如果 AI 改变交互界面，什么仍然稀缺、谁还能收费？

In my previous note, I wrote that AI doesn’t need to replace software to weaken its moat. I’ve kept thinking about that because several companies I own or follow have de-rated a lot. The question I’m now asking is: what is the economic bottleneck that AI still has to pass through? A business can survive AI without preserving its economics. The more important question is whether AI makes the scarce part of what it sells less scarce. An agent can take over part of the interface, reduce seat growth or make some software easier to replace.

Experian looks particularly interesting to me. An AI agent can change how a bank makes a credit decision, but it still needs credit histories, identity data and fraud signals. FY26 organic revenue still grew 8% and Benchmark EPS 15%. The interface can change quite a lot while Experian’s underlying data remains valuable.

RELX is similar. Lawyers and researchers may increasingly use AI agents, but those agents still need reliable legal, scientific and risk information. RELX grew underlying revenue 7% and operating profit 9% in H1 2026. In both cases, the valuable part sits underneath the interface, which seems much harder for AI to bypass.

Intuit is less clear. QuickBooks owns a lot of accounting history and sits inside important workflows. Online Ecosystem revenue grew 19% last year and QuickBooks Online Accounting 23%. But I can also imagine a small business owner barely opening QuickBooks anymore, with an agent doing much of the accounting work in the background. That makes payments, payroll and tax more important. Maybe those become the real bottlenecks rather than the interface itself.

Adobe looks less interesting to me after applying the same idea. The business is still growing — Q3 revenue was up 13% — but AI attacks something much closer to what Adobe historically monetized: creating and editing through its tools. Adobe still has PDF, Acrobat and strong distribution, but I’m less sure what remains truly scarce.

I also looked at MSCI and S&P Global, but I think they are less clean examples of a current dislocation. Both had already gone up enormously before their recent stagnation. A $100 investment in MSCI at the end of 2016 was worth roughly $819 by the end of 2021, while S&P Global reached about $460 over the same period. Some of the recent stagnation may simply be the businesses growing into valuations that had already moved far ahead.

So I keep coming back to Experian, RELX and Intuit. Adobe looks less compelling to me. The question I keep asking is: if AI changes the interface, what is still scarce — and who still gets paid?

No posts