Agentforce And

Prompt builder
 Agentforce and AI Agents
•What is AgentForce: AgentForce is a new platform developed by Salesforce that allows businesses to
create and deploy intelligent and Autonomous AI Agents.
These agents work alongside human employees to enhance customer experiences and streamline business
operations. Salesforce Provides OOB Agents for Sales, Services and Marketing etc but Customers can extend
these agents or build new agents as per the business needs, Agentforce Platform will provide the ability to
Partners to build and Deploy agents and agent Actions through AppExchange which customers can add in their
org.
• Agent:- Agents Are AI Assistants who can perform tasks, make decisions, and interact with customers or
internal systems autonomously. these AI Agents understand and process information, communicate naturally
with users, and take actions to perform specific task It uses Salesforce Atlas Reasoning engine to understand
the intent of the Question follow all the instructions defined for the particular task , find the appropriate
response and Action and then perform the task as requested.
Example: Customer Service agents :-
Scenario: A customer wants to return a pair of shoes purchased online.
AI Agent's Role: The AI Agent assists the customer by accessing their purchase history, initiating the return
process, and providing instructions for shipping, all without needing a human representative.
•What It Does:
• Answers questions , Takes actions , follows instructions in everyday language
• Works on multiple Channels in Salesforce, Agent for internal employee to help
users in the daily task , an agent for customer services deployed on
Community site, Agent for WhatsApp chat etc. agents for attending calls etc.
•How It's Different :- It's not just a regular chatbot. While bots follow strict scripts,
Copilot is more flexible and can handle a wide range of requests. It's built to understand
complex instructions and carry out multiple tasks by itself
 Agentforce and AI Agents
•Actions:
• Comes with many built-in tasks it can do right away they are called
Standard Actions (like drafting emails or finding customer records).
• You can also customize it to do special tasks specific to your
business needs(custom actions through Apex, flow and Prompt
template, External API’s).
•Trust and Security:
• Safety first: Agent makes sure only authorized users can see
sensitive data(Respect user Permission).
• It uses special Architecture defined to keep data secure and private,
following Salesforce's high standards for trust and safety(Einstein
trust Layer).

Key Features:
• AI Agents handle routine or repetitive tasks, allowing employees to focus on more
strategic and creative work
• AI Agents use data to provide customized experiences for each customer
• AI Agents can autonomously perform actions like answering customer inquiries,
processing orders, or scheduling appointments
• Businesses can easily build and deploy AI Agents using tools like Agent Builder
without needing advanced coding skills
• AgentForce includes safeguards to ensure AI Agents operate responsibly and protect
customer data
Agentforce and AI Agents
Agentforce and AI Agents
Agentforce and AI Agents
Agentforce and AI Agents
Agentforce and AI Agents
Agentforce and AI Agents
Agentforce and AI Agents
 Signing up and Activating Agents
•Sign up for the Free Developer org with
access to Agent (
https://trailhead.salesforce.com/content/l
earn/projects/quick-start-build-your-first-a
gent-with-agentforce/configure-an-agentf
orce-service-agent
) Salesforce might change the URL for
Signing up

•Go to Setup -> Einstein Setup -> Turn on

Einstein

•Go to Setup -> Agents-> Turn on Einstein

Copilot for Salesforce
 AI Agents Continue…
AgentForce utilizes a combination of
data, AI models, and business logic to
create AI Agents that can:
•Understand Roles: Know their specific
job functions within the company.
•Access Data: Retrieve and use data
from various sources to make
informed decisions.
•Take Actions: Execute tasks through
predefined workflows and processes.
•Interact on Multiple Channels:
Communicate with customers via chat,
email, voice calls etc.
 AI Agents Continue…
• Topics :- Topics Define the work that these AI agents will do , So each Topic
Defines Jobs that the agent needs to perform

• Instructions :- instruction are kind of rules that Agent should follow while
Solving customer queries for example you can give instruction that do not
accept Nagative values, you can define rule that Zip code length should be 5
digits, user should provide order number etc. Instructions are the guardrails
which ensures that Agent will Perform the task within the boundaries that we
have defined.

• Actions :-Actions are the one which actually performs the task, like Creating
a Draft order, Or booking an Appointment, fetching some Account details etc.
there are Standard and custom action that we will be creating in the system
as per the organization needs which will solve a particular task, the custom
actions can be created using Flows , apex and Prompt templates
 Atlas Reasoning Engine.
• Atlas Engine :- At its core, Atlas is a reasoning
engine designed to simulate human-like thought
processes within Salesforce's Agentforce
platform. Think of it as the "brain" behind AI
agents that can make decisions, take actions, and
continuously learn—all in real time

Key Components:
• Reasoning Capabilities: Atlas
doesn't just process data; it
understands and interprets it to
make informed decisions.
• Real-Time Processing: It handles
vast amounts of data instantly,
allowing for immediate action.
• Continuous Learning: Atlas refines
its responses over time, improving
accuracy and effectiveness.
How Agent Works
Architectu
re
 Salesforce Model Builder
Model Builder: Model Builder is a feature of the
Einstein 1 Studio that enables you to connect your data
with your preferred large language model.
Model Builder allows Salesforce professionals to create
custom AI models tailored to specific business needs
without requiring extensive data science expertise or
integrations, it's a No code tool to connect external
Models or to create a model from Scratch.

With Model Builder, you can leverage your Salesforce

data to train machine learning models for various
purposes. The platform provides a user-friendly interface
and guided workflows to facilitate the model-building
process.

You can use Salesforce’s Shared LLMs, Salesforce’s

hosted third-party LLMs, or bring your own model
(BYOM) from other platforms (e.g. Amazon SageMaker,
Google Vertex AI) and combine your company data with
preferred AI models
Salesforce Model Builder
 Event Logs/Audit
•Every User interaction(Session) with Einstein Copilot is logged in the system and can be referred to understand why Copilot took certain
decision
•This is part of the Einstein Trust layer i:e Auditing all the Interactions with copilot
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
Einstein Trust Layer
 Custom Topics and Custom
Actions
•Custom Actions :- Created for each business specific to their need using flows, Apex and Prompt builders

1. Apex: Salesforce programming language.

2. Prompt Template: Reusable prompts written with natural language.
3. Flow: No-/low-code automation tool.
4. External Services: No-code connection to external API.
 Custom Actions
•Custom Actions :- Steps to Create a Custom Action

1. Understand the business requirements for which you need to create a custom action.
2. Write the business logic in Apex, Flow, or a prompt template.
3. When creating the invocable method, make sure to provide a description of the method and include
multiple utterances. The method description will become the Copilot instruction for the new custom
action. During the planning phase, Einstein will use this instruction to decide which action or set of
actions can solve the current requirement.
4. Assign the custom action to the Copilot, ensuring that the Copilot is deactivated before assigning or
removing an action
5. Test the new action with multiple utterances in the Copilot Builder. If you do not get the desired output
or if the Copilot is unable to locate the custom action, make changes to the business logic or the custom
action instructions
6. Activate the Copilot
 Custom Actions Cont..
•Custom Actions :- Using Apex Programming Language

1.Every Apex Method which has @InvocableMethod Annotation can be used as Copilot Custom Action
2.To create the Custom Action Click on New Custom Action from Setup -> Einstein Copilot Actions .
3.Select the type as Apex and Select the Correct Apex Method
4.While creating Custom Action it is Super important to fill in instruction field with as
much Description as Possible , this will help the copilot During the planning Stage to
Decide if It should Use this Custom Action Or not.
5.When a Custom/Standard Action returns a record such as Account or Lead then it will
shown as a Tile on The UI
6.As a Best Practice give many example utterances on the instruction field so it will help
Copilot to decide if this action can be used for a task
7.As a Best Practice always Test the Action in Copilot builder before Rolling it out to the
users, while Testing try different utterances and prompt and check if Copilot is able to
find the correct Action Or Not.
 Custom Actions Cont..
•Custom Actions :- Using Flow

1.Every Invocable Flow can be used as Copilot Custom Action

2.To create the Custom Action Click on New Custom Action from Setup -> Einstein Copilot Actions .
3.Select the type as Flow and Select the Correct Flow from the Drop Down
4.While creating Custom Action it is Super important to fill in instruction field with as
much Description as Possible , this will help the copilot During the planning Stage to
Decide if It should Use this Custom Action Or not.
5.As a Best Practice give many example utterances on the instruction field so it will help
Copilot to decide if this action can be used for a task
6.As a Best Practice always Test the Action in Copilot builder before Rolling it out to the
users, while Testing try different utterances and prompt and check if Copilot is able to
find the correct Action Or Not.
 Einstein Data Library
•Einstein Data Library :-

The Einstein Data Library is a powerful tool designed to

streamline the setup and use of Retrieval-Augmented
Generation (RAG) within Salesforce. It simplifies the process by
ingesting data from Knowledge Articles or uploaded files,
automatically creating essential components like Data Streams,
Data Model Objects (DMOs), search indexes, and retrievers. This
capability ensures that organizations can efficiently organize and
leverage their data for enhanced AI-driven insights. Additionally,
the Einstein Data Library seamlessly integrates with Agentforce
Agents and Einstein Bots, enabling smarter and more
contextually relevant interactions in customer service and other
use cases.
Einstein Data Library
Einstein Data Library
 Prompts and Prompt Builder
•Prompts :- Generative AI can give correct output or meaningful Answers if it’s given effective prompts. Prompts are the way you communicate
with the large language models (LLMs). We know that designing an effective prompt is key to unlocking the benefits of generative AI.

•So, specific prompts lead to effective prompts. But getting specific can be a challenge for your business, because it requires you to add your business data to a
prompt. How do you do that for every customer or product? And is your business data safe? This challenge is where Prompt Builder and the Einstein Trust Layer
come in.

PROMPT Builder

•Build trusted generative AI with ease :- Create prompt templates bound to one or multiple objects to
generate summaries or content
•Enrich prompts with your data :- Create more relevant prompts grounded in your business data from
the CRM, Data Cloud, and external sources
•Preview and version templates :- Test and preview the resolution and response, and version them
•Use templates everywhere :- Build prompts once and reuse them everywhere in sales emails, record
pages, Flows, Apex and Einstein Copilot
Prompt Template Types
Invocation of Prompt templates
 Building Prompt Templates
Example of Sales Email Template Invocation
 Einstein Service Replies
•Einstein Service Replies Helps customer service teams respond to
customers more quickly and accurately by using AI to suggest what to say.
This can be during a live chat or when writing an email. The goal is to make it
easier for customer service agents to help people, which in turn makes
customers happier.
How It Works:
• Smart Chat Replies: When chatting with a customer, Einstein suggests
helpful responses, so the agent can reply faster.
• Email Help: When an agent is working on a case, Einstein can suggest what
to write in an email, using information from useful articles to ensure the
reply is accurate.
•Examples:
•Live Chat Example:
Customer: "I can’t log into my account."
Einstein Suggestion: "I’m sorry you’re having trouble. Have you tried resetting your password using
the 'Forgot Password' link?“
Email Example

Subject: Help with Installing Your Software Update

Dear [Customer Name],

Thank you for contacting us about your installation issue. Here’s how to get the update installed:

1.Download the Update: Ensure you have the latest version from our website [Link to Download].

2.Run the Installer: Double-click the file and follow the on-screen instructions.

3.Restart Your Device: After installation, restart your device to apply changes
Einstein Knowledge Creation
•Einstein Knowledge Creation : Einstein Knowledge Creation helps customer service
teams create helpful articles more easily and quickly. Based on a customer conversation,
Einstein drafts fluent, relevant knowledge articles that agents review, edit, and save. As your
knowledge base develops, agents can quickly find answers to issues and address emerging
customer pain points..

•Examples:

•A customer calls about trouble connecting their new smart TV to Wi-Fi. Once The Problem is
Resolved Agent Clicks on Create Einstein Article and Einstein uses the Content from the
conversation and Populates the fields of the Knowladge articles which Agent Validates and
Saves the article as Draft

•When an agent Completes a Case for a user who was having issue with Solar panel
Installation, He clicks on Create Article using Einstein and It generates and Populates most of
the Fields in the Knowladge article which Agent Validates and Saves the article as Draft
 Einstein Call Summaries
•Einstein Call Summaries : Einstein Call Summaries automatically generate
concise summaries of voice and video calls within Salesforce. These summaries
highlight key discussion points, customer feedback, and next steps, helping
teams stay organized and collaborate effectively.

Key Features:
•Automated Summaries: After calls, summaries are auto-generated and
editable.
•Key Elements: Each summary includes the main points, customer impressions,
and next steps.
•Collaboration: Summaries can be shared across teams for better alignment.
•Easy Access: Summaries are stored in the "Summary" tab and are linked to call
records.
•Language Support: Available for transcripts in English, French, German, Italian,
Japanese, and Spanish.

Examples:
•After a sales call, Einstein generates a summary capturing the key topics
discussed, the customer's feedback, and next steps like scheduling a follow-up
meeting.
 Einstein Call Explorer
•Einstein Call Explorer: Einstein
Conversation Insights users can ask
questions directly from voice and video
call records. Powered by Einstein, the
new Call Explorer allows users to quickly
gather information about a call, such as
product mentions or outstanding
customer questions.

To bring up Call Explorer, use the Explore

Conversation button.

Call Explorer delivers one answer per

query. Previous questions and answers
are not stored. Call Explorer is currently
supported for calls in English, French,
German, Italian, Japanese, and Spanish
Einstein for Developers
Einstein for Developers