“The only thing we’re going to do at Salesforce is Agentforce.” That’s what Salesforce CEO Marc Benioff said in September when introducing the AI agents. This week, the company showed what that means by unveiling a substantially upgraded Agentforce 2.0.
Among its new features, four stand out:
Here’s a look at them and how they can be used.
This library spans key business applications like CRM, Slack and Tableau, and integrations from AppExchange partners. These ready-to-use skills simplify the process of building tailored AI agents. It eliminates the complex, time-consuming integration issues. The library has pre-built Slack Actions like Create Canvas and Message Channel, which let teams create agents for their Slack workflows.
Here are some examples:
With this, Slack becomes the central point for employees to interact with Agentforce. They can access information, trigger workflows and manage tasks directly within Slack. That means they don’t have to jump between applications.
Agentforce can also automate tasks within Slack. For example, Accenture uses it to give employees quick updates on important accounts. It helps automate document retrieval and makes recommendations for solving business problems. It also streamlines proposal creation and visualization and enhance employee collaboration, including providing real-time updates, retrieving documents, offering proactive recommendations and creating proposals.
Dig deeper: SharkNinja embarks on its Salesforce AI journey
Agentforce also uses Slack Enterprise Search. It learns from conversations in Slack channels and direct messages. This lets agents provide more relevant responses and actions based on the context of conversations.
The Atlas Reasoning Engine has been improved with advanced reasoning and retrieval augmented generation (RAG) so agents can handle more complex, multistep questions and provide more accurate and precise answers.
For example, RBC uses Atlas and Data Cloud with Agentforce to help its financial advisors. Data Cloud integrates all client data into a unified system, giving advisors a comprehensive view of their clients. Atlas can answer complex client questions. It does this by using multistep reasoning. It also pulls from RBC’s data and business logic. This helps deliver a reliable answer.
There’s an old Russian saying, “Trust, but verify.” The enhanced RAG capabilities let users do that by giving inline citations with agent responses. Citing the exact sources used builds trust in the AI’s output. It also helps users check for accuracy.
The Testing Center lets businesses test their AI agents before they are used. This ensures the agents follow instructions, stay factual and work quickly. This helps mitigate risks and ensures agents perform consistently and reliably. It also provides tools that make it easier to test, audit and scale agents.
Salesforce is using the technology internally. It says agents at help.salesforce.com now handle 83% of customer support queries independently, with human escalations dropping by 50% since implementation two weeks ago.
“Suddenly, as a CEO, I’m not just managing human beings, but I’m also managing agents,” said Benioff. “There is an authentic agentic layer around the platform today. It’s not some vision fantasy in the future idea, it’s what is happening right now.”