What I like most about Salesforce Agentforce is that it moves AI beyond simply generating responses and enables agents to take meaningful actions within Salesforce. The ability to reason over trusted CRM data, understand context, execute business processes, and collaborate with users creates a much more practical form of enterprise AI.

I especially like its potential to automate repetitive work while still keeping humans involved for decisions that require judgment. Because it’s built into the Salesforce platform, Agentforce can leverage existing data, security, workflows, and business logic. That makes it easier to embed AI into real business processes, rather than treating it as a standalone tool. Review collected by and hosted on G2.com.

The biggest challenge with Salesforce Agentforce is that achieving reliable, production-ready outcomes can still take significant effort, especially around data quality, grounding, permissions, testing, and guardrails. The technology is powerful, but the gap between a compelling demo and a robust enterprise implementation can still be substantial.

I’d also like to see continued improvements in observability, debugging, cost predictability, and the overall simplicity of configuring and managing agents. As Agentforce evolves, it will be critical for enterprise adoption to make it easier for architects and administrators to understand why an agent made a particular decision or took a specific action. Review collected by and hosted on G2.com.