AI Agents Won’t Kill SaaS

Since AI Agents became popular, many have already started writing eulogies for SaaS. But I think it’s too early.

Investors are indeed panicking. In early 2026, panic about the end of SaaS swept through the tech world. At the end of January, when Anthropic merely released a feature update allowing Claude to call plugins, the market capitalization of the U.S. software sector evaporated hundreds of billions of dollars in the following three weeks.

Their panic logic is simple. They believe that since AI can already write code, find bugs, and even dynamically generate tools on its own, the cost of writing code is approaching zero. Once Agents can create various customized tools for enterprises anytime, anywhere, the moats painstakingly built by those software companies charging monthly subscriptions will naturally vanish.

Thus, from CrowdStrike to IBM, from Salesforce to ServiceNow, regardless of how impressive their earnings reports were, they are all experiencing severe sell-offs.

Meanwhile, countless AI entrepreneurs are holding their business plans, telling VCs they want to “build the middleware for the Agent era” or “start a business for Agents.”

They are all betting on one thing: building tools is the sexiest business of this era.

But if we shift our gaze away from those PowerPoint presentations and look at the real cross-section of how businesses operate, we find that this is not the case at all.

Software Was Never About Selling Code

In economics, there is a classic and repeatedly validated theory called the “Transfer of Factor Scarcity.” Every productivity revolution makes a previously scarce factor abundant while making another previously overlooked factor extremely scarce, with wealth subsequently concentrating in the latter.

Before the Industrial Revolution, labor was scarce; the steam engine made mechanical labor abundant, so scarcity shifted to capital and factories, making factory owners the wealthiest people of that era.

The internet revolution reduced the cost of information dissemination to zero, so scarcity shifted to users’ “attention,” making traffic a huge business.

Today, the AI revolution is making the ability to write code and build tools extremely abundant. In the Agent era where code is no longer scarce, where exactly has the scarcity shifted?

In fact, over the decades of software industry development, code itself has never truly been a moat.

Every line of the Linux system is free, but that didn’t prevent Red Hat from being acquired by IBM for a staggering $34 billion. MySQL is free, yet after Oracle acquired it, it could still sell expensive service contracts. Anyone can download PostgreSQL’s code, but AWS’s Aurora database service still takes in tens of billions of dollars annually from enterprise customers.

The code is free, but the business remains, and it’s doing quite well.

The most crucial things are actually these three: solidified business processes, customer data accumulated over years, and the resulting extremely high switching costs.

When you buy Salesforce, you are not buying the source code of that CRM system. You are buying the over 50 trillion enterprise customer records it manages and the process expertise of how it seamlessly integrates sales, customer service, marketing, and other links. This data is not lines of cold code; it is the living time and history of a business.

A company that has used Salesforce for ten years has every customer communication record, every transaction history, and every follow-up milestone for every sales opportunity stored within it. Migrating away is not just a matter of changing software; it’s equivalent to moving the company’s entire memory. This is why Salesforce can still report $41 billion in annual revenue and set a target of $63 billion for 2030.

Returning to the framework of factor scarcity transfer. Since Agents can create tools themselves and the cost of writing code has dropped to zero, what exactly is the scarcest factor in the enterprise services scenario?

What’s Choking the Agent

What truly chokes the Agent is not that it lacks hands, but that it lacks the “context” in its brain.

A super Agent with all tools is like a top-of-the-line juicer. It spins extremely fast, has sharp blades, but if no one throws fruit into it, it certainly can’t produce a glass of juice for you.

McKinsey’s annual report points out that 88% of enterprises are using AI, but only 23% have truly achieved the scaled implementation of Agent systems in a specific internal process. What’s holding them back is not that the large models aren’t smart enough, but that the enterprise data architecture isn’t ready.

SAP’s President of Data & Analytics, Irfan Khan, mentioned in an interview with MIT Technology Review: “An enterprise cannot throw away its entire general ledger system and replace it with an Agent, because an Agent can’t do anything without business context.”

The “business context” here refers to: where are this company’s financial compliance baselines, what are the regulatory requirements of this industry, the preferences and history of this specific customer over the past decade, the payment terms and default records of this supplier, the performance history and promotion path of this employee… These things are neither publicly available on the internet, nor can they be obtained through web scraping, and AI cannot generate them through text prediction.

Ashu Garg, a partner at Foundation Capital, holds the same view. He says Agents need not just data, but a “context graph,” a reasoning layer that can capture not just what an enterprise does, but also how it thinks. This kind of thing can only be precipitated from real business operations; it cannot be manufactured out of thin air.

Under this logic, scarcity has shifted from the “ability to create tools” to “possessing irreplaceable business context data.”

Since an Agent cannot conjure up a glass of juice on its own, who exactly is holding the fruit?

The Golden Age of Data Landlords

The answer points to those old-timers once thought to be disrupted by AI.

On February 23, 2026, Bloomberg launched an Agentic AI interface called “ASKB.” Bloomberg Terminal is one of the most iconic entities in the software industry. Although it has only 325,000 subscriber accounts globally, each account costs $32,000 per year, meaning Bloomberg takes in over $10 billion annually from these 325,000 accounts alone, accounting for over 85% of Bloomberg LP’s total revenue.

For the internet industry, which operates on “the more users, the better,” this is actually counter-intuitive. Bloomberg has built a solid commercial fortress with an extremely small number of paying users.

The only reason it succeeded is that Bloomberg possesses the world’s most complete, real-time, and deeply structured financial data. This data is the product of decades of continuous investment, including real-time market data, historical archives, news corpus, analyst reports, company financial data… Any institution wanting to make serious decisions in the financial domain cannot do without it.

For the newly launched ASKB, AI is the engine, and Bloomberg’s unique data is the only fuel. Any Agent that wants to function in the financial domain cannot fabricate this data out of thin air; it can only obediently connect to Bloomberg’s APIs.

WatersTechnology gave a very apt comment: Bloomberg’s Agentic strategy demonstrates “how those who own the data turn AI into their own ATM.”

This logic holds true across various verticals. Veeva holds the global pharmaceutical industry’s compliance and R&D data; any pharmaceutical company’s Agent handling clinical trials or regulatory submissions must call upon this data. Epic holds the medical health records of over 250 million patients in the U.S.; every diagnostic suggestion from a medical Agent needs this real patient data as its foundation. LexisNexis monopolizes a vast archive of legal documents; legal Agents performing case retrieval and compliance analysis cannot bypass it.

This data is the crystallization of decades of real-world business operations, the sedimentation of time, and irreplicable history. This is also the ultimate manifestation of the “transfer of factor scarcity”: when everyone possesses top-tier AI engines, what truly determines victory is whether you can find that oil field unique to you.

In the past, these subscription-based data services were sold to human analysts. A large institution might need to purchase 100 Bloomberg Terminal accounts. But in the future, when machines become the consumers of data, an institution might be running tens of thousands of Agents, frantically calling these proprietary data APIs within milliseconds.

This is an order-of-magnitude leap. The number of queries a human analyst can process in a day is limited, but the call frequency of Agents is far greater than that of humans. The demand for continuous, real-time, high-value data will experience exponential growth. The subscription-based business logic is not only not disrupted but is infinitely amplified by the greedy appetite of machines.

Code goes to zero, data starts collecting rent.

But does this mean all SaaS and data companies can rest easy?

Not All SaaS Companies Hold This Card

If this article is interpreted as indiscriminate bullishness on the SaaS industry, that would be a big mistake. What AI brings to SaaS is a brutal and significant divergence.

In early March 2026, TechCrunch interviewed several top VCs, asking them what they least want to invest in now.

Silicon Valley investors are already voting with their feet. Simple workflow encapsulation, horizontal tools applicable to any industry, lightweight project management—stories that once could secure a funding round now share the common fate of being directly passed over. The reason is simple: because Agents can handle these tasks easily. Software companies without proprietary data are quickly losing their eligibility to enter the capital’s field of vision.

This judgment splits the SaaS world in two.

One half consists of those tool-type products offering only thin encapsulation, wrapping publicly available data in a nice interface, or merely optimizing a single-point operational process. The moat of such products is essentially user habits and interface stickiness.

But as Jake Saper from Emergence Capital said: “In the past, getting humans to form habits in your software was a powerful moat. But if Agents are doing this work, who cares about human workflows?”

This type of SaaS indeed faces significant threats. The GTM (Go-To-Рынок) tool stack is a typical case. Gainsight, Zendesk, Outreach, Clari, Gong—these companies occupy adjacent functions like customer success, customer service, sales outreach, revenue forecasting, and call analysis, each requiring separate budgets, separate operations, and separate integrations. AI-native companies can now use a single Agent to connect all these links, significantly diminishing the value proposition of these point-solution tools.

The other half of SaaS is deeply embedded in core enterprise business processes, possessing irreplaceable proprietary data. These companies will not only not be replaced by Agents but will become more valuable because of their existence.

Take Salesforce as an example. In February 2026, Salesforce’s earnings report showed that Agentforce’s annual recurring revenue reached $800 million, a 169% year-over-year increase; it had cumulatively delivered 2.4 billion “Agentic work units” and processed nearly 20 trillion tokens; it had signed over 29,000 Agentforce customers, a 50% quarter-over-quarter increase. More crucially, the combined ARR of Agentforce and Data 360 exceeded $2.9 billion, growing over 200% year-over-year.

Marc Benioff said on the earnings call: “We have rebuilt Salesforce into the operating system for the Agentic Enterprise. The more AI can replace work, the more valuable Salesforce becomes.”

Salesforce has not been replaced by Agents; instead, it has become the soil in which Agents operate. Its value precisely comes from the business data and process context that Agents cannot bypass.

ServiceNow’s CEO, Bill McDermott, publicly announced in February 2026: “We are not a SaaS company.”

He wasn’t denying himself but actively distancing the company. His logic is that SaaS is a concept about “software delivery method,” and what ServiceNow aims to become is the orchestration and execution layer for enterprise AI Agents. AI can identify problems and give suggestions, but the actual execution of actions within enterprise systems still relies on platforms like ServiceNow that are deeply embedded in workflows.

Workday, on March 17, 2026, released “Sana,” a conversational AI suite deeply integrating HR and financial data. The core logic of this product is not to replace Workday with AI, but to feed AI with Workday’s data.

Workday holds the salary, performance, organizational structure, and financial budget data of thousands of enterprises. The depth and uniqueness of this data are something no AI-native startup can replicate in the short term.

Therefore, the real moat is not whether you have data, but whether the data you hold is something others cannot obtain, buy, or create.

The Next Decade: Who Collects the Rent?

In every technological revolution, those who ultimately take the largest profits are often not the ones who invented the groundbreaking new technology, but those who quietly mastered the scarce factors upon which that new technology depends for survival. In this era of rapidly developing AI, the capabilities of large models will grow stronger, and Agents’ ability to write code and create tools on their own will become increasingly widespread.

When these once-considered black-box technologies become infrastructure, the logic of “factor scarcity transfer” leads to only one conclusion: those desperately building tools for Agents are likely not the ultimate winners of this era.

Foundation Capital’s analysis in February 2026 stated that the overall market capitalization of the software industry will expand to 10 times its current size over the next decade. But this 10x growth will not be evenly distributed among all software companies; it will be highly concentrated among those players who can truly master the Agent era.

The real winners are those holding the data assets that Agents cannot bypass.

For today’s entrepreneurs and investors, entrepreneurs in this era have only two destinies: one is desperately building hoes for Agents, the other is first occupying that piece of land. You should know which one you are doing now.

Don’t stare at the Agent’s hands; go for its neck.

Эта статья взята из интернета: AI Agents Won’t Kill SaaS

Related: Hold Bitcoin Mid-Term Short Positions, HYPE Successfully Rides the Wave for Profits | Guest Analysis

Looking back at Bitcoin’s performance last week, the overall trend largely followed the previously established trading framework and rhythm. At the short-term strategy level, the market’s movement closely aligned with our short-term assessment. We executed one short position (1x leverage) based on the short-term trading strategy, achieving a single-trade profit of approximately 2.12% (see Table 2 for details). Regarding the medium-term strategy, Bitcoin continued to adhere to the previously established mid-term approach, maintaining the short position opened at $89,000 (1x leverage). As of last week’s close, the price had retreated to around $65,770, resulting in a paper profit of about 26.10% for this mid-term trade, with a maximum floating profit of around 32.58% during the period. From a market analysis perspective, Bitcoin continued to trade within a weak range-bound consolidation…