Trailhead recently rolled out the Agentblazer progress trails to help with navigating the information for Agentforce and guiding administrators and developers through AI in Salesforce.
Agentforce Agentblazer Levels
- Champion: This is the baseline of Agentforce concepts and skills.
- Innovator: taking the skills built in the champion trail and putting it into relevant use cases.
- Legend: the summary states that it will help with multi-agent systems, mentorship and designing AI systems.
Agentblazer Champion
The Agentforce Agentblazer Champion trail consists of concepts from the basics of Artificial Intelligence such as ethics, hallucinations, natural language processing and generative AI. You then actually get hands-on with prompt builder and Agentforce within Salesforce. These are guided modules and projects. So you don’t need to worry about having to write an apex class from scratch or build out a flow. However, you do walk end to end through the process of enabling, configuring and deploying an agent.
Agentblazer Innovator
This trail takes the Champion a step further and walks you through planning out an Agent, connecting Data Cloud, adding customizations for Agents, and Service/Sales Agent customizations. You still dont build out apex classes but you do work with flow directly and need to add components and customizations to your Agent project.
My Agentblazer Journey
I completed the first Champion trail and I’m currently going through the Innovator trail. Although I have been following Agentforce and Einstein AI since Salesforce started rolling out trails and modules for Einstein AI I felt that these trails really brought it all together. I felt like it fleshed out some of the details that I had been missing since starting with prompt builder. And the journey through the vacation resort put real world application behind some of the use cases that Trailhead had for the prompt builder modules but was more disconnected.
Agents out of the box?
One of the things to take note is that many of the flows, actions and apex classes are already done for you as part of the test AI playground. These are great references for your own business use cases. I’m uncertain if these come in a standard Einstein or Agentforce enabled Salesforce org. But these example components are great sources for administrators and developers to use outside of the trailhead examples.
I will also say that Agentforce and Einstein AI still seems very much dependent on these components. So I don’t believe an analyst or administrator would be able to build out an Agentforce use case from end to end with just some business requirements. In the TDX keynote the Salesforce team referenced a beta program for low-code creation of Agents using Einstein to build out an agent by scanning your org and choosing existing apex classes and flows. This of course means that your org already has existing flows/apex that will work for the agent . Realistically, this could help with building agents but in my opinion will still need advanced administrators and developers.
You can, however, build out an outline and some of the basic functionality with the flows and actions out of the box today. Which will get administrators or analysts 70% of the way to a usable AI agent in Salesforce. And I was impressed with the flows and actions that were available and they could be great starting blocks for development.