Dreamforce Attendees Say Last Year's AI Models Are Good Enough

At Salesforce's annual Dreamforce conference in San Francisco, keynote conversations between CEO Marc Benioff and the heads of Anthropic, OpenAI and Nvidia tackled the AI safety debate, with Nvidia's Jensen Huang urging frontier labs to "run as fast as you can."

But on the ground at the 50,000-person event, attendees said they're struggling to keep up with AI as it exists today. "It's already hard enough to keep up," said Alec Bronston, a senior Salesforce director at Chicago-based retail data company Spins, who said a slowdown would offer "a lot of opportunity to just even catch up."

The conference landed amid a dramatic moment in the AI boom. Days before kickoff, an Anthropic researcher resigned, saying top labs were "gambling with our lives," prompting Anthropic's Dario Amodei and OpenAI's Sam Altman to propose safety initiatives and push for a slower pace of development.

Salesforce customers and partners told CNBC that older, cheaper AI models are plenty powerful for everyday sales and customer service work. "The majority of agentic outcomes aren't driven by frontier capabilities," said Tim Sanders, chief innovation officer at G2. "They're driven by last year's AI."

Kevin Lee, technology chief at Nice, said his company doesn't count on high-end models for most workloads, noting that even one generation behind is "highly performant and effective."

Not everyone is holding back. Databricks released GPT-6 Astra to all 3,500 of its software developers this week, with engineering VP Patrick Wendell writing on X that it "unambiguously out performs" previous top models. Nagarro's Ram Reddy said his engineers wait about three months before adopting the latest models. "We are not one of those first early adopters," he said.