mile from demo reliability to production reliability, which matters more for companies selling completed work than for anyone else.

The companies that solve that last mile, price by outcome, and go after large underserved labor budgets, wherever in the world those budgets are, will be among the most valuable businesses of the coming decade. The investors who understand the distinction between selling tools and selling work are the ones positioned to back them before the rest of the market catches up.

Find the founders selling outcomes, not tools, in the markets others overlook. Explore active founders on SeedScope across 30+ countries. Start here →