A - DAgent Builder is a Salesforce tool for creating, managing, training, and customising AI agents; no coding skills are required.
The agent console is the main workspace in Salesforce where human service agents handle customer requests. With Agentforce, this console includes AI-generated suggestions and real-time guidance.
Apex code is Salesforce’s programming language used to add custom features or automate tasks.
A way for different software systems to talk to each other and share data.
Atlas Reasoning Engine is what makes Agentforce truly intelligent. Instead of just generating responses, it “thinks”, breaking down requests, pulling in relevant data, refining its plan and then taking action.
Records that track who accessed what data and when. Used in Agentforce to maintain transparency and support compliance.
Using AI to automatically review and classify new support cases, routing them to the right team or agent with no human effort.
A law that protects the personal data of people in California and gives them more control over how it's used.
A feature in Agentforce where AI reads a case’s content and generates a short, helpful summary so service personnel can get up to speed faster.
Software that runs entirely in the cloud (online), with no need for on-premise servers. Agentforce is fully cloud-native within Salesforce.
The process of ensuring that your systems and data handling comply with local laws and regulations, such as GDPR or CCPA.
Customer 360 is Salesforce’s approach to giving organisations a complete, unified view of every customer. This means teams can see the same, up-to-date information about each customer from their past interactions to their current needs and use it to deliver more personalised, consistent, and efficient service.
E - GEinstein AI is Salesforce’s built-in artificial intelligence. It analyses data and turns it into useful insights, predictions and actions. Instead of being a separate tool, it enhances everyday processes and ensures data is handled securely and responsibly through the Einstein Trust Layer.
A conversational AI assistant built into Salesforce. It can guide agents, automate tasks, and respond to queries using natural language.
Einstein GPT is Salesforce’s generative AI. It combines Salesforce’s trusted platform and customer data with the power of advanced language models to automatically generate content and responses. That could mean drafting emails, summarising case notes, generating chat replies or preparing reports.
Einstein Trust Layer is Salesforce’s built-in security framework for AI. It ensures that whenever AI agents use your data, it remains safe, compliant and traceable.
Tracks user activity and system events in Salesforce to help spot issues and improve security.
Controls that determine who can see or edit specific fields in a Salesforce record. Important when configuring what AI can access.
Salesforce’s low-code tool for automating business processes. Often used to design what actions an AI agent should take or suggest.
A law in the UK and Europe that protects people's personal data and privacy, making sure it's handled safely.
The set of tools and policies used to manage, monitor, and control how AI is used in a business, including approvals, audits and usage tracking.
Rules and boundaries are built into AI systems to prevent risky or inappropriate behaviour. In Agentforce, they protect data, accuracy, and brand voice.
Page menu:
H - MAn approach in which AI makes suggestions, but a human makes the final decision. Agentforce is built around this concept to ensure quality and control.
Data is sent between systems in a simple, structured format that both can understand.
A stored answer, guide, or how-to used by service teams. Agentforce can reference these when generating AI responses.
LLM is a form of AI trained on vast amounts of text so it can understand and generate language naturally. Instead of just retrieving information, it interprets what you ask, then produces clear, human-like responses. Einstein GPT uses a version of this kind of model.
A dashboard or system that tracks how well your AI agents are performing-accuracy, adoption, feedback, and more.
Support for different languages in AI interactions. Agentforce can generate responses and interpret inputs in multiple languages, depending on the setup.
N - SNext Best Action (NBA) is Salesforce’s way of guiding staff toward the smartest next step with a customer or case. It uses data and AI to suggest the most relevant action, whether that’s sending an update, escalating an issue or offering the right service.
When an agent accepts, ignores, or improves an AI suggestion, the system learns from this input to improve future results.
Permission sets are Salesforce tools that define what users (or AI) can access or do. Used to ensure AI agents don’t exceed their intended scope.
Prompt Builder is Salesforce’s tool for creating and managing the instructions (or “prompts”) that guide AI agents. Instead of writing code, you can design prompts in plain language and connect them directly to your organisation’s data.
A common way for different software systems to communicate and exchange data over the internet.
A security practice where permissions are assigned based on job roles. In Agentforce, this ensures the right people and AIs see the right data with flexibility and reduced management costs.
Logic used to automatically send new cases to the correct team, queue, or agent based on region, product, priority, or issue type.
An enterprise-grade security add-on that includes tools like encryption, field audit trails, and event monitoring-used in Agentforce to enhance data protection.
Salesforce’s customer service platform. Agentforce builds on Service Cloud by adding AI capabilities directly into the agent experience.
T - ZUsage data that shows how often features are used and how well they’re performing. Used to improve both AI and agent workflows.
Technical limitations on how much text an AI can process at once. Important for understanding prompt size and input length in GPT-style models.
Training data is information used to “teach” AI how to respond. In Agentforce, this might include past cases, customer data, or knowledge articles.
Salesforce’s built-in approach to responsible AI. Includes tools for monitoring, consent, redaction, audit logs, and model controls.
When a case is entirely resolved by automation, no human intervention is required. Often used for simple requests, such as status updates or password resets.
