Why AI Is Not Eating Enterprise Software
“AI is eating enterprise software” is the loudest story in tech. Until you look at what AI can’t actually chew through.
Due to popular demand, I am offering my stock analyses in both English and German (DE). If you would prefer to read this article in German, click here. If you would like to receive the German articles by email in the future, you can change the settings (notifications) in your Substack account here.
Fifteen years ago, Marc Andreessen published his famous essay, “Why Software Is Eating the World,” in the Wall Street Journal. In it, he explained a broad technological shift in which software was taking over entire value chains in traditional industries, from media (Netflix) and retail (Amazon) to cars (Tesla) and financial services. This shift was driven by two main factors: a massive increase in reach, thanks to broadband and smartphones, and a dramatic drop in start-up costs, thanks to the cloud and modern software tools. These changes enabled new software companies to attack established industries more quickly.
Now, many AI enthusiasts believe that “AI is eating the software industry.” Looking at the disastrous price developments of most enterprise software stocks, one might think that an implosion of these companies is imminent.
We’re not just talking about the significant declines of aging software conglomerates, such as Salesforce and Adobe, which for many years have grown primarily through acquisitions and haven’t experienced substantial organic growth in years. We are also talking about their attackers. I’m referring to younger, faster-growing software companies with high gross margins over 80% that have lost 50-70% of their value in the past 12 months. Examples include Atlassian, HubSpot, GitLab (all from my watch list), and Monday (from my portfolio).
All these B2B software companies have one thing in common: measured by revenue or cash flow multiples, they are more affordable today than ever before.
During the first weeks of the new year, investors fled these software stocks.
The reason: Many tech observers believe that, in the not-too-distant future, AI agents will perform a significant portion of the work in companies. What will the future of enterprise software look like then?
Will companies use AI-based vibe coding tools to program their own personalized enterprise software instead of buying expensive subscriptions from established SaaS providers?
With each new release and improvement in the coding capabilities of AI tools like Claude Code and Lovable, social media is flooded with self-proclaimed AI experts claiming that enterprise software is obsolete because, in the future, every company will be able to build its own ERP or CRM system with minimal effort.
The financial market apparently believes that the era of predictable growth for enterprise software providers based on recurring revenues is coming to an end. There is no other explanation for the poor performance of relevant SaaS stocks on the market.
With 25 years’ experience as a software entrepreneur, I believe it is time to address the panic among investors with this post.
First, the blanket statement “Enterprise SaaS is dead” is nonsense - if only because you can’t lump all enterprise software systems together. It’s important to first distinguish between a System of Record (SoR), a System of Engagement , and a System of Intelligence/Automation because these layers of software value creation face very different challenges in the age of AI.
System of Record (SoR) – the “Status” of the Company
Every company operates at least one core system (usually several) in which the organization’s central data and business processes are mapped. This SoR is the binding “source of truth” for a company, containing data models, transactions, authorizations, controls, and audit trails.
An SoR provides central master data (customers, products, employees, accounts), requires transaction security (e.g. order-to-cash and procure-to-pay processes), ensures governance (roles, policies, and approval chains), and is designed for traceability (who, what, when, and why).
Even in the age of AI, these central systems are not easily replaceable. AI can generate content for these systems and possibly make decisions, but it cannot render a legally robust and consistent state for companies.
Examples of classic SoRs are SAP and Oracle ERPs, Workday HR and Salesforce CRM. These legacy systems have (completely independently of AI) faced different challenges for years and have had to compete with new, younger, more flexible, and, above all, cheaper systems. However, fundamentally, they cannot be replaced by AI, only by another SoR. This usually only happens when the “pain” associated with the old system becomes unbearable.
System of Engagement (SoE) – the Work Interface for People
An SoE is used wherever people “work”: UI, workflows, forms, dashboards, and collaboration. This software is designed to optimize usability, thereby promoting user adoption and engagement.
A typical SoE provides task lists, inboxes, and case handling; defines process steps for teams (e.g., sales, support, HR, and finance); creates context; and promotes collaboration (e.g., via comments, attachments, and notifications). Examples of Systems of Engagement include Zoom, WIX, and Slack (now part of Salesforce).
AI is likely to bring significant changes to this area in the future. The classic “click-based” UI, with humans as click workers at its center, is partially being replaced by conversational UIs (chat/voice), proactive AI suggestions (”next best action”), and automatic documentation and follow-ups.
It’s important to understand that, while AI support can radically transform the SoE, it still needs an SoR underneath it.
System of Intelligence & Automation (SoI/SoA) – Agent Layer as “Operator”
In addition to the SoR and SoE, systems that sit between the two layers have existed since the 1990s. These systems are designed to automate processes and workflows. Initially rigid and deterministic, these workflow systems are now being supplemented by AI models and agents, which always require suitable orchestration via rules and tools. This layer contains the intelligence of the system and interprets goals, plans steps, and executes tasks in the SoR via APIs.
An agent layer can automate routine tasks, such as data maintenance, ticket classification, and quote creation, including orchestration across different systems (CRM ↔ ERP ↔ Support ↔ Billing).
In the future, “Vibe workflows” will enable non-programmers to formulate actions using natural language. Agents defined in this way can support decision-making, but they always require clear authorizations and limits specified by the SoR.
The Future of Enterprise Software
Enterprise software will not disappear with the advent of agents, but it will change as more actions are performed by AI than by humans.
The system of record will maintain the status quo and become more important.
The system of engagement will lose importance as a human work interface and has to change.
The system of intelligence/automation will become the valuable layer that performs work across all systems.
If you are a software investor who wants to address the vulnerability of the SaaS companies in your portfolio to potential AI disruption, ask yourself the following three questions:
1. Does the company have a system of record, or is it just about the user interface (system of engagement)?
The winners are the providers who maintain the binding state - that is, the binding “state layer” that humans and agents must agree on. Products whose main value lies in a fancy UI and manual click work are more likely to lose out when agents perform tasks triggered by text or voice because the advantage of a “nice UX that people enjoy working with” shrinks.
2. How deep is the data model, and how unique is the data?
Winners have complex, domain-specific data models with many dependencies and collect unique data over time, such as process and transaction histories, asset and configuration data, and contract and billing data. This data becomes even more valuable in the AI world because it is the only way agents can act reliably. Losers are providers with thin data models and data that is easily exportable or generated in other systems. If the data does not account for a large part of the added value, the UI layer can be more easily replaced.
3. Can the company monetize the use of agents without cannibalizing its own human user base?
The enterprise software business model will have to evolve over time. Currently, the vast majority of SaaS products are billed per user, which usually reflects the majority of software usage by end users. However, as AI agents take on a larger share of interaction and work with the software, enterprise systems will need to evolve to support a consumption- and usage-based model.
The winners will strike a balance by enabling agents to perform more tasks and monetizing usage (e.g., transactions, automations, API calls, and managed workflows) without eliminating the existing human user base. Ideally, agents will increase overall usage, allowing more value to flow through the system. Losers are software providers who focus too heavily and inflexibly on seat sales and cannot offer a compelling alternative to usage-based pricing.
In the future, we will see an increasing mix of monetization components. Users will continue to be “seats” in the software, and AI agents, which represent increased usage, will be monetized through consumption.
This is already happening in many products, such as HubSpot and Monday, where the end-user component is combined with a consumption model.
It remains to be seen on a quarterly basis how successful these companies will be in their transition into an AI-supported enterprise software. However, the stock prices of these companies already reflect a great deal of negative developments surrounding AI disruption that I have not yet seen.
The Example of monday.com
At a current price of $125, Monday (MNDY 0.00%↑) shares are valued at an Enterprise Value of only $4.7 billion. This equates to just three times the expected revenue for 2026 and 14 times the cash flow over the past 12 months.
Thus, the company is worth significantly less on the stock market than it was at the time of its IPO in June 2021. That was almost five years ago. At that time, the company had revenues of less than $300 million. Today, that amount has quadrupled.
We’re talking about a company with a gross margin of nearly 90%. It has grown extremely efficiently, with organic revenue growth at 28% in 2025 and a free cash flow margin over 25%.
I now view this as an exaggerated downward trend in the stock market and have recently cautiously increased my position in MNDY. I also believe that Monday Software, with its history as a work management solution, is being misjudged prematurely by many investors as a pure “system of engagement.”
Monday has evolved considerably in recent years with its successful modules for CRM and service management, and is now much more than a “one-trick pony” for work management. When I speak with Monday’s customers, I find that, for small and medium-sized companies, Monday has become their “single source of truth,” and they cannot imagine operating without it. In other words, it is for them a true “system of record.”
Of course, we will have to wait and see if the company can maintain its position in the market amid the rise of AI agents and adapt its current seat-based monetization model accordingly. This will all take time, just as the introduction of agents in general will not be as quick or smooth as AI enthusiasts claim today.
Despite the drop in share price, I remain invested in Monday and will closely monitor its further development in the Enterprise SaaS sector here on my Substack.
*Disclaimer: The author and/or related individuals or entities own shares of monday.com. This article is an expression of opinion and does not constitute investment advice.
I work for a software co and also use various enterprise apps. This is my major investing theme for '26, as the current narrative is 100% backwards for the reasons you cite. The B2B apps that control and can use the data with agents will be the AI winners.
The System of Record v/s System of Engagement framework is a very good framework to analyze these companies. If Monday is indeed operating as a System of Record for established companies, then it could be an opportunity. However, what I worry about is increased competition for these kind of products themselves, since it has become easier to build them from scratch.