On June 13, The Information published a story about Anthropic blindsiding its own business partners. Weeks before launching Claude Design, a tool that builds UI mockups and prototypes directly inside Claude, Anthropic asked Figma and Canva to co-announce the product as “partners.” The catch: Claude Design competes directly with both companies. Figma’s stock dropped 5% on the news. Its shares are down roughly 50% over the past year.
This is not an Anthropic story. This is the pattern that will reshape every SaaS invoice sitting in your finance team’s queue right now.
The platform always eats the layer above it
Since January, ETFs tracking public software companies have fallen 30%. Salesforce, Adobe, Intuit, ServiceNow, Veeva, all down 25 to 30% in weeks. Andreessen Horowitz published an analysis in May arguing that AI will collapse the distance between intent and execution so completely that the traditional application layer loses its reason to exist.
The evidence is already in the stock prices.
Claude builds prototypes. ChatGPT writes and runs code. Gemini fills forms, books appointments, and compares options inside Chrome without the user ever opening another app. Every one of those capabilities used to require a separate product with a separate license and a separate onboarding process.
The AI platforms are not partnering with your software vendors. They are replacing them. The partnership announcements are the courtesy call before the competition begins.
What this looks like in practice
Take design. A year ago, if your marketing team needed a landing page mockup, they opened Figma, coordinated with a designer, iterated through rounds of feedback, and exported assets. That workflow justified a $45-per-seat subscription and a specialized role on your team.
Today, Claude Design produces a working prototype from a conversation. It is not as polished as what a senior designer builds in Figma. But for 80% of the internal work that used to require a design tool, it is good enough. And “good enough from the AI platform you already pay for” beats “excellent from a tool that costs extra and requires a specialist.”
This is the pattern. Not replacement at the top of the quality curve. Absorption at the middle, where most of the volume lives.
The same thing is happening with code editors, with data visualization tools, with customer support platforms, with project management software. Each time a frontier AI model ships a new capability, it absorbs a thin slice of what used to be a standalone product. No single slice is fatal. But they compound.
The vendor strategy question nobody is asking
Most organizations I work with have between 40 and 120 SaaS subscriptions. Each one was justified individually. Each one has an owner, a workflow, and a renewal date. Nobody is looking at that list and asking: which of these sit directly above the AI layer, doing work the AI platform can now do natively?
That is the question.
Not “should we switch to AI?” You already use AI. The question is whether you are paying for three tools when one of them absorbed the other two six months ago and nobody updated the vendor list.
Here is how I think about it. Every tool in your stack falls into one of three categories right now:
Tools that own unique data or workflows the AI cannot replicate. Your ERP, your CRM with 10 years of customer history, your industry-specific compliance platform. These are safe. The switching cost is the moat.
Tools that sit on top of the AI layer and add a specialized interface. Design tools, writing assistants, analytics dashboards, scheduling platforms. This is where the absorption happens. The AI platform is building that interface into itself, one feature at a time.
Tools that are already redundant but nobody has checked. The screen recording tool your team stopped using when the AI started generating walkthroughs. The translation service that still auto-renews even though Claude handles it. Every organization has these. Most do not know how many.
The operations problem underneath
This is Belief 1 territory. Handling this well has nothing to do with having the best AI models or the most technical team. It requires a quarterly vendor audit with a simple filter: does this tool do something our AI platform cannot?
That is an operations discipline, not a technology decision. It requires someone in procurement or ops to sit with the team leads who own each tool and pressure-test the answer. Not once. Quarterly. Because the AI platforms ship new capabilities every few weeks, and every release moves the line.
Anthropic added design. OpenAI added deep research that competes with consulting deliverables. Google added agent-powered browsing that competes with task automation platforms. Three releases. Three categories of SaaS tool suddenly in the crosshairs. And that was just the last 60 days.
The real risk is inertia
The Figma story is instructive not because Figma is doomed. Figma will adapt. The company has genuine workflow depth that matters to professional designers.
The risk is for the businesses paying for Figma seats that are used 4 hours a month by people who are not designers. That is the spending that disappears first. Not because someone made a strategic decision, but because a team lead realized Claude already does the thing and stopped submitting the expense report.
Getting ahead of this means making it a process instead of an accident. Audit the stack. Identify the tools sitting one layer above the AI. Decide deliberately which ones still earn their seat.
The AI vendors are not waiting for you to figure this out. They are shipping features that make the decision for you. The only question is whether your operations team names the overlap before your finance team discovers it in next quarter’s renewal cycle.