A year ago you could describe Anthropic in one line: "the company that makes Claude." One model, one chat, an API for developers. By mid-2026 that description no longer holds. Anthropic is shipping vertical AI products built for specific professions: Claude Code for developers, Cowork for office work, Claude for Teachers for educators, Claude for Financial Services for banks, Claude Science for researchers. The model underneath is shared. What they sell is the workflow of a profession.
We ship AI into businesses every week, and we track this shift for a practical reason: it changes the answer to "what should you even buy." Below we cover what Anthropic actually launched (with sources and dates), why "not a new model, but a workflow" is printing billions, and what it means for you if you're adopting AI in your own industry.
What changed: Anthropic no longer sells "one model"
The core point is simple: Anthropic stopped being a model provider and started building the operating layer for individual professions. TechCrunch puts it bluntly — the company is "increasingly betting its growth on vertical, workflow-level products rather than just raw model capability," aiming to "own the operating layer for specific industries, the way Claude Code has become the operating layer for software development" (TechCrunch).
The clearest tell is how Anthropic describes its science product. Claude Science, the company says, is "not a new AI model and not a more capable model for biology" — it's a workbench that pulls a researcher's databases and pipelines into one window (TechCrunch). The value isn't a smarter model for biology. The value is a model wired into the biologist's workflow.
"Operating layer" is the phrase to hold onto. For many teams Claude Code became less an assistant than the environment where code gets written and shipped — the whole development loop lives inside it. Anthropic wants to repeat that in every profession: to be the window where a lawyer runs a case, an analyst builds a report, a scientist runs an experiment. Own that window and you capture most of the value, even when everyone underneath runs the same base model.
Claude now lives in four modes — Chat, Cowork, Code, and Projects — plus industry bundles: Legal, Small Business, Marketing Ops. Anthropic runs vertical products across four domains at once — healthcare, financial services, legal, life sciences — with dedicated engineering teams for each. This isn't "one model for everything" anymore. It's a portfolio of apps per profession.
The break from last year is real. In 2025 Anthropic competed on benchmark numbers, and any edge went stale with a competitor's next release. In 2026 it competes on how deeply Claude sits inside a specific role's daily work — an advantage you can't copy overnight. The model became the foundation, not the product.
The lineup per profession: Code, Cowork, Teacher, Finance, Science
Here's Anthropic's vertical AI lineup as of July 2026 — who it targets and what it closes:
| Product | For whom | What it does | Launch |
|---|---|---|---|
| Claude Code | Developers | Writes, debugs, ships code; became the "operating layer" of development | GA May 2025 |
| Claude Cowork | Office roles (non-coders) | Multi-step work across files and apps: reports, checklists, spreadsheets | GA April 2026 |
| Claude for Teachers | K-12 teachers (US) | Lesson plans, standards for all 50 states, no training on chats | July 14, 2026 |
| Claude for Financial Services | Banks, analysts | Pitchbooks, KYC, models, month-end close; Excel/PowerPoint plugins | May 2026 |
| Claude Science | Scientists | Workbench over 60+ scientific databases; genomics, protein structure, chemistry | June 30, 2026 |
Every row is a real launch, not a roadmap. Anthropic gave Claude for Teachers to verified US school teachers for free, with a library of teaching skills, curricula mapped to standards in all 50 states, and a promise not to train the model on their conversations (Anthropic, Chalkbeat). Claude for Financial Services arrived with prebuilt agents for investment banking and wealth management, runs right inside Excel and Outlook, and is already in production at JPMorgan, Goldman Sachs, Citi, and Visa (Anthropic, Fortune).
Note the pace, too. In July 2026 Anthropic brought Cowork to phone and web: sessions now run remotely and keep working even after you close the laptop. On top of that came narrow bundles — Legal, Small Business, Marketing Ops. Anthropic is packaging not just the big professions but specific scenarios inside them. Eighteen months from "one model" to a fan of apps.
Why it works: where the money actually is
The move into applications isn't a nice strategy deck — it's revenue. Anthropic reported a $30 billion annual run rate after growing dozens of times over, with roughly 85% of revenue coming from businesses rather than individual subscribers (VentureBeat). That's exactly the audience that needs a workflow, not a bare model.
The sharpest example is Claude Cowork. When Anthropic released usage data, it turned out most users don't code at all: 33.4% of tasks were operations work (pulling scattered updates into one report, reconciling spreadsheets) and 16.4% were content creation and copywriting (VentureBeat). A product that grew out of an engineering tool reached a far larger audience — finance, HR, marketing, admin.
And Claude Code, which started as a developer utility, hit roughly $2.5 billion in annual run rate by February 2026 — up from $1 billion just three months earlier (Anthropic 2026 overview). A vertical product monetizes far better than plain API access "on general terms."
For us that's the headline signal. Money in AI flows to where a concrete professional job gets done end to end, not to where the most powerful model is simply available. The market votes for the workflow, and it votes with cash.
What it means for business: buy the process, not the model
The practical takeaway is one we repeat to clients constantly: competitive advantage has moved out of picking a model and into how the model is wired into your work. If Anthropic — the company with some of the best models on earth — is betting on workflow instead of "+2 more benchmark points," a business has even less reason to buy "AI in general."
We've written about this before, unpacking how AI agents are changing work: value gets created where the agent is connected to your data, tools, and rules. Anthropic's industry bundle wins not because its model is smarter, but because it knows the shape of a pitchbook or a lesson plan. The same principle holds inside a single company.
The framing flips cleanly. The old question was "which model is best" — and the answer expired every couple of months with a new release. The right question now is "what process do we wire it into, and what stays ours when the model changes." Anthropic answers that for whole industries. Inside your own company, you answer it.
There's a risk here, too: not every profession will get its own "Claude for …." Anthropic's vertical products cover large horizontal markets — code, finance, science, education. Your niche, your rules, your integrations with an internal CRM won't likely make that list. Which means the vertical layer for your function is something you'll build yourself.
How to build a vertical AI for your own function
The logic is the same as Anthropic's, just at the scale of one company: take a strong base model and wrap it in the process of a specific role. On projects we do it in four steps.
- Pick one profession, not "the whole business." An AI agent for one role's job (sales, accounting, support) pays back faster than a universal assistant for everyone at once.
- Embed it in the real tools. Just as Anthropic moved into Excel and Outlook, your agent should live in your stack — CRM, email, internal systems — not in a separate chat window.
- Load the role's data and rules. RAG over your documents and templates turns a generic model into "an employee who knows your practice."
- Don't marry one model. Design so the base model can be swapped while the accumulated process and data stay yours.
Here's a real project example. Instead of "add AI to the sales team," we start with one bottleneck — say, an agent that drafts a proposal from a CRM deal card and the message history. One role, one process, measurable hours saved each week. From there the same setup expands to neighboring tasks, but the entry point is always a single profession, not "the whole business at once." That's exactly how Anthropic started with developers and only then moved into the office, finance, and science.
A vertical AI isn't "buy a more powerful product," it's assembling the workflow of a profession on top of a model. If you want to figure out which role in your company would pay back such an agent first, look at our AI implementation services or discuss your project — we'll help you pick the starting point.
Anthropic has shown where the market is heading: from "one model" to apps per profession. The winner isn't whoever has the newest model, but whoever wires it into a specific job first.