Benioff calls AIforce an interface revolution

Marc Benioff opened Dreamforce 2026 in San Francisco on Tuesday with a keynote about what he called an interface revolution. According to the Salesforce CEO, AI models alone cannot run a business and the software industry is not disappearing but changing shape.

The keynote at the Moscone Center followed a morning of announcements around AIforce, the layer that makes Salesforce data and workflows available in Claude, Slack and other AI environments.

No end of software

Marc Benioff, Chair and CEO of Salesforce: "We've been hearing all this kind of crazy nonsense about the SaaSpocalypse, especially over the last six months. That's been a bit wild. But the SaaSpocalypse is not about the end of software. It may be about the end of software that makes humans do all the work."

SaaSpocalypse is the term for the fear among investors that AI agents will make traditional business software redundant. Benioff used the stage to counter the idea that Salesforce is slowing down. In August the company raised the revenue forecast for fiscal year 2027 to between 46.1 and 46.4 billion dollars.

"Next year we're entering into the 50 billions in revenue."

Europe lags behind

Benioff described San Francisco, where robotaxis from Waymo are part of the street scene, as a place where AI is already visible. He contrasted that with a recent trip.

"When I was in Europe for two months, I didn't see a single autonomous vehicle. That was kind of a metaphor to me. Some countries have more AI than others. This is a very real thing. But it's also true in our customers too."

That unevenness also shows up inside companies.

"Yes, the door is opening, but not everyone has gone through it yet. In many cases, some of our customers are kind of stuck. They're kind of stuck in the old world. Maybe they're even using ChatGPT or Claude Cowork during the day at home. But when they get to their businesses and they get to their enterprises, they still don't see that enterprise transformation."

Models need a source of truth

"Models alone cannot run the enterprise."

Benioff described language models as probabilistic: they produce plausible answers that are not tied to fixed company data. He illustrated the point with an anecdote.

"I have a friend back home and they've built this amazing app on Claude Cowork. Every time he shows the app, he goes: well, this is this amazing app, but this number, I know it's wrong. I said: that's because it's not grounded in your single source of truth."

The Salesforce answer consists of four layers: data, applications with the business logic and definitions of a company, agents and the interface. The AI in the top layers remains probabilistic but draws on the fixed data and rules underneath.

No head to cut off

Earlier this year Salesforce introduced Headless 360, which makes the platform available to other systems and agents without the Salesforce screen. Benioff addressed how that was received.

"I loved it when we announced headless and everybody thought we somehow had cut the head off of our apps. There was never any head on the apps. It was always in the metadata layer."

Fields, relationships, permissions and business rules are stored as metadata. From there Salesforce builds the screen for a browser or an iPhone and now also for AI environments such as Claude Cowork, Slack and Lightning.

"It's an interface revolution. We've seen a lot of interface revolutions. We went from DOS to GUIs and we went from GUIs to web and we went from web to mobile. Today you're going to see AI interfaces that are dynamic and intelligent, that are composable and alive."

"I'm confident you will see this pattern repeat everywhere else."

Data stays with the customer

Benioff also addressed what happens to customer data in AI models. Salesforce uses the term zero data retention for the principle that models do not store customer data or train on it.

"When you're working in Salesforce apps and when you're working on our platforms, your data is your data. It does not go in the models. We've audited it. We've tested it. We've tried it. Regardless of what you've heard in the media or from other vendors, let me assure you that Salesforce products are built with zero data retention, tested over and over again by our security teams. You're not training any other model when you're using our products."

Salesforce as test case

Benioff presented Salesforce as the first user of the Salesforce agents. Piper qualifies prospects on salesforce.com, Hunter looks for new sales opportunities.

"Hunter went out and generated 500 million dollars of pipeline for me last quarter. Casey answered 5 million customer service issues for me so far, since I launched just about 18 months ago."

On 10 September Salesforce completed the acquisition of Fin, formerly Intercom, for about 3.6 billion dollars. The Fin customer service agent now also runs on help.salesforce.com.

"We got that running, I think, in about 12 days."

Slack and one bundle

"It's become our very fastest growing product. So many AI companies here in the Valley, like within a mile or two of this building, all run on Slack."

"By the time we get back a year from now, we'll have dozens of live interfaces running on Salesforce."

Benioff also presented AIforce Max as a single product that combines all functionality. Earlier this month Salesforce introduced Core, Advanced and Max editions for Agentforce Sales, Agentforce Service and Agentforce Industries, which bundle AI, Slack and Tableau Next in one purchase. Existing Agentforce 1 Edition customers can move to Max at no extra cost.

"One product, AIforce Max, with all of our functionality built together."

What agencies should note

For agencies that build AI tools, dashboards or reports for clients, the anecdote about the wrong number is the practical point: output generated by a model is only as reliable as the data it is connected to. Salesforce positions the CRM as that source. The screen itself becomes something that is generated on request, in Claude, in Slack or in Salesforce itself. Salesforce will also release an SDK with which companies can build their own AIforce applications.

Benioff handed over to Patrick Stokes, President, Applications and Marketing at Salesforce, for the product demonstration. Dario Amodei of Anthropic and Sam Altman of OpenAI speak later in the programme.

 Follow Marketing Report on LinkedIn

 Subscribe to the Marketing Report Newsletter

 Send us your press releases, columns, interviews…

 Join Marketing Report in our LinkedIn Groups