Why Secure AI Starts With You: What Admins Must Know About Agentforce

Why Secure AI Starts With You: What Admins Must Know About Agentforce

 

Today on the Salesforce Admins Podcast, we talk to Sri Srinivasan, Senior Director of Information Security at Salesforce.

Join us as we chat about what admins need to know about Agentforce and how to build secure AI experiences.

You should subscribe for the full episode, but here are a few takeaways from our conversation with Sri Srinivasan.

Quick heads-up before we get started: This episode may include forward-looking statements—things we're excited about, but not yet available. So please make any purchasing decisions based only on products and features that are currently available. For all the legal details, visit salesforce.com.

The hidden job of AI security: admins build the brakes

Sri gave a great TDX presentation about AI security and the crucial role admins play in the future of Agentforce. “Admins are key to everything that we do,” he emphasizes, “they understand everything that's happening within their environment. They know which actions, what permissions, what they do, and agents are just another avenue to expose and interact with this crux of it.”

As Sri puts it, Agentforce is like a sports car in terms of what it can do with your data. But how fast would you drive a sports car with no brakes? That’s why admins are so important in the age of AI. We can build the brakes for Agentforce to make sure our agents are behaving correctly.

Five questions to ask when building secure Agentforce experiences

Security conversations can get very scary very quickly, but Sri boils it down to five questions admins should ask when they’re building with Agentforce:

1. What is the agent’s role and scope?

2. What data will the agent have access to?

3. Which actions should be public and which should be private?

4. Do you need to build any extra guardrails?

5. Which channels will the agent use?

The key here is practicing the principle of least privilege. And for admins, that comes down to managing permissions and profiles in Salesforce and following security best practices. Every agent runs as a user—and that user needs to be tightly scoped.

Test before you trust: scaling with the Agentforce Testing Center

Going back to the idea of brakes, Sri cautions that just because you built an agent fast doesn’t mean that it’s ready. Luckily, his team has been hard at work on new tools to help you make sure your agents are working as intended.

The new Agentforce Testing Center helps simulate and validate agent behavior at scale—without needing a QA army. You’re also able to peek under the hood to understand why an agent made a certain choice—turning debugging into decision-making clarity.

Be sure to listen to the full episode for more on what admins need to know about Agentforce. And don’t forget to subscribe to the Salesforce Admins Podcast so you never miss out.

Podcast swag

• Salesforce Admins on the Trailhead Store

Learn more