Salesforce and Anthropic Team Up So You Never Have to Open Salesforce Again

The other day, I came across a piece of news and stared at it for a good while.

Salesforce announced it is expanding its partnership with Anthropic and launching something called Claudeforce. Its first product is called Salesforce in Claude.

In plain terms: putting Salesforce inside Claude.

From now on, looking up customers, editing records, checking opportunities, running processes — all of it happens in a chat box.

No need to open Salesforce.

A company that has spent twenty-plus years building enterprise software has, with its own hands, turned its most important door into an optional extra.

Why is this worth talking about? Because the ambition here is far bigger than "yet another AI integration."

What Exactly Is Claudeforce?

Start with the product itself.

Salesforce in Claude is aimed mainly at sales for now. A salesperson can simply tell Claude: get me ready for tomorrow's client meeting, take a look at my pipeline, how healthy is this deal, update the follow-up notes.

No logging in, no clicking through menus. Just talk.

It ships with 37 preset sales skills at launch. Right now it is open only to a small group of pilot customers, with a public beta starting in September. In the months after that, marketing, customer service, and e-commerce capabilities will follow one after another.

You might ask: how is this any different from bolting AI onto an ordinary CRM?

The difference is in the motive.

The Facade Can Come Down. The Foundation Stays.

Patrick Stokes, president of applications and marketing at Salesforce, already said it outright. The gist of his words: Salesforce's value doesn't live in the UI, not in the app itself — it lives in the data and the metadata.

What are data and metadata?

It's who your customers are, where an order stands, who can see which records, and what the process looks like after an order is placed. All of this normally hides behind the interface, where nobody sees it.

But that's the family fortune.

What is the interface? A facade. A facade can be renovated, relocated, even done away with entirely.

The foundation stays.

Back in March, Salesforce started paving the way with Headless 360, letting AI agents bypass the interface and get at data and processes directly through APIs and MCP servers (Model Context Protocol, a standard that lets AI agents connect to external tools and data). The thing is, that earlier toolkit was built for developers. What Claudeforce does is package it into a product ordinary business users can actually use.

So here's the question: if AI gets hold of all your data, what's to stop it from going rogue?

Salesforce thought of that. An administrator connects the two sides once. From then on, permissions carry over exactly as they are.

Records you can't see in Salesforce, Claude can't see either.

Orders you can't change, Claude can't change either.

Every action still executes inside Salesforce, and every business rule still applies.

AI puts on a new skin; not a single rule comes off.

There's another demo I've watched over and over. Someone describes to Claude what they want to see, and Claude builds a custom dashboard on the spot, using data from Salesforce.

Not picking one from a list of preset reports — building one, right there.

Of course, Salesforce is careful to point out that this demo isn't an official feature yet. But the direction it points is laid out in plain sight.

AI Is Becoming the Interface

What makes Claudeforce truly worth chewing on is the shift it points to: AI is becoming the interface of enterprise software.

Think about what a marketer's day actually looks like. Ping-ponging between a dozen-plus systems: hunting for data, configuring workflows, pulling reports, executing actions. By the end of the day, the window-switching alone is exhausting.

What if all of it could be handled with one sentence spoken to an AI?

But note: platforms won't disappear. They just switch roles. Data, permissions, APIs, business logic — the things AI needs to do its work — matter more now, not less.

Here's an analogy. How long has it been since you last walked into a bank branch? And yet your money sits in that bank, every cent of it.

The branches have gone quiet; the vault is still there.

What Salesforce wants is to turn itself from branch into vault.

CEO Marc Benioff has been anything but polite about the fashionable "SaaSpocalypse." In one interview he called it pure nonsense. Frontier models depend on CRM — they don't replace it. Salesforce is, first of all, a data company.

The more capable AI gets, the more it needs the well that feeds it data.

By the way, Salesforce is also inviting Claude further inside its own house: Claude will become the reasoning model for Agentforce, drive Agentforce Vibes and Agentforce Coworker by default, and serve as the default model for Slack.

How do these two entry points divide the work?

Agentforce is machines working on their own — automated execution, facing the customer directly. Salesforce in Claude is humans working — employees' daily jobs, with AI as the assistant.

On one side, the employee's day-to-day; on the other, the machine's automation. Two lines that converge in the same place: Salesforce's data.

Charging by the Seat, or Charging by the Work?

One more thing, and if you're in marketing operations, pay extra attention: how the money works.

Claudeforce runs on Salesforce's headless consumption model — how much API capacity you get depends on your edition. Claude's own usage is settled separately, with Anthropic.

Behind this sits a completely different business logic.

How does traditional SaaS charge? By the seat. A 50-person team buys 50 licenses.

But what if the ones doing the work change from people to AI agents?

An agent doesn't sleep and doesn't take vacations. It pulls data, runs processes, and calls APIs 24/7.

Do the math: how many clicks does a 50-person team manage in a CRM over a year? How many API calls does a tireless agent fire off in a single day?

Sooner or later, per-head pricing becomes more and more awkward, and usage-based pricing is simply where things land.

For marketing ops teams, this will rewrite how you look at cost: the system nobody logs into may send the biggest bill.

Because the AI in its back office never stops working.

A Closing Word

For now, Claudeforce starts with the sales use case, and its impact on marketing largely still lives in the imagination.

But Salesforce's bet is already out in the open:

CRM doesn't need to be where you work — but it can remain the system your work depends on.

Dig one level deeper: a piece of software's value will increasingly hinge on what AI can safely accomplish with it, not on how long you sit inside it each day.

It may not be long before you genuinely go an entire day without opening Salesforce.

And it ends up making more money than ever.

Here's wishing you a business whose foundation sits where no one can tear it away.