Software ate the world. Now the agents are eating software.
Software companies fear AI agents could disrupt their entire industry:
A record 27 US public software firms identified AI agents as a competitive risk in their filings in Q1 2026.
The number has more than doubled since Q2 2025.
By comparison, in Q4 2024, just 2 firms flagged this risk in their filings.
Anthropic and OpenAI are releasing AI superagents which can use enterprise software the way humans do, but in a fraction of the time and without supervision.
AI agents could replicate existing software or reduce subscription revenue, as companies that need fewer employees also need fewer software licenses.
Investors are already pricing in the risk, with the Software ETF, $IGV, down -21% year-to-date.
AI is reshaping the software industry.
The SEC filing is the canary in the coal mine. When 27 public software companies name AI agents as a material risk in the same quarter, it's not speculation — it's disclosure-grade certainty. The software industry isn't debating whether agents will disrupt their business; they're legally obligated to tell shareholders it's already happening.
The Kobeissi Letter's breakdown of Q1 2026 SEC filings reveals a tectonic shift in how the software industry views its own future. The jump from 2 companies flagging AI agent risk in Q4 2024 to 27 in Q1 2026 — a 13.5x increase in five quarters — tracks almost perfectly with the deployment cadence of Anthropic's Claude and OpenAI's agent products. The Software ETF ($IGV) dropping 21% year-to-date isn't a blip; it reflects institutional investors repricing the entire category. When Salesforce, ServiceNow, and their peers admit in legal filings that AI agents could "replicate existing software or reduce subscription revenue," they're describing the end of per-seat licensing as the dominant SaaS business model.
The second-order effect is even more consequential: fewer employees means fewer software licenses. McKinsey's 2025 estimate that AI agents could automate 30% of knowledge work tasks implies a corresponding reduction in SaaS spend — not because the software gets cheaper, but because fewer humans need it. Companies like SapienEx that operate at the strategy layer understand this dynamic intimately. The winners won't be the firms clinging to seat-based pricing while agents erode their user counts. They'll be the ones who architect agent-native workflows from day one, treating AI agents as the primary user and humans as supervisors. The risk filings are the establishment admitting what builders already know: the agent era doesn't augment software — it replaces the need for most of it.
The bottom line: When 27 software companies tell the SEC that AI agents are a material threat, the disruption isn't theoretical — it's a line item in the 10-K.
Our Take
The SEC filing is the canary in the coal mine. When 27 public software companies name AI agents as a material risk in the same quarter, it's not speculation — it's disclosure-grade certainty. The software industry isn't debating whether agents will disrupt their business; they're legally obligated to tell shareholders it's already happening.
The Kobeissi Letter's breakdown of Q1 2026 SEC filings reveals a tectonic shift in how the software industry views its own future. The jump from 2 companies flagging AI agent risk in Q4 2024 to 27 in Q1 2026 — a 13.5x increase in five quarters — tracks almost perfectly with the deployment cadence of Anthropic's Claude and OpenAI's agent products. The Software ETF ($IGV) dropping 21% year-to-date isn't a blip; it reflects institutional investors repricing the entire category. When Salesforce, ServiceNow, and their peers admit in legal filings that AI agents could "replicate existing software or reduce subscription revenue," they're describing the end of per-seat licensing as the dominant SaaS business model.
The second-order effect is even more consequential: fewer employees means fewer software licenses. McKinsey's 2025 estimate that AI agents could automate 30% of knowledge work tasks implies a corresponding reduction in SaaS spend — not because the software gets cheaper, but because fewer humans need it. Companies like SapienEx that operate at the strategy layer understand this dynamic intimately. The winners won't be the firms clinging to seat-based pricing while agents erode their user counts. They'll be the ones who architect agent-native workflows from day one, treating AI agents as the primary user and humans as supervisors. The risk filings are the establishment admitting what builders already know: the agent era doesn't augment software — it replaces the need for most of it.
The bottom line: When 27 software companies tell the SEC that AI agents are a material threat, the disruption isn't theoretical — it's a line item in the 10-K.