My First Experience with Agentforce: What I Learned and What I Wish I Knew

‍

It’s been over ten years since I last worked with Salesforce as a developer. In that time, Salesforce has introduced a lot of new technology designed to reduce friction for developers, making it easier than ever to build on the platform. So when I first dove into Agentforce, I wasn’t sure what to expect. What I discovered was a mix of exciting new tools, a few unexpected hurdles, and a realization that Agentforce fits comfortably within the existing LLM ecosystem. 

‍

Getting Started: Scratch Orgs to the Rescue

‍

One of the biggest improvements Salesforce has made over the years is the introduction of Scratch Orgs. These lightweight, disposable environments made it incredibly easy for me and another developer on my team to create fresh Salesforce instances with all the necessary configurations turned on. This was a game-changer—it meant we could spin up new environments instantly and start testing Agentforce without having to deal with the complexities of setting up a traditional Salesforce org from scratch.

Developer Workflow: Iteration Made Easy

‍

I was able to do all of this comfortably from VS Code with the SFDX extensions and using the SFDX CLI tool from Salesforce. The developer experience felt polished and intuitive, making it easy to navigate the platform and get up and running without unnecessary roadblocks.

‍

That said, it wasn’t all smooth sailing. Even with Agentforce installed on a brand-new Scratch Org, I quickly realized there were additional steps required to get everything up and running properly. For instance, I needed Data Cloud enabled, and the Salesforce team helped get my Developer Hub Org set up with that capability. But even after that, I had to manually enable Data Cloud in each Scratch Org and assign the correct permissions to my user profile before I could access the Agentforce agent.

‍

Once that was all figured out though I could quickly write a script to automate this process for my fellow Wayfounders 

‍

sfdx org:assign:permset -n CopilotSalesforceAdmin CopilotSalesforceUser -b wayfound1

‍

Once I worked through those initial setup steps, everything clicked into place. The process became seamless, and I was able to quickly build my own agent using the Agentforce GUI. This is a great option for admins, offering an intuitive way to create and configure agents. But for developers who need deeper customization, Agentforce also provides powerful APIs that allow for more complex integrations.

Exploring Agentforce: The Power of the GUI and Data Cloud

One of the most important things I learned was the value of using the Agentforce GUI to get a quick understanding of all the different features at my disposal. The GUI provided an intuitive way to explore Agentforce’s capabilities before writing any code. I could configure agents, set up tools, and test interactions in a way that was fast and visual.

Once I had my agents up and running, I was able to dive into the platform’s features. I could view conversation history, see the different tools set up for each agent, and even use a testing harness to create automated tests. This was one of the most exciting discoveries—being able to automate testing for these agents added a whole new level of confidence in their behavior and performance.

Beyond just building agents, viewing where data is ending up in Data Cloud was another key insight. By analyzing the data stored in Data Cloud, I gained a deeper understanding of how Agentforce operates under the hood. It gave me a clear picture of how user interactions were processed, how different tools were leveraged, and what kind of data was being captured. This was crucial because it provided a broader sense of how Agentforce functions as part of the Salesforce ecosystem.

What struck me most was how familiar Agentforce felt compared to other LLM-powered tools. The conversational flow, the tool calling, and feedback, mirrored what I had already seen with platforms like OpenAI, Anthropic, etc. The API was intuitive and structured similarly, making it easy to integrate into our existing Wayfound system. The concept of using tools and actions to connect to external services or internal data sources was also familiar, reinforcing the idea that Agentforce is built on the same core principles that define today’s leading AI agent platforms.

Final Thoughts: A Smooth Transition into the Salesforce AI Ecosystem

‍

In the end, my experience with Agentforce was a rewarding one. While there were a few initial hurdles in configuring Data Cloud and permissions, once those were handled, I was impressed by how seamlessly everything came together. The platform’s familiar structure, powerful APIs, and frictionless developer experience made integrating it into our workflows at Wayfound a natural fit.

‍

If you’re just getting started with Agentforce, my biggest piece of advice would be to take advantage of Scratch Orgs—they will save you time and effort. Also, ensure that you have all necessary permissions and features enabled upfront to avoid unnecessary troubleshooting down the line.

‍

I’d love to hear your experiences with Agentforce as well! If you’re attending TrailblazerDX (TDX) this week, come find me—I’d be happy to share insights and learn from your journey as well.

‍