Skip to main content
MindStudio
Pricing
BlogAbout
My Workspace
Claude for ExcelClaude for financial servicesAnthropic enterprise products

Claude's Enterprise Push: How Anthropic Is Squeezing SaaS Stocks

Anthropic keeps launching Claude products for finance, legal, security, and design, and each launch has rattled a SaaS incumbent's stock price.

Edited by Luis Chavez-Mattos, Director of Product RSS
Claude's Enterprise Push: How Anthropic Is Squeezing SaaS Stocks

What is Anthropic actually doing to SaaS companies?

Anthropic has spent the last two years releasing a steady stream of Claude products aimed at specific industries and job functions, including finance, legal, security, and design. Each launch narrows the gap between what Claude can do out of the box and what a dedicated enterprise software company charges a subscription for. Several of these launches have landed on the same day as visible drops in the stock price of the SaaS company whose territory Claude just entered, including Thomson Reuters, CrowdStrike, and Figma. Anthropic has also disclosed in its IPO filing language stating it “may from time to time compete with enterprise software and application companies,” which is about as direct a warning as a company can give its own ecosystem.

TL;DR

  • Anthropic has rolled out a wave of vertical Claude products, including Claude for financial services, Claude for Excel, Claude for life sciences, a legal plugin suite, Claude Code Security, and a design-focused release.
  • Stock moves followed several launches: Thomson Reuters dropped 16% after the legal plugin suite, CrowdStrike fell 11% days after Claude Code Security shipped, and Figma fell about 7% the day Claude’s design product launched.
  • Anthropic’s own IPO filing states the company may compete directly with enterprise software and application companies, even though its usage terms bar customers from using Claude to build competing products.
  • The mechanism is data: enterprise customers and API users feed Claude their workflows and prompts, and Anthropic’s Economic Index report analyzes over a million business transcripts to see what tasks AI can already automate.
  • Companies worried about this dynamic are shifting workloads to open weight models they run on their own servers, trading some capability for control over their data, sometimes called knowledge alpha.
  • Thin “wrapper” products built as a light interface on top of a foundation model are the most exposed, since Anthropic and OpenAI can rebuild that same interface as a plugin in days.
  • Builders trying to survive this cycle are advised to go deeper than a UI layer, target niches the big labs won’t bother chasing yet, and avoid feeding proprietary SOPs into prompts they don’t control.
Self-host
Remy.
On-prem or in your private cloud. Remy and every app you build stay inside your walls.

How did Claude’s product launches line up with SaaS stock drops?

The pattern described by industry observers goes like this: Anthropic ships a Claude product aimed at a specific vertical, and within a day or two, a publicly traded SaaS company serving that same vertical sees a stock decline. Thomson Reuters, which sells legal research and workflow software, fell 16% around the time Anthropic launched its Claude Co-works legal plugin suite. CrowdStrike, a major cybersecurity vendor, dropped 11% a few days after Claude Code Security arrived. Figma, the design and prototyping platform, fell roughly 7% on the day Anthropic released a Claude product aimed at design work.

None of this proves Anthropic single-handedly caused those drops. Markets react to many signals at once, and correlation over a few days isn’t formal causation. But the pattern is consistent enough that investors appear to be pricing in a real threat: that Anthropic can use its existing model and distribution to replicate a chunk of a SaaS company’s value proposition essentially overnight, without needing years of product development or a dedicated sales team.

Why does Anthropic keep launching into the same verticals?

Anthropic’s rollout reads like a checklist of profitable enterprise software categories: Claude for financial services, Claude for Excel, Claude for life sciences, Claude Sheets, Claude Whiteboards, Claude Data, Claude Meetings, Claude for Manufacturing, and Claude for financial advisors, among others. OpenAI has made a parallel move, launching ChatGPT for sales and listing it inside a plugin library alongside established sales tools like Apollo, Clay, and Gong.

The logic is straightforward. These labs already have the model, the infrastructure, and hundreds of millions of users. Building a plugin or vertical product on top of that base costs them far less than it costs a standalone SaaS company to build a comparable product and acquire customers from scratch. They also don’t have to market it as a new startup entering the space. It ships as a one-click feature inside a product people already use every day.

Where does the product-planning data actually come from?

The most important detail in Anthropic’s strategy is what happens to the data that flows through Claude in the first place. On most plans, conversations with Claude can be used to train future models unless a user explicitly opts out, a policy that mirrors what OpenAI, Google, and others disclose in their own terms. But opting out of model training doesn’t mean Anthropic learns nothing. Even without training on the content, the company can still observe usage patterns: which plugins get used, how often, by what kind of customer, and for what kind of task.

The enterprise platform for building software with AI.
Secure, compliant, auditable apps, built to scale.

Anthropic formalizes part of this analysis in a periodic report called the Economic Index, where Claude reviews large volumes of business API transcripts (reportedly over a million at a time), sorts them by task type, and evaluates which tasks AI is already capable of automating or absorbing. That analysis doubles as a demand map. It shows Anthropic exactly which workflows people are already trying to get Claude to do, which is also a shortlist of what to build next.

There’s a notable asymmetry here: Anthropic’s usage terms prohibit customers from using Claude to build a directly competing product, while its own IPO filing reserves the right to compete with enterprise software companies. One rule applies to users, a different one applies to the platform itself.

Is switching to open weight models a real defense?

For companies worried about their internal processes becoming someone else’s product roadmap, the emerging response is running open weight models (models whose weights can be downloaded and hosted on a company’s own infrastructure, rather than accessed through an API). Keeping inference in-house means sensitive prompts, documents, and workflows never leave the building, cutting off the data exhaust that feeds a frontier lab’s product decisions.

AT&T has been cited as an example of this shift at scale, reportedly directing a meaningful share of its AI usage toward open models with a stated goal of pushing that share much higher. Surveys referenced in industry discussion suggest a notable share of companies want open weight models in production, and a majority have already moved at least some workloads off closed, API-based systems for this reason. Cost is part of the motivation too: running your own compute at scale can be cheaper than paying per-token API pricing for frontier-level intelligence, especially once a company’s usage is high enough to benefit from batching efficiencies on its own hardware.

The tradeoff is real. Open weight models generally trail the best closed models on raw capability, and self-hosting requires infrastructure and expertise most companies don’t have in-house. It’s a bet on data control over peak performance.

What should builders do if they’re exposed to this?

A few practical responses show up repeatedly in how people are thinking about this shift. Thin wrapper products, meaning a basic interface sitting directly on top of a foundation model’s API with little added value, are the most vulnerable, since a lab can replicate that interface as a built-in feature. Building something that goes deeper than a UI layer, whether that’s proprietary data, physical-world integration, regulatory accountability, or hard-to-automate relationship work, makes a product harder to absorb.

Picking a narrow niche rather than a broad category also helps, since large labs tend to prioritize big, obviously profitable verticals first and leave smaller niches alone, at least for now. And on the data side, feeding detailed standard operating procedures or proprietary playbooks into a hosted model’s prompt is worth thinking twice about, since that information lives in a transcript the provider could draw on later. Tools that let teams switch between hosted models like Claude and self-hosted open models, depending on how sensitive the task is, are becoming part of that defensive toolkit.

Frequently Asked Questions

What is Claude for financial services?

It’s one of Anthropic’s vertical product launches aimed at the financial industry, part of a broader pattern of releasing Claude variants tailored to specific professions and industries rather than a single general-purpose chatbot.

Did Anthropic’s launches actually cause Thomson Reuters, CrowdStrike, and Figma to drop?

Developers have their tools. Enterprises have Remy.
Developers · AI coding tools
Claude Code
Codex
Cursor
GitHub Copilot
Enterprises · built into Remy
⌂Self-hosted
✓SOC 2
✚HIPAA
☰Audit log

The timing lines up closely (a 16% drop for Thomson Reuters, 11% for CrowdStrike, and about 7% for Figma around their respective Claude launches), but stock prices move for many reasons, and public reporting has framed this as a correlation investors are watching rather than a proven direct cause.

What does Anthropic’s IPO filing say about competing with SaaS companies?

It states that Anthropic may from time to time compete with enterprise software and application companies, language that signals the company isn’t ruling out entering markets currently served by its own customers and partners.

What are open weight models and why do companies use them for this?

Open weight models are AI models whose weights can be downloaded and run on a company’s own servers instead of accessed through a provider’s API. Running them in-house keeps sensitive data and workflows off the provider’s systems, addressing concerns about proprietary processes ending up in a competitor’s product roadmap.

Is building a wrapper product on top of Claude or GPT still viable?

It’s increasingly risky if the product is just a thin interface with no proprietary data, niche focus, or deeper capability behind it, since large labs can and do replicate basic wrapper functionality as built-in features within their own ecosystems.

Editorial standards

Presented by MindStudio

No spam. Unsubscribe anytime.