Deploy the Agent to Production
Learning Objectives
After completing this unit, you’ll be able to:
- List the prerequisites for deploying an agent.
- Move an agent from a sandbox org to a staging or production org.
Almost Live
Warren has connected the agent to chat and voice, but everything so far lives in a testing environment. Two things stand between Northern Trail Outfitters (NTO) and its Agentforce go-live: proving the agent works end to end, and moving it safely to the production org where real customers can use it.
Validate Your Agent
When you finish configuring your deployment in your sandbox environment, test to make sure that the agent performs as expected in each channel. If you’re using adaptive response formats, validate how they look in a test channel.
Check out the Agentforce Testing Tools and Strategies badge for an overview of the testing process. And see the Resources section for even more information about testing.
Before You Deploy
Warren runs through a prerequisites checklist before starting.
- A built and tested agent: Deployment assumes the agent already works.
- An agent user: A service agent runs as a dedicated agent user: a Salesforce integration user whose profile and permissions govern what the agent can access on customer channels. Warren confirms this user exists and has the right permissions.
- Channel readiness: Each channel has its own prerequisites (for example, a website needs a place to embed the chat).
Deploy to Staging or Production
When you’re sure that your AI agent is ready for prime time, you can package the agent and deploy it to your target org, whether that’s a staging or production environment.
Here are the basic steps.
- Retrieve the latest metadata from your sandbox org using Salesforce CLI and generate a package file for your agent. Include any custom Apex, flows, and prompt templates for your agent actions. Also include any permission sets assigned to the agent user.
- In your target org, verify you have the necessary licenses for your agent.
- Verify that Einstein, Agentforce, and Data 360 are enabled in the target org.
- Create the agent user in the target org.
- If you’re using Enhanced Chat, Embedded Service Deployments, or Omni-Channel Flows, deploy that functionality to the target org.
- Deploy the Agentforce metadata to the target org using Salesforce CLI.
- Verify that the agent is active.
- Add the deployed permission sets to the agent user.
- Make sure all deployed flows have the correct version active.
- If your agent uses Embedded Service Deployments, publish the web deployment in the target org.
After deploying to his target org, Warren checks that all of the subagents, instructions, variables, and filters look right. He also verifies that all permissions have been deployed or updated appropriately, and that the agent is responding to user requests. The NTO service agent is officially live!
Up Next: Monitoring and Improvement
Congratulations! You’ve learned the basics of deploying an AI service agent that can assist your customers 24/7. But how do you know if your newly deployed agent is really doing what you intended? Check out the Agent Analytics and Monitoring badge to dive into the monitoring stage of the agent development lifecycle and find out how to measure and optimize your agent’s performance.