AI is shaking up software, but not as fast as stock prices suggest

Salesforce's figures illustrate the gap. In results announced August 26, second-quarter revenue for fiscal year 2027 came in at $11.3 billion, up 11% from a year earlier, and the company raised its full-year revenue guidance by $200 million to a range of $46.1 billion to $46.4 billion. Combined recurring revenue from its AI agent products, Agentforce and Data 360, grew more than 210% year over year to nearly $3.9 billion.

Workday sits in a similar spot. The company, which sells HR and finance software, has also seen its stock pressured by worries that AI-driven headcount cuts would shrink per-seat fees. But according to the article, its results have remained financially solid. For both companies, the worry is about the future, while the numbers reflect the present.

That's where the paradox lies. Even as fears that AI would replace Salesforce dragged the stock down, AI products were the single fastest-growing driver of Salesforce's revenue. The money moved first toward AI as an add-on, not as a replacement.

That said, AI has also shown up on the cost side. Salesforce expects to spend $300 million on Anthropic in 2026, and that spending kept its full-year operating margin forecast stuck at 20.1%. Deputy Chief Financial Officer Mike Spencer explained that the company no longer needs to use the newest model for every task, and has moved into a phase of picking cheaper models job by job.

The cost-cutting approach is concrete. Internally, Salesforce is running Claude, OpenAI, and Cursor side by side, while also testing Grok, choosing different-priced models depending on the task. Spencer said older-generation models are good enough for most jobs, and that choice feeds directly into margins.

The article's conclusion — that things aren't moving "as fast as the worry suggests" — combines these two pictures. AI is displacing enterprise software more slowly than stock prices have priced in. Meanwhile, the pace at which these software companies are selling AI and spending on AI is faster than expected.

That's also why investors and the companies themselves seem to be running on different clocks. The market has already priced in the idea that AI agents doing the work instead of people will break the per-seat software business model. The companies, meanwhile, are busy attaching those very agents to their own products and selling them. Which side turns out to be right will hinge on how much recurring revenue shifts from being counted by seats to being counted by volume of work done.

The same question applies to Korean enterprise software companies. In the first scenario, a company that has been making money on per-seat fees adds an AI agent, customers cut headcount, and fees shrink accordingly. In the second scenario — the one Salesforce is demonstrating — the agent itself becomes a new billable product. So far, the numbers favor the second scenario.

That's why next quarter's results matter. If agent revenue keeps growing at triple-digit rates while per-seat revenue holds up, the second scenario becomes the dominant story. If per-seat revenue falters first, then the market's worry will have been right all along.

In the end, this isn't a story about the worry being wrong — it's about the sequence being different. AI is first changing software companies' costs and adding to their revenue; disrupting the business model comes later. Stock prices have already priced in that final stage, while the actual results are still sitting in the first one.