Artificial Intelligence

Is Salesforce Giving Anthropic Too Much Power?

Thomas Morgan

By Thomas Morgan

Back in 1999, Oracle CEO Larry Ellison backed Marc Benioff when Salesforce was just an idea. Ellison reportedly put in $2M, joined the board, and supported Benioff while he built a company that would eventually help shift enterprise software away from the old on-premises model Oracle had grown up with. Salesforce didn’t destroy Oracle, of course, but it did help define the cloud era in a way Oracle initially found very hard to match.

Fast forward 27 years and Benioff now finds himself on the other side of a very similar-looking relationship. At Dreamforce ‘26, Anthropic CEO Dario Amodei was on stage with him following the recent announcement of Claudeforce, while Salesforce spent the majority of the week showing off a much deeper integration with Claude. Salesforce presented that they want its data, workflows, and business logic to be available wherever people choose to work, including inside Claude itself.

And this is where we start to see some parallels. Benioff knows better than most what can happen when a new layer of technology changes how people consume enterprise software. The question, now, is whether he’s helping Anthropic build the next one, or whether Salesforce has learned enough from its own history to make sure Claude still needs it underneath.

Dreamforce Showed What Salesforce Is Actually Betting On

As expected, Dreamforce was packed with several announcements and demos to truly demonstrate their new long-term message – you don’t have to interact with the traditional Salesforce UI to use Salesforce anymore.

The newly named AIforce took center stage, presented as the way for Salesforce to “come to you”, whether that be in Claude, Slack, Lightning, or wherever people really prefer to work. You can even update records, trigger workflows, and query CRM data, all without needing to open Salesforce at all.

Patrick Stokes talked attendees through this idea during the Claudeforce demo, arguing that Salesforce’s real product is not simply its interface, but the trust customers place in it to hold their data, workflows, and business processes. So what is Salesforce actually trying to protect, and why is it so willing to separate the two?

If someone can begin their day in Claude, ask it what is happening with an account, or update an opportunity without really having to think about Salesforce, then Claude basically has full ownership of the actual user experience. On paper, this does feel like a risk – the more people use Claude, the less important the Salesforce interface starts to feel.

But Salesforce seems to be betting that the interface can change completely, provided the important stuff underneath it does not. Claude can become the place where the user asks the question, but Salesforce still holds all the important stuff that makes it irreplaceable.

In other words, Salesforce may be willing to become less visible if it means becoming harder to remove.

READ MORE: Salesforce’s First CRM Reasoning Model ‘Koa’ Is Revealed at Dreamforce ‘26

Has Benioff Learned What Ellison Didn’t?

To put it bluntly, this is sort of where the Oracle comparison starts to break down. The easy version of the story is that Larry Ellison helped fund Salesforce, Salesforce went on to challenge the model Oracle had built its business around, and now Benioff could be doing exactly the same thing by getting so close to Anthropic. But the big difference is that Benioff is not ignoring the changes around him, but encouraging his company to move toward it as fast as it can.

This, as many of our readers know, has not always been smooth. Agentforce has had its fair share of criticism, and there is still plenty of pressure on Salesforce to prove customers are getting enough value/ROI from the AI products it has spent so much time promoting.

But the response has not been to retreat back behind the walls of its own platform. Salesforce has gone the other way and pivoted when necessary, now bringing Claude closer and accepting that its own technology may not need to (or not be able to?) own every part of the experience.

Ultimately, this may be the lesson Benioff took from the cloud era – you do not necessarily have to build the thing that replaces your interface, just make sure whatever replaces it still needs you.

And as impressive as Claude may be, it still definitely needs Salesforce underneath it – the access to the customer data, workflows, business logic, and so on, that companies have spent years building inside Salesforce’s platform.

READ MORE: Comparing Salesforce’s AIforce, Headless 360, and the Enterprise AI Harness

This then also raises questions around the idea that AI could ever really simply recreate products like Salesforce from scratch. You can vibe-code something that looks like a CRM fairly quickly, sure, but recreating years of company data, integrations, and processes is another battle entirely, and not one Anthropic seems ready to risk.

So perhaps Salesforce is not handing Claude too much power after all, but giving Claude just enough access to make sure Salesforce remains part of whatever comes next.

Final Thoughts 

Does anyone really think Larry Ellison knew exactly what Salesforce would become when he backed Marc Benioff? Probably not, but that is what makes the comparison useful. Big technology shifts are obvious in hindsight and much harder to read while they are actually happening.

Benioff at least seems determined not to stand still – Salesforce is giving Claude more room, but it is also trying to make sure the data, workflows, security, and business logic underneath still belong to Salesforce.

READ MORE: Marc Benioff Weighs In on AI Slowdown: “They Should Be Held Accountable”

The Author

Thomas Morgan

Thomas Morgan

Thomas is a Content Editor & Journalist at Salesforce Ben.

Leave a Reply