AIforce Is Our Force.com Moment

I’ve been in the Salesforce ecosystem long enough to remember when “the platform” was a new idea. After Dreamforce ’26 (September 15–17 in San Francisco), I came home with the same feeling I had back then: we’re at the start of a new era.

Where we’ve been

In September 2007, Marc Benioff went on stage at Dreamforce and gave Salesforce’s platform-as-a-service a name: Force.com. The promise was that you could build any application, database or logic and run it on demand. Salesforce stopped being just a CRM and became a place to build things.

Visualforce came out at that same Dreamforce. For the first time we could design our own pages instead of accepting the standard layouts. Then Force.com Sites, announced in 2008 and generally available in June 2009, took those apps outside the company, onto public websites running on Salesforce.

Each launch removed a limit:

Force.com took away the limit on what we could build.
Visualforce took away the limit on how it looked.
Sites took away the limit on who could use it.

Where we are now: AIforce

This week Salesforce launched AIforce, and I think it removes the biggest limit so far: the interface itself.

Patrick Stokes said it best:

“The value of Salesforce has never been in the UI. It’s been in the platform that stores the way our customers encode their business.”

Everything we’ve spent years building (data models, permissions, flows, business logic) can now be used by AI agents wherever the work happens:

Claudeforce brings Salesforce into Claude, with prebuilt sales skills and a dedicated MCP server.
Slackforce lets people update records and log notes with a prompt, without leaving the conversation.
Agentforce Coworker is an AI teammate inside Lightning that respects the permissions we already set.
Headless Toolkit makes the platform available through MCPs, APIs, plugins and skills.

Why MCP is the real story

To me, MCP (Model Context Protocol, a standard way for AI tools to connect to other systems) is the Visualforce of this era, only much bigger. Visualforce let us shape one page. MCP lets us connect the whole world’s tools and data (your CRM, your data warehouse, your docs, your partners’ systems) into one intelligent interface.

For the first time, we can bring together the full picture of a customer, ask questions in plain language, and act on the answers in the same place. That means we can:

Understand customers using context that used to be spread across a dozen tabs
Drive customer outcomes by moving from insight to action in one step
Advance business goals by pointing agents at real metrics instead of dashboards nobody opens

Add Koa, the new CRM reasoning model Salesforce built with NVIDIA, and Benioff’s promise that “your data is your data. It does not go in the models,” and the base for this feels solid.

A new way to work with software

For twenty years, working with software meant learning its screens. We trained users, designed page layouts and built apps so people could find their way around.

That’s changing. Software now adapts to us. We describe the outcome, and the platform works out how to get there.

Force.com made us builders, Visualforce made us designers, and Sites made us publishers. AIforce makes us something new: people who direct what the software does. The skills this community has built (clean data, sound security, careful business logic) are exactly what these agents run on. None of that work goes to waste. It’s what makes all of this possible.

I’m excited, and I haven’t felt this way about a Salesforce launch in a long time.

The next ten or more years start now.