Software Stocks Going to Zero? by Adam Khoo

Software Stock Sell-Off: An Analysis of AI Disruption

Key Concepts:

• Claude Co-work: An autonomous AI agent system developed by Anthropic, capable of accessing local files, operating browsers, and executing multi-step workflows.
• System of Record (SOR): A central, authoritative source of data for an organization, acting as a “single source of truth.”
• System of Engagement: The user interface and front-end experience of software, designed for employee interaction.
• System of Intelligence/Automation: The agentic layer of software, interpreting goals and autonomously executing tasks.
• Vibe Coding: The ability to rapidly create software using AI agents without extensive programming knowledge.
• Seat-Based Pricing: Traditional software licensing model charging per user/employee.
• Consumption-Based Pricing: A pricing model based on usage or outcomes, rather than per-user licenses.
• Agentic AI: AI systems capable of acting autonomously to achieve specific goals.

I. The Current Sell-Off & Initial Concerns

Software stocks have experienced a significant sell-off, particularly in 2024, with companies like Salesforce (-28%), Intuit (-34%), ServiceNow (-33%), and Adobe (-23%) seeing substantial price declines. The core concern driving this downturn is the emergence of advanced AI agents, specifically Anthropic’s Claude, and its potential to disrupt the software industry. The fear is that these AI agents could fundamentally alter the value proposition of existing software companies, potentially driving down revenue and even leading to bankruptcies.

II. Claude Co-work: The Disruptive Force

Anthropic’s Claude, backed by major shareholders like Amazon, Nvidia, Microsoft, and Google, has released “Claude Co-work,” an autonomous AI agent system. This system functions as a “digital colleague” – akin to Tony Stark’s Jarvis – with the ability to access local files, operate browsers, and execute complex workflows without human intervention. Claude Co-work can read, edit, move, rename, and create files in applications like PowerPoint, Excel, and PDFs, as well as navigate websites, extract data, fill forms, and book travel. It’s powered by 11 open-source starter plugins for legal, sales, marketing, and finance applications.

III. The Two Primary Fears Driving the Sell-Off

The presenter identifies two key fears fueling investor anxiety:

1. Democratization of Software Development (Vibe Coding): Claude Co-work enables individuals without programming expertise to create custom software rapidly. The example given is building a dashboard tracking company coffee consumption from 50 spreadsheets in minutes. This raises the question: why purchase software from established vendors if equivalent functionality can be created quickly and cheaply?
2. Collapse of Seat-Based Pricing: Traditional software pricing relies on per-employee licenses. As AI agents automate tasks previously performed by human employees, companies are reducing headcount. This translates to fewer required licenses, potentially significantly impacting revenue for companies like Adobe and Salesforce. The fear is a 50-80% drop in sales and profits, potentially leading to company failures.

IV. Historical Parallels & Cautionary Notes

The presenter emphasizes that similar “sky is falling” predictions have often proven inaccurate. He cites two recent examples:

• DeepSeek (January 2023): DeepSeek claimed to build a large language model for $3 million in 3 months, challenging the billions spent by Meta and Amazon. This sparked a 43% drop in Nvidia’s stock, but Nvidia ultimately recovered and thrived.
• ChatGPT Threat to Google (May 2023): Fears that ChatGPT would destroy Google’s search business led to a 35% drop in Google’s stock, but Google’s search market share, ad revenue, and cloud business subsequently grew, and the stock reached all-time highs.

These examples highlight the tendency for market overreactions and the importance of avoiding irrational fear.

V. The Three Layers of Enterprise Software Value

The presenter argues that the true value of enterprise software lies beyond the software itself. He identifies three key layers:

1. System of Record (SOR): The central data repository and “single source of truth” for a company, tracking all transactions and ensuring data integrity. This layer is difficult to replicate due to strict governance, audit trails, and legal/tax requirements. AI can use the SOR, but cannot be the SOR.
2. System of Engagement: The user interface (dashboards, forms, etc.) that employees interact with. This layer is the most vulnerable to disruption by AI agents, as AI can automate manual tasks and provide conversational interfaces.
3. System of Intelligence/Automation: The agentic layer that interprets goals, executes tasks autonomously, and provides specialized intelligence (e.g., classifying support tickets).

Successful software companies provide one, two, or all three of these layers.

VI. Company-Specific Resilience Assessment

The presenter provides a resilience ranking (1-10, 10 being most resilient) based on the three-layer framework:

• Microsoft (9.5): Strong across all three layers, with a robust AI infrastructure (Azure).
• ServiceNow (9.2): Spans all three layers, with a growing focus on consumption-based pricing.
• Constellation Software (8.5): Diversified portfolio of niche software companies, many acting as systems of record.
• Salesforce (7): Provides all three layers, but is more vulnerable than Microsoft or ServiceNow due to a thinner data model and ongoing transition to consumption-based pricing.
• Adobe (5.5): Most vulnerable – primarily a system of engagement company, susceptible to disruption from AI-powered creative tools like Canva and Midjourney.

VII. Key Takeaways & Conclusion

While the fear of AI disrupting software companies is valid, it’s not a foregone conclusion. The impact will vary significantly depending on the company’s ability to provide a comprehensive suite of services (SOR, engagement, intelligence) and adapt to consumption-based pricing models. Companies heavily reliant on seat-based licensing and offering primarily user interface-focused software are most at risk. Investors should carefully analyze individual companies, considering their strengths and vulnerabilities in the face of this evolving technological landscape. The presenter suggests a cautious approach, adding to resilient companies slowly and avoiding those most susceptible to disruption. He reiterates that predictions are often inaccurate and encourages a nuanced perspective.