What is Salesforce Agentforce, and what does it change for a business that already runs on Salesforce? The name comes up in every vendor briefing, but the explanations tend to be either too technical or too promotional. This guide sticks to what Salesforce itself publishes, in seven points.

NSquare delivers Salesforce Cloud services, so read this as a partner's plain summary. It is not a product review, and Salesforce's pages change often, so check them before you decide anything.

1. What Is Salesforce Agentforce? A Platform, Not a Chatbot

Salesforce describes Agentforce as an AI agent platform whose agents are autonomous, proactive applications built to carry out specialized tasks for employees and customers. In practice that means software that can answer questions and take actions, not only suggest text.

2. Agents Do Tasks Across Service, Sales and Operations

Salesforce lists use cases across customer service, contact centers, field service, employee service, sales and IT service. Its example tasks include resolving an open case, drafting a follow-up email, scheduling a technician visit and qualifying an inbound lead.

3. You Instruct Agents in Plain Language

Salesforce's training on Agentforce guardrails and trust patterns explains that each agent works through topics. A topic holds actions, which are the tools available, and instructions, which tell the agent how to decide. Some Salesforce pages call these subagents.

Instructions are typically phrased in plain terms, such as "always", "never" or "if this, then that". Admins control the guardrails, and internal leadership typically signs them off.

4. The Data Underneath Decides the Answers

Salesforce says agents use trusted business data, including Salesforce CRM data and external data from Data 360, plus the knowledge you give them. Connectors and flows link them to other systems.

If departments define a term differently, an agent inherits the confusion, as we show in why AI gets "enrollment" wrong.

5. Trust Controls Come Built In

According to Salesforce's training on the Einstein Trust Layer, it masks sensitive details such as card and phone numbers before a prompt reaches the model, grounds answers in your data, checks responses for harmful content and keeps an audit trail.

Salesforce also says the model provider does not retain the prompt, the customer data or the response.

6. Humans Stay in the Loop

Salesforce says agents can escalate to a human when an issue is too complex or a customer is becoming frustrated, and topic instructions set when that happens. Its guidance is to define clear handoff rules and keep monitoring after launch, not only before it.

Its overview page names Agentforce Observability for tracking agent performance. Human approval for risky actions matters on any platform, as we cover in our Copilot Studio implementation guide.

7. Pricing Is Not a Simple Per-Seat Model

Salesforce's overview page lists Flex Credits, Resolutions or per-user licensing. It also says every Salesforce customer can get started for free through Salesforce Foundations.

We do not quote prices here because they change, so check Salesforce's own pricing. Budget for the work around the licence as well, such as data cleanup, integration and monitoring.

What We Have Seen

We have no Agentforce delivery results to quote, so this guide sticks to what Salesforce publishes. From our Salesforce work, discovery comes first, and data definitions and ownership are easier to settle before an agent depends on them.

If you are shortlisting partners for any Salesforce project, see our guide to choosing a Salesforce implementation partner.

FAQs

• What does Agentforce do? 

It runs AI agents that answer questions and take actions on tasks such as resolving cases, drafting emails, scheduling visits and qualifying leads.

• Is Agentforce free? 

Salesforce says every customer can get started for free through Salesforce Foundations. Its pricing options include Flex Credits, Resolutions and per-user licensing.

• What is the Einstein Trust Layer? 

Salesforce's built-in AI security layer. It masks sensitive data, grounds answers in your data, checks responses for harmful content and keeps an audit trail.

• Does Agentforce replace human agents? 

Salesforce's design includes escalation to humans. What changes in your team depends on how you scope the agent.