Is software being disrupted? The twilight of SaaS, or the dawn of "Service as Software"?

For two decades, the SaaS model has been the gold standard in tech investment due to its high margins and predictable recurring revenue, with San Francisco giants exemplifying the success of the "per-seat" pricing model. However, with the exponential leap in AI, this once-secure crown is showing irreversible cracks. A fundamental upheaval is underway: the rise of AI Agent SaaS is dismantling the economic foundation of traditional software. Traditional SaaS growth relies on client headcount expansion—more employees mean more subscriptions—but in the era of AI automation, the goal has shifted to using agents to reduce human intervention. This creates a fatal "efficiency paradox": as software becomes smarter, clients need fewer humans, leading to inevitable revenue shrinkage for vendors clinging to legacy billing models. We stand at a historic turning point, witnessing a paradigm shift from "Software as a Service" to "Service as Software." In this new era, software is not merely a tool assisting humans but a digital workforce delivering direct business outcomes. This transformation forces tech companies to rethink their value propositions: when AI performs tasks independently, pricing must shift from "seats" to "outcomes." For entrepreneurs and investors, embracing outcome-oriented pricing and digital workforce logic will determine whether they sink in the twilight of SaaS or rise in the dawn of Service as Software.

Introduction: The Crisis of the SaaS Model and the "Twilight" Narrative

Over the past two decades, SaaS (Software as a Service) has been the undisputed commercial crown jewel of the tech world. The Salesforce Tower in San Francisco is not merely a building, but a symbol of the heyday of this business model: distributing software via the cloud and charging a monthly fee for every "head" within an enterprise, thereby building massive and predictable Annual Recurring Revenue (ARR). However, this once impregnable model is now facing an unprecedented existential crisis—a narrative known as the "Twilight of SaaS" is spreading rapidly across investment circles and technical communities.

This is not alarmism, but a keen reaction by the market to changes in underlying logic. The core of the crisis lies not in whether software itself still has value, but in the fact that the rise of AI Agents is fundamentally dismantling the economic foundation of "Seat-Based Pricing." Traditional SaaS growth relied on client enterprises hiring more employees to use the software, but in the AI era, the goal of enterprises is exactly the opposite: to use AI to reduce human intervention and achieve automated operations.

This friction is already reflected in concrete market data. According to Revenue Wizards' analysis, the adoption rate of the pure "pay-per-seat" model dropped from 21% to 15% in just 12 months, while companies sticking to this old model face a customer churn rate 2.3 times higher than that of hybrid models. When AI can complete tasks independently without humans clicking through interfaces, the "Average Revenue Per User" (ARPU) metric that SaaS vendors take pride in loses its meaning.

What we are experiencing is not the end of the software industry, but a paradigm shift from the "Tool Age" to the "Workforce Age." Traditional SaaS consists of tools sold for human use, whereas the new generation model—"Service as Software"—is a digital workforce that directly delivers results. When software is no longer a copilot assisting human work, but a driver directly taking over the work, the old business structure must be rebuilt.

Core Conflict: Why AI Is Destroying "Seat-Based Pricing"

Over the past two decades, the golden rule of the SaaS industry has been built on a simple and solid formula: Revenue = Users (Seats) × Unit Price. This model, known as "Seat-Based Pricing," is tightly bound to corporate personnel expansion—the more people a customer hires, the more money the software vendor makes. However, with the rise of AI Agents, this commercial cornerstone is crumbling, triggering a profound "Efficiency Paradox."

The Efficiency Paradox: When Software Starts "Stealing" Users' Jobs

Traditional SaaS growth logic relies on the inefficiency or expansion of the customer organization. When a company needs to hire more sales, customer service, or finance personnel because of busy operations, SaaS vendors cheer because it means more account subscription fees.

But the core value of AI lies in automation and eliminating human labor. When AI Agents can complete tasks independently, they are no longer human "Co-pilots" but directly become "Drivers." This leads to a fundamental conflict of commercial interests:

The SaaS Efficiency Paradox
The smarter the software, the fewer human employees the customer needs; the fewer people the customer hires, the lower the revenue for SaaS vendors based on "seats."

If vendors stick to the old charging model, they are effectively being "punished" for providing higher efficiency. According to data from Revenue Wizards, in just the past 12 months, the proportion of enterprises adopting a pure "Seat-Based Pricing" model has dropped from 21% to 15%, while the customer churn rate (Churn) for companies sticking to this model is as high as 2.3 times that of other modes.

Crunching the Numbers: Taking the Customer Service Scenario as an Example

To specifically illustrate the collapse of this model, we can refer to the customer support scenario analysis mentioned by a16z:

Assume a company uses customer service software like Zendesk, with 100 customer service representatives, each with a monthly subscription fee of $115.

• Under the traditional model: The SaaS vendor receives $11,500 in revenue per month.
• After AI intervention: After introducing high-level AI Agents, the system can automatically resolve 50% of tickets without human intervention. The enterprise consequently reduces the customer service team to 50 people.
• Result: The SaaS vendor's revenue is instantly halved to $5,750, even though its software now shoulders a heavier workload and creates higher value.

In this example, the "Seat" is no longer the minimum atomic unit for measuring software value. If software can directly deliver "Outcomes" (resolved tickets), then continuing to charge based on the "number of people logging into the software" is no longer logically tenable.

Capital Market Reactions and Model Restructuring

Capital markets have keenly captured this crisis. Investors are beginning to realize that for SaaS companies unable to shake off their dependence on the "headcount tax," their growth ceiling will be lowered by AI. In contrast, shifting to Usage-Based or Outcome-Based pricing models has become an inevitable choice.

Data supports the superiority of this trend: m3ter's analysis shows that companies adopting usage-based pricing models typically have a Net Revenue Retention (NRR) between 115-130%, while traditional fixed subscription models are only 95-105%. This is because under the usage model, revenue expands automatically with the growth of the customer's business (such as API call volume, amount of data processed, or number of tasks completed) without relying on the customer increasing headcount.

This conflict is not just an adjustment of pricing strategy, but a rewriting of the essential definition of software: shifting from "selling tools for humans to use" to "selling services that directly deliver results."

2026 Prophecy Analysis: From "Assistive Tools" to "Digital Workforce"

In discussions regarding the collapse of the SaaS business model, a frequently mentioned specific date is February 17, 2026. This "prophecy" did not come out of thin air but stems from the projections of industry observers (such as Haris Odobasic of Revenue Wizards) regarding the speed of SaaS decline. In this envisioned future scenario, AI Agents will no longer merely be human assistive plugins but will formally become "digital employees" within the corporate organizational structure.

We need to strip away the sci-fi narrative color from this prediction and examine its underlying realistic significance through market logic. This time point actually symbolizes the critical tipping point where software crosses over from Co-pilot to Agent.

1. Core Shift: From "Human-Machine Collaboration" to "Human-Machine Delegation"

Most current AI applications remain in the Co-pilot stage: the user is still the core of operations, with AI providing suggestions or drafts, and humans ultimately clicking "send" or "confirm." In this mode, the traditional SaaS "Seat-Based Pricing" remains barely tenable because human employees still need to log in to the system.

However, the future Agent-as-a-Service (AaaS) model will completely shatter this logic. At this stage, the role of software undergoes a qualitative change:

• Current (SaaS + Co-pilot): Software is a tool; humans are operators. Companies buy Salesforce to let sales personnel record data.
• Future (Service as Software): Software is labor; humans are supervisors. Companies buy AI Agents to let them directly complete the cleaning, contacting, and conversion of sales leads.

2. Why 2026?

2026 is not an absolute deadline but an estimation based on the current speed of AI iteration and corporate adoption cycles. According to analysis by First Analysis, as AI technology matures further in 2026, traditional SaaS valuations may face restructuring, and mergers and acquisitions (M&A) will accelerate, because the old per-seat revenue model will appear completely uncompetitive when facing new software capable of "automatically executing tasks."

3. Reality Reflection: The "Replacement" Already Happening

We do not need to wait until 2026 to see these signs. In the field of Customer Support, companies have already started shifting from "buying Helpdesk software for 50 customer service representatives" to "buying AI customer service to automatically resolve 80% of tickets."

• Old Logic: The harder your software is to use, the more people I need to hire, and the more seats you sell (the efficiency paradox).
• New Logic: The smarter your software is, the fewer people I hire; you must charge based on the outcome of solving problems, rather than the number of users.

This shift means that software companies must stop viewing themselves as "providers of digital tools" and instead reposition themselves as "dispatchers of digital labor." For those traditional SaaS vendors that rely solely on increasing feature stacks but cannot deliver independent work results, 2026 may indeed become the twilight of their business models.

Concept Reconstruction: What is "Service as Software"

Over the past two decades, SaaS (Software as a Service) has redefined the delivery method of software—turning on-premise CD-ROMs into cloud subscriptions. However, "Service as Software" is redefining the essential purpose of software. This is not merely an inversion of word order, but a fundamental reconstruction of the value chain.

The traditional SaaS model is essentially about leasing tools. No matter how powerful Salesforce or Microsoft 365 are, they remain "empty shells" passively waiting for human operation. Software provides capabilities, but humans must provide the labor to produce results. As noted in Saventech's analysis, we are transitioning from "software you use" to "software that works for you."

The core definition of "Service as Software" is: Software is no longer just a tool to assist humans in completing work; the software itself is the "employee" completing the work.

In this model, customers no longer purchase the "possibility of improving efficiency," but "certain results." For example, you no longer subscribe to CRM software for salespeople to enter data (SaaS), but hire an AI agent that automatically scrapes leads, sends emails, and schedules meetings (Service as Software). This shift scales and automates traditional service delivery (usually done by humans) through software, just as Webapper summarized: SaaS changed how software is consumed, while Service as Software changes how services are created and delivered.

In-depth Comparison: SaaS vs. Service as Software

To clearly understand this paradigm shift, we need to deconstruct both from their business models to their underlying logic. If the core of SaaS is "efficiency improvement," then the core of Service as Software is "replacement."

Below is a detailed comparison between traditional SaaS and the emerging Service as Software (also known as Agent-as-a-Service):

From "Selling Shovels" to "Selling Dug Holes"

This table reveals a brutal reality: existing SaaS moats—complex workflows and highly sticky UIs—may become liabilities in the AI era. In Standard Beagle's view, future software design must be "Design for delegation." When the primary user of software becomes an AI agent, those complex interfaces designed to retain human users will no longer hold value, replaced by APIs and agent capabilities capable of directly delivering results.

This shift requires companies to rethink their fundamental business logic: Software deserves to be called "Service as Software" if and only if it can independently close the loop to complete a valuable service.

Deep Comparison: SaaS vs. Service as Software

To understand the essence of this transformation, we must look beyond the superficial differences in terminology and delve into the core of the business model. Traditional SaaS (Software as a Service) and the emerging Service as Software are not merely simple version iterations, but two distinct logics of value delivery.

In short, SaaS is about renting tools, aiming to make "people" work more efficiently; whereas Service as Software is about buying results, aiming to let "software" complete the work directly.

Core Dimensions Comparison Table

The table below summarizes the key differences between the two in terms of user entities, business models, and product forms, clearly demonstrating the leap from "tool attributes" to "service attributes":

1. From "Human-Machine Collaboration" to "Machine Labor"

Under the traditional SaaS model, the value of software is limited by the skills and time of human operators. For example, Salesforce is a powerful CRM tool, but it still requires sales personnel to manually enter data and follow up on leads.

In the Service as Software paradigm, the role of software is inverted. AI Agents are no longer tools passively waiting for instructions, but autonomous "digital employees." As industry analysis points out, future software will not merely be assistive but will directly take over manual workflows via AI Agents. Users no longer need to learn complex operational processes; they simply set a Goal, and the software handles the execution path.

2. The Disappearance of the Interface (The Vanishing Interface)

Traditional SaaS is often known for rich features and complex interfaces, where users need to click through dashboards to extract value. However, when software begins to work autonomously, complex UIs actually become a hindrance.

The ideal form of Service as Software is "imperceptible." If software can automatically handle expense approvals or code deployment, users do not need to log in to the backend. Therefore, the focus of product design is shifting from "designing better buttons" to "designing better automation logic," with frontend interfaces gradually devolving into simple chat boxes or pure backend API calls.

3. Restructuring the Business Model: Pay for Results

This is the most painful transition for traditional vendors. The cornerstone of the SaaS golden age was "Per Seat Pricing," where companies paid for accounts based on the number of employees. But in the AI era, charging by seat may fall into a "death spiral"—because the introduction of AI is intended precisely to reduce human seats.

The future direction is Outcome-Based Pricing. Enterprises no longer pay to own an account, but pay for "successfully booking a meeting," "successfully resolving a customer service ticket," or "generating a qualified legal contract." This model directly binds the vendor's revenue to the client's actual business results, thoroughly changing the economic logic of the software industry.

Transformation Path: How Enterprises Can Cross the "Twilight"

For existing SaaS enterprises, the "Twilight of SaaS" does not mean the end of the industry, but rather a forced evolution of business models. In traditional SaaS logic, revenue is pegged to the number of client employees (Seats); however, the core value of AI Agents lies precisely in reducing dependency on manual operations.

This contradiction creates a dangerous "death spiral": the more efficient your software is (i.e., AI completes more work for the client), the fewer people the client needs to hire, and your revenue actually declines. To cross this twilight, enterprises must fundamentally reconstruct their pricing and delivery models, shifting from "selling tools" to "selling results."

Bidding Farewell to the Seat Economy: Embracing Outcome-Based Pricing

In the AI era, software is no longer merely a productivity aid, but a direct labor substitute. Therefore, the pricing anchor must shift from "how many people use it" to "how much work is completed." EY's analysis points out that this shift requires enterprises to adopt "success-based variable fee arrangements." For example, clients no longer pay for every seat in a customer service system, but for every complaint ticket successfully resolved by AI; they no longer pay for sales software accounts, but for every Sales Qualified Lead (SQL) incubated by AI.

This model completely aligns the vendor's incentive mechanism with the client's success: clients are not only willing to pay a higher premium for determined results, but this pricing model is also uncapped—the more tasks AI processes and the greater value it creates, the higher the software company's revenue becomes.

Implementation Strategy: Hybrid Models and Attribution Challenges

Although outcome-based pricing is the ideal endgame, a direct transition poses huge financial risks and execution difficulties. According to research by BCG, the safest path currently is to adopt a hybrid pricing model.

1. Hybrid Transition (Hybrid Approach): Enterprises can retain basic platform subscription fees to maintain cash flow stability while introducing "agent workload" as a value-added charge item. For example, Salesforce or HubSpot could continue to charge basic CRM fees, but charge extra based on execution count or conversion results for marketing campaigns or automated responses autonomously executed by AI. This method manages revenue risk while allowing clients to gradually adapt to the new value exchange logic.
2. Solving the Attribution Puzzle (Attribution): The biggest obstacle to outcome pricing lies in how to define "credit." If a deal is closed, is it the credit of AI or the impact of the macroeconomy? BCG's survey shows that 47% of buyers struggle to define clear, measurable outcomes, and 25% find it difficult to agree with vendors on value attribution. Therefore, SaaS vendors must establish a transparent, auditable "proof of work" mechanism to clearly demonstrate the specific contribution nodes of AI Agents within the business chain.
3. Redefining Core Metrics (Redefining KPIs): As business models change, metrics originally used to measure SaaS health (such as MAU, DAU) will gradually become obsolete. As pointed out by SavenTech, future core metrics will revolve around "task completion rate," "percentage of automated execution," and "cost per unit outcome."

The key to crossing the twilight lies in whether enterprises dare to disrupt themselves—proactively offering a new solution that allows clients to reduce Total Cost of Ownership (TCO) by "buying results" before clients benefit from AI by cutting seats.

Pricing and Product Form Transformation: Outcome-Based

In the era of "Service as Software," the core value proposition of SaaS is undergoing a fundamental inversion: customers no longer pay for "access to tools," but for "work results." This shift requires enterprises to fundamentally reconstruct pricing models and product forms; otherwise, they face the risk of being marginalized by "digital employees."

Farewell to "Per-Seat Fees": Pay-for-Outcome

Traditional SaaS business models rely on Per-Seat subscription fees, premised on software serving as a tool to assist human work. However, when AI Agents begin to independently undertake complete business processes, the concept of a seat loses its meaning—Agents do not need to log in to accounts, nor do they need to rest.

Future pricing strategies must be directly linked to delivered value. Enterprises need to establish measurement standards based on specific business outputs, rather than software usage time or user count.

• Customer Support Domain: No longer charging by the number of support agent accounts, but billing per Resolved Ticket.
• Sales and Marketing Domain: Fees charged to customers are based on Per Scheduled Meeting or Per Qualified Lead.
• Development Tools Domain: Shifting from charging per developer headcount to billing per Successful Merge or the number of automatically fixed bugs.

This outcome-based pricing model, while potentially bringing longer sales cycles and initial revenue uncertainty, greatly aligns the interests of suppliers and customers. As industry analysis points out, if your AI can deliver results superior to humans at a lower cost, you possess extremely strong pricing power; conversely, if you merely provide homogenized AI capabilities, the price will eventually fall towards the marginal cost of computing resources.

Product Form: From "UI First" to "API and Reliability First"

In the old SaaS model, the usability of the User Interface (UI) was the product's core moat. But in a future dominated by AI Agents, the importance of UI will decline significantly. Agents do not need exquisite dashboards or complex click flows; they need structured input, output, and extremely stable execution capabilities.

Therefore, product forms must evolve towards "Headless":

1. API as a First-Class Citizen: The core competitiveness of a product lies in the richness and response speed of its API, ensuring Agents can smoothly call functions, read data, and return results.
2. Model Reliability as Product: Customers are no longer buying functional modules, but the accuracy and robustness of the model in specific scenarios. For example, the value of a financial AI Agent lies in its zero-error rate when processing invoices, not the aesthetics of its interface.
3. Observability and Intervention Mechanisms: The product focus should shift to providing a backend for managing Agents through monitoring and metrics, allowing human managers to take over or fine-tune when an Agent makes a mistake, rather than directly operating business processes.

Beware of the "Pseudo-Transformation" Trap

Many traditional SaaS companies are currently falling into a dangerous misconception: simply attaching an AI Chatbot via API to existing complex software and declaring they have completed the transformation. This is not "Service as Software," but merely adding a natural language interaction entry point.

True transformation is not about letting users operate old tools through a chat box, but letting AI directly take over workflows. If your product still requires human users to frequently switch between a chat box and a traditional interface, or if AI-generated results still require tedious secondary confirmation by humans, then this remains a "tool" rather than a "service." To achieve a true outcome orientation, one must go deep into the orchestration of the underlying architecture, solving "dirty work" such as cross-system authentication, error retries, and compliance checks, ensuring AI can independently deliver final results like a responsible employee.

Conclusion: Not Extinction, But Evolution

When we talk about the "twilight of SaaS," we are actually talking about the end of the old era's "Rent-a-Tool" model. Software itself has not died out; it is undergoing a species evolution from a "passive tool" to an "active service."

For the past twenty years, the core logic of SaaS has been to provide humans with sharper "hammers" and charge enterprises a "hammer usage fee" on a per-head basis. However, in today's rise of AI Agents, customers are no longer satisfied with buying hammers; they want the result of "driving the nail," or even a "digital employee" capable of independently completing carpentry work. As industry analysis points out, we are shifting from Software as a Service to Service as Software. This is not just a reversal of word order, but a fundamental reconstruction of the way value is delivered.

For all developers, entrepreneurs, and investors, this is both a wake-up call and a bugle call before the dawn:

1. Stop merely manufacturing "better tools": If your software still relies on humans clicking, typing, and correcting errors within a UI interface, then its value ceiling will be locked within the category of "efficiency tools" and face the risk of being replaced by automated processes.
2. Start building "digital employees": Future software should be like an excellent colleague (Smart Colleague), capable of understanding intent, planning tasks, and delivering final results. As emphasized by Workato when analyzing AI Agent evolution, the real value lies in orchestration (Orchestration) and action, not just conversation.
3. Embrace a result-oriented business model: Dare to price for results, not for seats. When software can assume responsibility (Accountability) like an outsourcing service provider, it will also be qualified to share a larger piece of the commercial value cake—this is not only an expansion of the software market but also software's devouring of the trillion-dollar "service industry" market.

SaaS is not dead; it is merely shedding its skin. Those enterprises capable of crossing this chasm and upgrading software from "assisting humans" to "acting on behalf of humans" will welcome a "Service as Software" era even vaster than the past twenty years. Do not go gentle into that good night; go create code that can truly "get the job done" for the world.