If the idea of Salesforce Multitenancy is important to understanding Salesforce as a platform, then the logical next step is to consider its use of metadata. Metadata is fairly ubiquitous in the internet age. We all know it’s important to tag data and content with information and descriptions so that browsers, search engines, and, these days, agents can properly comb the internet and understand it.
But what is its role when it comes to a complex system designed to allow multiple administrators to create unique experiences for a company full of users? The simple answer is that it’s the very backbone. And while most of the magic of metadata happens behind the scenes, understanding its basics and the role it plays can be helpful for the solutions you as an admin provide your users, especially when it comes to Agentforce.
Think about your Salesforce implementation like it’s your own luxury apartment. Let’s say you’ve just bought the apartment next door and you want to knock down the wall between them, clear out its existing rooms, and expand your home to fit your new needs and guests.
When you hire your contractor, you aren’t going to just wave your hands at the newly demolished wall. You’ll have them measure the floorspace, note things like electrical outlets and plumbing, and then, before building on anything, agree upon a detailed blueprint to know precisely how the new expansion will work.
Now with Salesforce, that blueprint is being automatically created for you whenever you customize anything on the platform. Add a new object? Blueprint. New field on the object? Blueprint. Add a flow building on top of that object? Blueprint. There are a lot of advantages to having an ever-changing blueprint that automatically expands with your customizations. One of the most apparent benefits is packaging. That blueprint layer allows you to take a subsection of your implementation and share the blueprint for that section to move or share it elsewhere.
Metadata is essential to how Agentforce works. Context is everything to AI and it comes in two ways—the original model that the AI is trained on, and the adjacent media that it has access to read. Without metadata, the large language model (LLM) isn’t able to understand any of your customizations to Salesforce. With the universal nature of metadata, it doesn’t matter if we’re talking about custom fields or custom flows—Agentforce has the capacity to understand what those customizations look like. If we go back to the robot from the Multitenant Analogy, ask questions about how that robot works in a highly customized environment. If you ask the robot to get you a fork, how does it know which drawer in which room to find that fork? It’s because it can load up a blueprint at any time that will give it all the details of all the customizations that led to someone putting that fork in the drawer in the first place.
An interesting fact about that blueprint is that the Description field is super powerful in the age of Agentforce. All admins know that a good Description field is helpful when their human users need clarity on what a certain object or field might be for. It’s even more important to Agentforce, as it provides easy-to-find context for it to make decisions.
Remember that Agentforce’s planner has a library of actions it can take, and the more context it has, the better the chances it will choose and utilize the right one the right way. Consider all these places you can add a simple layer of text in Salesforce: custom fields, custom objects, formulas, custom agent actions, etc. The descriptions can follow the best practice you already know; they don’t have to be a block of text, just enough to be descriptive to humans and AI alike.
On top of all the normal descriptions you might use, remember the power of the agent’s own topic. AI relies on human text as a guiding tool for how it should behave, and the topic is your chance to control the outcome of prompts by giving it clear instructions in plain text—no code or other tools required. Examples of things you can control in the topic include the kind of tone the AI should take, the basics of how the responses should be formed, and what to do in edge cases.
Now you can use metadata as your secret weapon to supercharge your agents. Want to learn more? Sign up for an Agentforce NOW workshop.
Internal teams often reach out to admins with the same types of questions, like how to reset a password, where to check their paid time off balance, or how to find a policy. These requests may be simple, but they take time and often build up during the day. Agentforce can help by automatically handling […]
As someone completely new to the Salesforce world, I know exactly how it feels to be faced with a powerful new technology like Agentforce. When I was asked to explore what it means to build an artificial intelligence (AI) agent, my first reaction was to look over my shoulder to see who they were actually […]
When you go to login.salesforce.com, do you consider what happens technically once you’ve logged in? In many ways, you don’t need to because Salesforce thinks about it for you. But if you want to better understand how Salesforce empowers your enterprise applications, there’s no better place to start than its concept of tenancy. Think of […]
