Your Guide to Agentforce: Basic Info, Use Cases, and ROI Tips

Author: Janet Elliott

We gave away some of the details with this blog’s title, but there is a lot more going on with Agentforce, aside from it being an AI tool that works with your Salesforce platform.

Agentforce is an ever-evolving topic (and product) that we’re uncovering more about on a daily basis, thanks to the Kicksaw team’s enthusiasm for keeping up-to-date about all things solution-related for clients, and their firsthand experience with Agentforce implementations. 

We’ll be using that insight to distill Agentforce information down into a guide that will help you:

• Understand the basics about Agentforce and why it’s important
• Build up your knowledge about what it does/how it works, along with how it can be used (including some real-world use case examples)
• Maximize ROI if you’re looking to invest in Agentforce for your organization

‍

‍

Agentforce 101: Understanding the Basics

In the most basic of terms, it’s a low-code AI tool for the Salesforce platform that allows organizations to build customized AI assistants. 

These assistants are what Salesforce calls ‘agents’, and they support employees and customers by using data, configuration, and set guidelines to answer questions and complete tasks using NLP (Natural Language Processing). That means you can interact with an AI agent by just writing out your question or request. 

If you’re thinking, “That describes a lot of AI tools, what makes Agentforce special?” — the answer is its autonomous capabilities and the Salesforce trust layer:

• Although there’s been a lot going on with AI in recent years, advancement has accelerated to the point where AI’s involvement in the workplace can now further expedite workflows
• For Agentforce, this is due to its ability to make decisions and act on them, within set parameters
• Meaning it's able to assist with tasks/answer questions, and complete the likely follow-up actions using data, set criteria, and set parameters for an extremely wide range of roles, industries, use cases, etc.
• It’s also able to function across a variety of channels (e.g., an organization’s website, CRM, mobile app, Slack, etc.)
• And the trust layer allows for zero data retention, prompt defense, data masking, toxicity scoring, and audit capabilities (meaning your data and customer information if safe and protected)

We mentioned above that AI has been quickly advancing — Agentforce is a clear example of this in terms of both capabilities and value for users. If you’ve seen any news about Agentforce 2.0, it's referring to Agentforce’s expanded suitability for roles across an entire organization. With this development, Salesforce is able to position Agentforce as digital labor, meaning it's probable you’ll find it useful if you’re:

• Facing labor shortages
• Want to increase the quality of service your organization can provide, without increasing the burden on your existing employees
• Trying to ensure the employees you have can focus on personalized interactions and/or high-priority tasks
• Looking into burnout mitigation and ways to increase efficiency while minimizing turnover
• Planning ahead for upcoming gaps in your workforce due to retirement, transitions in operations, etc.

‍

Agentforce: What Makes It Different and How It Works

To understand what makes Agentforce different from other AI tools on the market, it helps to look at its predecessors. If you’re familiar with basic chat bots, you know they take you down a predefined path based on your responses. 

• On the backend, this requires extensive mapping of every possible path
• Which means tons of work for your team to set up and a very limited and rigid experience for your customers

Next, we had tools like Salesforce’s Einstein Copilot, which delivered a much better conversational experience, but still had limitations when it came to follow-up inquiries and answering questions.

With Agentforce, technology has progressed to the point where both the user experience and backend functionality are more fluid (both major upgrades). You define a scope and the guardrails for an agent, and topic and related actions it can leverage (e.g., flow, Apex, etc.). From there, the Atlas Reasoning Engine takes care of determining the best course of action within those parameters. How it works is tied directly to what makes it different, because it follows a sophisticated reasoning process:

• Query Evaluation - assessing and refining user queries for clarity
• Data Retrieval - retrieves the most pertinent data from various sources
• Plan Building - using the retrieved data, Atlas constructs an execution plan
• Plan Refinement - further process refinement to ensure accuracy, relevance, and grounding in trusted data
• Autonomous Execution - allows Agentforce to autonomously reason, make decisions, and complete business tasks

‍

‍

From a user perspective this results in the agent being able to ‘roll with the punches’. The user can get directly to the point, typing out their question or request (without having to select the appropriate category) — the agent doesn’t follow a predetermined path, it identifies the question or request within its parameters and takes the appropriate action (which is built during set up/configuration).

This could mean answering a question based on knowledge base articles, summarizing a record, scheduling an appointment, etc. 

For example, say a customer or patient is planning out their day and realizes a later reservation or appointment would better suit their schedule. They go to the company’s (provider’s, patient’s, etc.) authenticated portal and ask the agent (which looks like a website chatbot) if there are any times available later in the day:

• The agent looks at the data, and uses that data, along with its configuration and set parameters to understand that this means the person may need to reschedule
• It can then act by providing available times within its set parameters
• Asking them to select an available time if they would like to reschedule
• And then rescheduling them once they have selected a new time 

The agent does this without the customer or patient needing to prompt it after every step, just as a live agent would if they were handling the interaction. How does it do that?

• Each agent created with Agentforce has a defined role (i.e., a purpose and goals to work towards)
• The data it uses is what Salesforce calls ‘trusted data’, which can include resources like company knowledge articles, CRM data,  public websites, and external data via Data Cloud, etc.
• Its actions are predefined tasks (e.g., running flows, using templates, or following Apex code/Salesforce configuration) based on a trigger or instruction
• It completes actions within the parameters mentioned previously; examples include LLM (natural language) instructions to escalate things to an individual or security features built into the Einstein Trust Layer 

‍

Use Case Examples

Aside from the example given in the ‘how it works’ section, and using product recommendations paired with shopper data to help Matthew McConaughey not look like a 90’s-era Jameriqui impersonator, what are some ways Agentforce can deliver value for clients?

The number of use cases it can cover is expansive, but we’ve also got some great examples for both internal and external facing use cases from client projects (and we’re constantly exploring more use case options).

‍

Industry: Technology, Tech, SaaS

Agent Goal: Streamline customer service, deflect cases, provide 24/7 support 

Function: Resolve common inquiries and escalate complex issues

Agent in Action: An Agentforce Service agent is deployed on a customer facing Experience Cloud site; it:

• Is able to answer a wide variety of customers’ questions by referencing a client’s Knowledge Base in Service Cloud, through an Einstein Data Library in Data Cloud
• Escalates to a live customer service rep using Omni-Channel and chat functionality when appropriate
• Operates within preset guidelines, but customers have access to a broader amount of information and complex issues can be escalated more quickly

‍

Industry: Any organization looking to increase Sales team efficiency 

Agent Goal: Streamline business development efforts, free up Sales reps to handle relationship building and high-value/strategic tasks 

Function: Automate lead follow up/outreach, answer company and produce/service questions, provide white papers, schedule rep appointments, and handle outreach cadence 

Agent in Action: The agent leverages the out-of-the-box Agentforce SDR Agent to:

• Define engagement rules and data sources (through an Einstein Data Library in Data Cloud), including files and Knowledge Articles that will autonomously respond to incoming leads as defined
• Connect to Einstein Activity Capture, which allows the SDR Agent to schedule rep appointments when warranted 

‍

Industry: Any organization looking to elevate internal processes

Agent Goal: Ensure employees are aware of and following standard operating procedures (SOPs)/best practices  for a variety of goals/outcomes (e.g., safety, efficiency, increased sales/customer satisfaction, etc.)

Function: Act as a resource on SOPs/best practices for employees, offering guidance and support

Agent in Action: Together, the out-of-the-box Agentforce (default) agent, General FAQ Topic, and related Actions are used to answer employees’ questions around SOPs/best practices, by referencing:

• A client’s Knowledge Base in Service Cloud
• Uploaded PDF documents in Data Cloud (think SOP Manual), using RAG (Retrieval Augmented Generation)

‍

Industry: Healthcare

Agent Goal: Streamline patient services, provide 24/7 support

Function: Answer general practice and provider questions from patients as well as provide scheduling services 

Agent in Action: An Agentforce Service agent is deployed on an authenticated patient portal, where it can:

• Answers patients’ inquiries about the practice and providers by referencing org data and a Salesforce Knowledge base
• Provides scheduling services to set, reschedule, and cancel appointments
  
  

Industry: Medical Devices and Technology

Agent Goal: Streamline potential customer inquiries and provide product or technical support

Function: Deflect cases, increase customer service or product satisfaction, provide 24/7 support

Agent in Action: The agent leverages device data through Data Cloud to ground itself in customer and device-centric data; this allows it to:

• Provide more contextual customer service
• Answer device and functionality questions based on a client’s Knowledge Base
• Review agent feedback and inquiries to identify possible product bugs or needed improvements 

‍

Where to Start: Agentforce Impact & ROI

We’ve discussed above how broad the number of use cases are, and it’s easy to get caught up in the art of the possible, but when looking at ROI, don’t overlook the quick win use cases that provide high value — efficiency and increased satisfaction for customers, patients, and employees.

This brings us to figuring out where and how you can leverage Agentforce, and identifying the ROI for your specific organization:

• Take a step back and look at your business processes and flow of work
• Review the roles and tasks of your employees, as well as how they currently interact with Salesforce
• Consider what type of interactions your customers need and how those play out today

From there:

• Ask yourself where generative AI like Agentforce can add value (e.g., creating summaries, drafting emails, streamlining digital labor, etc.)
• Identify and breakdown workflows that require a lot of clicks into logical sequences of events; for example, finding events relevant to a customer, scheduling the customer to attend, and sending a follow-up email

When you’re going through this analysis and identifying areas for potential improvement with Agentforce, those quick-win use cases are a great place to start and there are quite a few standard Agent Actions that come out-of-the-box (including everything from Answer Questions with Knowledge, draft an email, summarize record, draft a case response, and log a call).

• We recommend working with an implementation partner to understand which require more setup and which are more streamlined, as well as the nuances between internal-facing and external-facing agents, when separate agents are necessary, security best practices, etc.
  
• If you want to do some research before speaking with someone, Salesforce has a library of use cases that will help you brainstorm areas to use Agentforce in your own org., many of which leverage standard actions
• Partners who have Agentforce experience (like Kicksaw!), will likely have even more use cases and configuration examples for you to choose from, and if you’re looking for something unique, they’ll need to spend less time figuring out the basics
• While grounding your agent in your CRM data is the first place to go, with Data Cloud you can expand the data your agent can leverage (beyond data streams to unstructured data), which creates a bigger ROI potential 
• We highly recommend checking the accuracy of your org’s data as well as duplicates, completeness, and validations, because they all play a critical role in the success of an Agentforce initiative

Lastly, a final note about ensuring ROI, a critical factor for the long-term success of your Agentforce initiative, is an AI Agent Manager. Agents have a role and a job to do, and they need evaluation, coaching, and sometimes a performance improvement plan (in this case, refining instructions, topics or prompts). 

• Currently you can leverage various Agentforce reports for this, but an Agentforce Agentforce client project details and resources.
• Conversation Explorer is coming soon and will make agent management easier

‍

Summary and Next Steps

We know that was a lot of info. The TLDR is that:

• It’s a big leap forward in terms of AI tools
• We are already seeing the value and impact that Agentforce is having with our clients
• New features and updates are coming from Salesforce at breakneck speed, outside of their traditional 3x a year release schedule
• It’s delivering some really great ROI and will continue to do so, thanks to a big push for Agentforce investment and innovation by Salesforce

If you’re evaluating Agentforce, are ready for next steps, or have any questions, please feel free to contact us. Kicksaw is staying up-to-date, actively working on Kicksaw projects, and building out proofs of concept to help our clients realize the impact of the new ‘agentic’ wave (sorry, we had to get the word ‘agentic’ in here somehow). 

‍