Marc Benioff says “AI is doing 30% to 50% of the work at Salesforce”, like software engineering and customer service, and its AI product has hit ~93% accuracy
Salesforce Inc. Chief Executive Officer Marc Benioff said his company has automated a significant chunk of work with AI …
Bloomberg
Context & Ripple Effects
Salesforce’s claim follows its hard pivot to Agentforce, positioning AI agents inside the company’s existing enterprise software rather than as a stand-alone tool. The company had also planned to expand its AI-agent sales force, signaling that internal deployment and external monetization were advancing together.
The reported 30%–50% automation range and roughly 93% product accuracy matter because they turn Salesforce’s agent strategy into an operating-model claim across engineering and customer service, two functions central to both product delivery and customer relationships.
First-order effects
Salesforce can shift more software-engineering and customer-service tasks to AI workflows, with human staff focused on exceptions, review, and higher-complexity work.
The reported accuracy level gives Salesforce a concrete performance claim for selling and deploying its AI product, while raising the importance of monitoring errors in customer-facing use cases.
Second-order effects
Salesforce customers evaluating embedded agents will have a stronger incentive to compare automation quality and operational savings against incumbent support and development processes.
Competing enterprise-software vendors face pressure to demonstrate not just AI features but credible adoption inside their own operations; Salesforce’s existing AI-agent sales hiring plan reinforces that go-to-market contest.
Third-order effects
If enterprise vendors can repeatedly automate meaningful shares of routine work at acceptable quality, AI agents may become a standard layer of business software rather than a premium add-on.
The durable constraint will be the cost and reliability of completed tasks: vendors that combine agent quality with distribution in established workflows are likely to have an advantage, while human oversight remains necessary where errors carry material consequences.
The trend:Enterprise AI is moving from pilots toward embedded agents that vendors use internally to validate and sell workflow automation.
A) This is an obvious lie — B) Is this why Heroku, Salesforce's cloud service, had a completely unacceptable 12 hour downtime two weeks ago? [embedded post]
If you thought Salesforce was an expensive bloated piece of shit before, get ready cause they're about to prove it can and will get worse lol [embedded post]
1. this is a flat out lie — 2. even if it was true it makes you and your company look dumber, like a useless husk of whatever it was that it did before. running a potemkin company [embedded post]
This is obviously not really true, but if I were a business that used Salesforce, just him making this claim would be grounds for me to start the process of migrating to a competitor absolutely as soon as possible [embedded post]
Meanwhile Slack is asking for full-enterprise-cost subscriptsion from the communities that built their brand like Techqueria and Rands Leadership Slack [embedded post]