Salesforce banking on autonomous agents with Agentforce 2dx

Salesforce released the next version of its Agentforce platform Tuesday, along with a passel of developer tools to enable generative AI agent creation.

Most of the features are available now or will be later this month. A few from the upcoming fall release are in preview from Salesforce.

In Agentforce 2dx, bots can now proactively act on triggers that invoke agents to perform work -- a key to enabling users to build autonomous agents. Through the Agentforce API and connections to automation toolboxes such as Salesforce Flow and Apex orchestration tools, customers can now plug agents into business logic rules that allow Agentforce agents to make decisions on low-level tasks. For example, this could include making refunds in customer service, screening resumes for HR and automating contracting in healthcare settings, said Adam Evans, executive vice president and general manager of AI at Salesforce.

The demand for agentic AI is strong, said Dion Hinchcliffe, an analyst at Futurum Group. However, his company's surveys have found that customers don't want to make their own agents and instead are looking for integrated platforms.

In CX, the demand revolves around automating customer service, marketing and sales. But creating agents from scratch requires more attention to security, data access and governance than many organizations want to do themselves, Hincliffe said.

"[Also, a] concern is that generative AI is a little bit probabilistic compared to the ways that we used to automate things," he said. "[CIOs] really want dependability. They want something that's proven and working, and that's another reason why they're not necessarily rolling their own -- so they don't have to figure out how to do all of that. They're looking for partners that already have those issues resolved."

CRM consultant Brent Leary said that while many vendors, such as Oracle and Microsoft, have released autonomous agents, Salesforce has somehow grabbed all the mindshare.

The technology as a whole has been moving at lightning speed since the debut of ChatGPT generative AI two and a half years ago, he said. Technology buyers need to carefully analyze how agent implementations will work for their particular organizations before buying into the latest and greatest.

Autonomous agents have both good and bad consequences for enterprises, Leary explained. On the good side, companies can create workflows and have the AI not only use these workflows but also improve them.

"That's good, but it also could be a potential net negative if your organization isn't really ready to let that happen -- or doesn't know the ramifications of autonomous agents," he said. "It can get away from you quickly."