Physix FrontierHow to Tell If a Software Company Will Be Disrupted by AI
A friend recommended this 'AI Impact Health Check' spreadsheet for software companies. I'm curious to see if it's actually useful. I usually work on compiler optimization on Ascend 910B chips, so my habit is to look at inputs and outputs first, then identify the bottleneck. This sheet works similarly: first look at how the company makes money, then see if AI can replace those revenue points.
Day 1, you only need to do three things.
1. Create a new spreadsheet. In the first row, enter: Company Name, Revenue Model, Is Customer Paying Per Seat?, What Can AI Do?, Migration Cost, Judgment.
2. Pick three enterprise software companies you're familiar with, like office, customer service, or finance tools. Go to their official websites and find the Pricing page. If you see per user, per seat, or "charged per person per month," mark "Yes" in the Per-Seat Payment column. A "seat" is basically an account slot; the company buys one account for one employee. If you see usage-based, "charged per project," or "charged by call volume," mark "No." Usage-based means you pay based on usage—the more you use, the more you pay.
3. If you can't find a pricing page, check the Investor Relations page. Look for terms like subscription revenue or net revenue retention in recent quarterly reports. This way, you can at least understand where the money comes from.
On Day 3, add two columns. The first column notes whether AI can directly complete tasks humans used to do. The second column notes how difficult it is for customers to switch away from them. Here, you need to look at feature pages and case studies. For example, in a customer service software, if the website says an AI agent automatically handles tickets, it might reduce the need for human agents. If it says it helps agents summarize tickets, it's more of an enhancement tool. An AI agent can be simply understood as a program that reads tickets, looks up information, and fills out systems on its own. Implementing enterprise software requires considering permissions, approvals, historical data, and compliance checks.
Some institutions have evaluated over 500 enterprise software companies, suggesting 24% might not survive. Others worry that after a sell-off wiped out nearly $1 trillion in market cap, the panic has gone too far.
After a week, I finished filling out the sheet and felt this method was more reliable than just reading news. Anyone who does compilers knows that operator fusion isn't about forcibly combining operations with similar names; you have to see if intermediate data can be written to memory fewer times. It's the same for AI's impact on software companies: the key is whether it reduces a paid seat or increases call volume. If seats are reduced, companies charging per head face significant revenue pressure; if call volume increases, they might actually make more money.
The pitfalls are obvious. Beginners tend to focus only on stock prices. A big drop in stock price doesn't mean customers will leave tomorrow. Beginners also often mistake AI support for a moat. Many companies just plug a large model next to a search box without changing the workflow. You need to see if the company has integrated AI into legacy systems—can it reduce form entries, cut approval steps, or eliminate an account? News also mentions that some software companies have seen revenue accelerate again recently, with no clear sign of customers being directly taken away by AI.
Next step: pick a CRM or financial software you use regularly, fill out the sheet for it, and compare if its revenue model has changed between last year and this year.
📌 This article is compiled from WSJ Markets. Original text: https://www.wsj.com/tech/ai/ai-is-disrupting-software-companiesbut-not-as-fast-as-many-feared-a380443a?mod=rss_markets_main
Copyright belongs to the original authors. This is a compilation and independent analysis based on public reporting.
