AI agents as helpers, not replacements • The Register
Salesforce CEO Marc Benioff heralded the arrival of the agentic era during his keynote at the CRM giant’s annual Dreamforce conference.
Speaking in San Francisco, California, on Tuesday, without the protection of the National Guard troops he had invited to police the city, Benioff said, “The agentic revolution, there’s no question, it is here.” He promised attendees they can see it for themselves at Agentforce City, the exhibition area to show off the company’s Agentforce AI platform.
The agentic era refers to the way that AI vendors apply wishful thinking to this particular moment in time. Agents are AI models given access to software tools to pursue some goal with limited supervision. These agents fail a lot – 70 percent of the time by one recent measure. Even so, Salesforce, among others, remains convinced it can sell AI agents to enterprises so they can automate business processes, enhance productivity, and perhaps replace human workers.
That is, other companies have touted AI as a way to replace human workers – a dubious claim, according to Yale researchers. Salesforce, having perhaps perceived that people want to keep their jobs, is now saying, “AI doesn’t replace people, it elevates them.”
Benioff said Salesforce is saving $100 million annually using AI-powered customer service. And as further proof-of-life, the biz highlighted how Williams Sonoma has committed to use its Agentforce 360 Platform to build AI agents that will provide online customer support. The retailer expects that it “will be able to autonomously resolve more than 60 percent of chat inquiries.”
Put another way, that’s a 40 percent failure rate. While that may be an acceptable figure when an organization has human failover in place to resolve issues that bots cannot, such calculations need to be considered. Salesforce on Monday touched on this issue in a post titled, “How Much Free Will Should Your AI Agents Have?”
We note that AI agents – as software applications – are not alive and cannot have free will. The question being asked might be better worded, “how much risk from unexpected behavior can your business tolerate?”
As Salesforce observes, “Unchecked autonomy can be risky. One rogue response from an agent can draw regulatory scrutiny or cause reputational harm to the brand. But overcontrol can strangle innovation, reducing agents to glorified chatbots.”
Such concerns are evident in MIT’s recent study that only 5 percent of enterprises have successfully integrated AI tools into production environments. Benioff acknowledged those findings during his keynote.
“And we all saw the MIT study. We all saw just making our own models, or rolling it ourselves, or DIYing it, that isn’t going to do it, is it?” he said, arguing that Salesforce can help companies get to the promised AI land. “That in our vision is the Agentic Enterprise.”
Central to that vision is Agentforce 360, the company’s AI customer relationship management (CRM) platform, which is now generally available. It consists of AI agent building tools, the Data 360 layer to provide data persistence and context to models, Customer 360 apps, and the company’s Slack messaging service.
Among 12,000 claimed Agentforce 360 customers, Salesforce pointed to Reddit as an example of the potential savings to be had by having AI chatbots handle customer service interactions with its advertising clients.
Reddit is said to have “deflected 46 percent of support cases and cut resolution times by 84 percent, reducing average response time from 8.9 minutes to 1.4 minutes” by using Salesforce’s platform. According to John Thompson, VP of sales strategy and operations at the social media site, the improved efficiency increased advertiser satisfaction by 20 percent and allowed human sales reps to handle more complex questions.
Salesforce also celebrated various partnerships amid its Dreamforce news dump. Anthropic and Salesforce expanded a prior partnership focused on tuning Claude for the needs of regulated industries like financial services. OpenAI did a similar deal, expanding an existing relationship to make Agentforce 360 accessible from within ChatGPT while making OpenAI’s latest models (e.g. GPT-5) available within Agentforce 360.
Also, Salesforce and Stripe agreed to collaborate to build an Instant Checkout integration using the new Agentic Commerce Protocol (ACP). In theory this will help merchants using Agentforce Commerce support customers who want to use AI agents or bots to make programmatic purchases.
There were also more personal moments, as is often the case at Dreamforce. Benioff, having walked back his call for troops even as Elon Musk has echoed that call for a clampdown, asked the audience to put aside political differences and come together to listen to a video about technology and sustainability narrated by his recently deceased friend and prior attendee Jane Goodall.
Ironically, Goodall herself declined to be apolitical in the recent Netflix documentary Famous Last Words – she contemplated launching President Trump, Elon Musk, and assorted autocratic leaders on a one-way trip into space ®.
