AI Over SaaS: The TWR. Analysis of YC's S26 Request for Startups

by TWR. Editorial Team | Wednesday, April 29, 2026 for The Weekend Read.

---

There is a moment in every cycle when the language lags behind the reality. In the early 2000s, companies still talked about “installations” even as the internet was quietly reshaping distribution. In the early 2010s, executives debated whether the cloud was safe while their competitors were already moving core systems off-premise.

We are in another one of those moments now. The language of software still dominates boardrooms. Product. Seats. Licenses. Roadmaps. Yet the underlying economics that made those concepts durable are starting to loosen.

The Summer 2026 Requests for Startups from Y Combinator reads like it understands that tension. Not loudly. Not rhetorically. But structurally. It points to a world where software is no longer the primary unit of value. Systems are.

---

The SaaS Seat Model and Its Limits

For two decades, enterprise software scaled on a simple premise. Companies paid per user. More employees meant more seats. More seats meant more revenue.

It was clean, predictable, and easy to model. Investors rewarded it. Public markets reinforced it. Entire organizations were built around expanding usage inside customer accounts.

But the model carried an assumption that often went unspoken. Work scaled with people.

If a company hired more accountants, it needed more accounting software seats. If it expanded its sales team, it needed more CRM licenses. Growth in labor translated directly into growth in software spend.

That relationship is beginning to break.

Year to date in 2026, software companies are delivering strong earnings but losing market value as capital rotates toward AI model layers, chips, and infrastructure. The divergence reflects a structural repricing from seat-based software to systems that execute work.

AI systems do not require seats in the traditional sense. They do not log in, navigate interfaces, or consume software the way people do. They operate across systems, often invisibly, executing tasks that previously required human input.

When one system can handle the workload of many individuals, the link between headcount and software revenue weakens. The question is no longer how many people use the tool, but how much work the system performs.

This is not theoretical. Across categories, companies are already asking a different question: not “how many licenses do we need,” but “how much of this function can be handled without adding headcount.”

Public Markets Are Starting to Notice

Over the past year, software multiples have compressed relative to their peak. Growth expectations have adjusted. Investors are looking more closely at retention, expansion, and the durability of revenue tied to usage.

Companies that once relied on steady seat expansion are encountering a more complicated environment. Customers are scrutinizing spend. Consolidation is becoming more common. In some cases, entire categories of tools are being replaced by fewer, more integrated systems.

Before the market decides where software value is moving, leaders can pressure-test the model themselves below:

This is not a collapse of software as an industry. It is a re-pricing of what kind of software deserves a premium.

Businesses that remain tied to surface-level features or incremental productivity gains are finding it harder to justify their position. Those that are embedded in core workflows, or that directly influence outcomes, continue to command attention.

The shift is subtle but important. Value is moving from access to execution.

From Interfaces to Systems

Historically, software companies competed through interfaces. Better dashboards. Cleaner workflows. More intuitive navigation. The interface was the product.

The RFS suggests a different center of gravity.

Consider the idea of a company brain. It is not a tool that someone logs into. It is a structured representation of how decisions are made across an organization. Refund policies, pricing exceptions, incident responses. All of the small but critical processes that keep a company functioning.

These processes rarely live in one place. They are distributed across documents, messages, and individual experience. Humans bridge those gaps through judgment and memory.

AI systems cannot operate that way. They require structure. They require consistency. They require context that can be accessed and acted upon programmatically.

"Each of these shifts on its own would be significant. Together, they suggest a broader reorganization."

Once that structure exists, the interface becomes less important. The system can execute tasks directly. It can monitor outcomes. It can adjust behavior. The human role shifts from operating the system to supervising it.

This is where software begins to recede into the background. Not because it disappears, but because it becomes part of a larger system that is defined by what it does, not how it looks.

Services Begin to Move

If software is losing its position as the primary unit of value, where does that value go?

The RFS points clearly toward services.

Accounting, insurance brokerage, compliance, healthcare administration. These are not new markets. They are large, established, and deeply embedded in the economy. They are also expensive, labor-intensive, and often inefficient.

Software has historically supported these industries rather than replacing them. Tools improved workflows, but the underlying service model remained intact.

AI changes that boundary. Once exposure is clear, the next question is where the opportunity actually lives.

When systems can handle multi-step processes with a level of reliability that approaches human performance, the distinction between software and service starts to blur. A company can move from selling a tool that assists accountants to delivering the accounting function itself.

This is not simply a pricing change. It alters the structure of the business. Revenue becomes tied to outcomes rather than usage. Margins reflect automation rather than labor. Scale is driven by system performance rather than headcount.

The implications extend beyond startups. Incumbents in these industries are not just competing with new tools. They are competing with new models.

The Return of the Physical Constraint

One of the more striking aspects of the RFS is how frequently it moves beyond software.

Agriculture, defense, hardware supply chains, semiconductors, space. These are domains where physical constraints dominate. Materials, logistics, manufacturing cycles. Areas that software once touched lightly, if at all.

What has changed is not the nature of these industries, but the ability of systems to operate within them.

"When a small team can deliver a solution that addresses a specific, high-value problem quickly ... the cost of experimentation decreases. The potential upside increases."

Computer vision can identify individual plants in a field. Robotics can act on that information with precision. Biological advances introduce alternatives to traditional chemical inputs. In defense, distributed systems can coordinate responses in real time. In semiconductors, visibility across complex supply chains becomes possible in ways that were previously impractical.

These are not incremental improvements. They reshape cost structures and decision-making processes.

For a long time, software founders could ignore these domains because they were slow, capital-intensive, and difficult to penetrate. That assumption is becoming less reliable. The combination of AI with advances in hardware and materials is opening spaces that were previously inaccessible.

Every startup in Y Combinator’s current batches still receives $500,000 in funding, a structure that continues to define the baseline for early-stage capital in 2026.

A Different Kind of Buyer

Another shift, less visible but equally important, concerns who is buying.

Large enterprises have traditionally been difficult customers for early-stage companies. Long sales cycles, complex procurement, and a preference for established vendors created barriers that were hard to overcome.

AI is changing that dynamic at the margins.

When a small team can deliver a solution that addresses a specific, high-value problem quickly, the calculus shifts. The cost of experimentation decreases. The potential upside increases. Decision-makers become more willing to engage earlier.

This does not eliminate the challenges of enterprise sales. It changes the starting point. Instead of requiring years to build a product that meets enterprise standards, companies can reach that threshold more quickly, and in some cases, with fewer people.

For startups, this opens a path that was previously narrow. For incumbents, it introduces a new source of competition.

What Comes Next

The patterns in the RFS are not predictions in the narrow sense. They are observations about where constraints are moving.

Software is becoming easier to produce. Interfaces are becoming less central. Services are becoming addressable by systems. Physical industries are becoming more responsive to digital control. Buyers are becoming more open to new entrants.

Each of these shifts on its own would be significant. Together, they suggest a broader reorganization.

Companies will be defined less by the products they sell and more by the systems they operate. Revenue will be tied less to access and more to outcomes. Value will accrue to those who can integrate across layers rather than optimize within one.

For investors, this requires a different lens. Evaluating a product is not enough. The question becomes whether a company can capture and control a meaningful portion of a workflow.

For founders, it changes where to look. Opportunities are not just in building better tools, but in identifying where entire processes can be restructured.

For leaders inside existing organizations, it raises a more immediate concern. Whether the company’s current structure allows it to adapt to systems that require clarity, consistency, and integration at a level that many organizations do not yet have.

TWR. Last Word: When workflows collapse into systems and seats stop defining value, the real divide emerges between those who adapt their models and those who defend them.

---

Insightful perspectives and deep dives into the technologies, ideas, and strategies shaping our world. This piece reflects the collective expertise and editorial voice of The Weekend Read  — 🗣️Read or Get Rewritten  | www.TheWeekendRead.com

Nomenclature

Seat-Based Pricing: A revenue model where software growth is tied to the number of human users rather than the amount of work performed or value created

Workflow Ownership: Control over an entire operational process from input to outcome, rather than participation in a single step

Outcome-Based Model: A pricing and delivery structure where value is tied to measurable results such as cost reduction, revenue generation, or time saved

AI-Native System: A system designed from the ground up where AI executes core functions, not just assists human users

Company Brain: A structured, continuously updated representation of how an organization operates, enabling systems to make and execute decisions consistently

Agent: A software system capable of autonomously executing multi-step tasks across tools, data sources, and environments

Dynamic Interface: A software layer that adapts to user or system context rather than presenting a fixed UI

Service Replacement: The transition from software assisting labor to software fully performing the service itself

Inference Layer: The operational layer where AI models process inputs and generate real-time outputs within workflows

Closed-Loop System: A system that continuously observes performance, makes decisions, executes actions, and refines itself based on results

Sources

Y Combinator. (2026). Requests for startups (Summer 2026).https://www.ycombinator.com/rfs 

Y Combinator. (2025). What surprised us most in 2025.https://www.ycombinator.com/blog/what-surprised-us-most-in-2025

Crunchbase News. (2026). Global venture funding trends show AI dominance.https://news.crunchbase.com/venture/funding-data-third-largest-year-2025/

Carta. (2025). State of private markets: Q1 2025.https://carta.com/data/state-of-private-markets-q1-2025/

Fenwick. (2026). 2025 Venture Beacon: Key VC market trends.https://www.fenwick.com/insights/publications/q2-2025-venture-beacon-key-vc-market-trends

Reuters. (2025). Healthcare AI startup Abridge raises $250 million.https://www.reuters.com/business/healthcare-pharmaceuticals/healthcare-startup-abridge-raises-250-million-enhance-ai-capabilities-2025-02-17/

Reuters. (2024). Enterprise AI search startup Glean raises $200 million.https://www.reuters.com/technology/us-enterprise-ai-search-startup-glean-raises-200-million-plans-hiring-spree-2024-02-27/

Reuters. (2026). Defense tech firm Anduril reaches $30.5 billion valuation.https://www.reuters.com/business/anduril-secures-305-billion-valuation-latest-fund-raise-2025-06-05/