Your first Agentforce use case does not need to transform the entire company.
That was one of the clearest takeaways from my Salesforce Admins Podcast conversation with Kacie Molina. It’s easy to leave an Agentforce demo thinking you need to invent one big, impressive agent. A better place to begin is with a small problem users already experience and then use Agentforce to solve it in a focused, secure, and measurable way.
In this blog, you’ll learn how to get started by finding one moment of friction, giving Agentforce one clear job, and defining an outcome you can measure. A focused starting point is easier to test, easier to explain to stakeholders, and more likely to build the confidence you need to go further.
The best Agentforce ideas rarely come from formal brainstorming sessions. They show up during support requests, user observations, stakeholder conversations, and everyday complaints.
Listen for phrases like:
These statements reveal where users are spending time searching for information, bringing details together, or interpreting data before they can act. Any one of them could be the starting point for your first Agentforce use case.
In the podcast, Kacie shares an example involving client preferences stored across related Salesforce records. Agentforce’s Summarize Record standard action, which uses a Record Summary prompt template, could be used to bring the relevant information into one concise summary. That way, instead of opening several records to understand the client, the user could ask Agentforce for the details they need. The idea doesn’t require redesigning the entire business process. It gives the agent one focused job with a clear outcome.
Pay particular attention when users copy or export Salesforce data into consumer AI tools that are not approved or governed by your organization. That behavior is a strong signal that users have an unmet need for summarization or interpretation, but it can also create security and data-handling concerns. The opportunity may be to replace that unmanaged workaround with a governed Agentforce experience that gives users the help they need while preserving Salesforce permissions, policies, and business logic.
The next step is to move from noticing those moments to documenting them. To do that, start by keeping a friction log. For at least a week, capture those moments when users struggle to find, summarize, interpret, or move information. Note what they were trying to accomplish, what slowed them down, the workaround they used, how often the problem occurs, and what a better outcome would look like. You’ll use that list in the next step to compare and stack rank your potential use cases.
Finding a frustrating process does not automatically mean you’ve found an Agentforce use case. First, determine whether Agentforce is the right tool for the problem.
Use Agentforce when the system must first understand a user’s intent, interpret unstructured information, summarize context, or determine which action is appropriate. For example, a user might ask what they should know before contacting a customer, and Agentforce can bring together the relevant preferences and history from several records.
Before you evaluate any additional criteria, you should treat Agentforce fit as a qualifying question, not simply another score. Ask yourself, “Does the use case require the system to interpret natural language, summarize information, understand context, or determine the appropriate next step?”
If the answer is no, choose another solution. A problem may be frequent, valuable, and measurable but still be better suited to Flow.
Once you’ve confirmed that Agentforce is the right fit, evaluate whether the idea is a strong first use case. Score each remaining criterion from 1 to 5. A score of 5 represents a stronger candidate.
Criteria | Question to Ask |
Data readiness | Is the necessary data accurate, current, and accessible? |
Business value | Would solving it save time, improve service, or support a better decision? |
Frequency | How often does this problem occur? |
Manageable risk | Can the idea be tested safely, with limited consequences if it fails? |
Measurable outcome | Can you clearly determine whether the experience improved? |
These criteria do not carry equal weight. Start with data readiness. Even a strong Agentforce use case can be a poor first pilot if the information it depends on is incomplete, outdated, or inconsistent. Use the remaining criteria to find an opportunity that matters to your users, is contained enough to test safely, and produces an outcome you can measure.
For example, summarizing customer preferences may score highly for frequency and business value. But it’s a poor first use case if those preferences are stored in outdated notes, duplicate records, or inconsistent fields. Agentforce would only summarize the confusion faster.
At the other end of the spectrum, an agent that rewrites a short internal status update may be low risk and easy to test. But if it saves very little time or doesn’t improve an important outcome, users may have little reason to adopt it.
Use the scorecard as a conversation starter, not a formula that makes the decision for you. First, confirm that the problem genuinely requires Agentforce. Then, look for a problem that is valuable, frequent, supported by reliable data, contained enough to text safely, and tied to an outcome you can measure.
Once you’ve selected a strong first use case, check whether a standard action already performs the job. If you need different behavior, you can create a custom action powered by a prompt template or Flow.
Action Approach | Best Used When | Example |
Standard agent action | Salesforce already provides an action that performs the job. | Summarize a record. |
Custom agent action powered by a prompt template | You need a tailored generative response grounded in business data. | Create a pre-call brief using customer history and preferences. |
Custom agent action powered by Flow | The agent needs to initiate a deterministic business process with defined criteria and steps. | Validate and update a delivery address using the approved process. |
Often, the best solution combines Agentforce with Flow. Agentforce can interpret that a customer wants to change a delivery address and collect the necessary information. A flow can then perform the approved checks, update the record, or route the request for review.
Use Agentforce instructions when the request requires interpretation. Use Flow when the next step must follow a deterministic business process.
Your first use case should solve one recurring problem for one user group and give Agentforce one clear job. You should be able to explain what Agentforce will do, why it’s the right tool, what a successful outcome looks like, and how you’ll measure the improvement.
You don’t need the perfect idea. Your first Agentforce use case is probably already hiding in a support request, spreadsheet export, or sentence beginning with “I wish Salesforce could…”
When you start building agents in Agentforce, your first instinct might be to make one agent do everything. One set of instructions, one agent to handle billing questions, delivery updates, promotions, and escalations all at once. It seems like the right thing to do. But the more you ask a single agent to do, the […]
You shipped an agent. It’s working — and now everyone wants a piece of it. “Can it also handle billing?” “What about account inquiries?” “Can it do onboarding too?” Before long, your focused agent is being asked to do everything at once, and that’s where things get messy. Multi-agent orchestration is how you fix that: […]
Every admin knows the rule: Test in a sandbox before it breaks in production. Or as the saying goes, measure twice and cut once. When new or updated solutions are released to your users, you want to be able to predict that they will have the experience you intend. This also goes for ensuring that […]
