Salesforce has been renaming, rebranding and repositioning its AI components several times since 2023. If you’re confused, you’re in the same boat as I am. This glossary is a source of truth — a definitive source of truth that maps old names to new names and defines each and every major component of the current Salesforce AI architecture.
What Got Renamed? A Quick Reference
This table maps significant Salesforce AI naming changes since 2023.
| Old Name | Current Name (Feb 2026) | When | |
|---|---|---|---|
| Einstein 1 Platform | → | Agentforce 360 Platform | 2023 → 2025 |
| Einstein Copilot | → | Agentforce | Apr 2024 → Oct 2024 |
| Einstein 1 Studio | → | Builder Tools | Early 2024 → Late 2024 |
| Copilot Builder | → | Copilot Builder (name retained; now extends Agentforce agents) |
Unchanged |
| Salesforce Platform (interim name) |
→ | Agentforce 360 Platform | 2024 → 2025 |
| Customer 360 | → | Agentforce 360 | 2025 |
A
Agent Builder
Low code tool for creating custom Agentforce agents. Administrators can define the behavior of agents, select the data sources, set the guardrails, and link the business processes without writing code.
Agent Script
A hybrid reasoning approach that incorporates both deterministic workflows (scripted steps) and LLM flexible reasoning. This enables organisations to define which aspects of an agent’s behaviour are predictable and which can be dynamic.
Agentforce
Salesforce’s autonomous AI agent platform. It provides prebuilt agents for sales, service, marketing, and commerce, plus tools to build custom agents. Agents can plan, reason, and execute tasks independently using the Atlas Reasoning Engine.
Pricing at launch: $2 per conversation
Versions: 1.0 — Oct 2024 · 2.0 — Dec 2024 · 2DX — Mar 2025 · 3.0 — Jun 2025
Pricing at launch: $2 per conversation
Versions: 1.0 — Oct 2024 · 2.0 — Dec 2024 · 2DX — Mar 2025 · 3.0 — Jun 2025
Agentforce 360 Platform
The current top level branding for Salesforce’s complete AI platform. It contains all of the layers of AI, from Agentforce agents on the top to Hyperforce infrastructure on the bottom. The “360” branding stems from a previous concept, Customer 360.
AgentExchange
A marketplace for prebuilt AI agents, agent actions, and agent templates. Built by Salesforce Partners and independent software vendors (ISVs). It is an extension of the concept of AppExchange to AI agents.
Agentic AI
A contemporary paradigm of AI (2024 and further) of autonomous systems that can plan, reason, and execute tasks independently. This differs from Generative AI (which creates content) and Predictive AI (which forecasts outcomes). In Salesforce, Agentforce represents agentic AI.
Atlas Reasoning Engine
The basic thinking engine for the Agentforce agents. Features System 2 reasoning (deliberative decision-making), plan generation, data evaluation, and feedback-driven self-correction. Uses a ReAct (Reasoning + Acting) architecture rather than simple chain-of-thought prompting. Supports multi-agent orchestration and built-in guardrails.
See also: ReAct, System 2 Reasoning
B
Builder Tools
Salesforce’s low-code AI customization toolkit. Includes Prompt Builder (reusable prompt templates), Model Builder (bring your own LLM), and Copilot Builder (extend agent capabilities via APIs). Official name: Low Code Builder Tools for Generative and Predictive AI.
BYOM (Bring Your Own Model)
Salesforce strategy that can be used to allow customers to connect their preferred LLM provider instead of one default model. Examples include OpenAI, Anthropic Claude, Google Gemini, and IBM Granite. Managed through Model Builder.
See also: Multi-Model AI Layer, Model Builder
C
Cloud AI Functions
AI features embedded directly into Salesforce products such as Sales AI, Service AI, Marketing AI, Commerce AI, and Dev & Admin AI. These are capabilities that are built into the system that a user will interact with without having to configure agents or models.
Command Center
The monitoring dashboard for Agentforce agents. Provides agent health monitoring, performance analytics, and detailed inspection of agent interactions. Critical for organizations that have many autonomous agents.
Copilot Builder
Part of Builder Tools. Used to extend Agentforce agents with Apex, Flow, and MuleSoft APIs. Originally designed for Einstein Copilot but retained after the Agentforce transition.
D
Data Cloud
Salesforce’s real-time hyperscale data engine. Unifies customer data across systems and provides the foundational data layer for AI. Enhancements since 2024 include vector database capabilities, zero-copy architecture, and support for unstructured data (emails, PDFs, images). Agentforce agents are given context grounding from Data Cloud.
Dynamic Grounding
A technique to make sure AI responses are associated with verified enterprise data at the time of inference. Reduces hallucinations and ensures outputs reference real customer records and data.
E
Einstein AI
The generative and predictive AI layer of Salesforce. Includes Einstein GPT, Prediction Builder, lead and opportunity scoring, conversation insights, Einstein Search, and Einstein Bots. Einstein = intelligence layer · Agentforce = autonomy layer.
Einstein Bots
Rule based and AI assisted customer service chatbots. They are older than generative AI and are useful for organizations that need to have predictable scripted workflows.
Einstein Copilot
Legacy term. Salesforce conversational AI assistant, April 2024. Originally used chain of thought reasoning to help users. It was updated to Agentforce in October 2024 with autonomous capabilities.
See: Agentforce
Einstein GPT
Salesforce’s generative AI capability. Uses LLMs to generate emails, summarize call transcripts, create knowledge articles, and produce marketing content. Part of the Einstein AI layer.
Einstein 1 Platform
Legacy term. Announced at Dreamforce 2023 as a unified Salesforce AI platform. Later renamed to Salesforce Platform (interim), then to Agentforce 360 Platform (2025).
Einstein 1 Studio
Legacy term. Original name for Salesforce’s AI builder toolkit introduced in 2024.
See: Builder Tools
Einstein Trust Layer
Security and governance layer for Salesforce AI. Features include PII masking, toxicity detection, zero data retention with LLM partners, audit logs of agent actions, and dynamic grounding. Essential for controlling autonomous AI behavior.
H
Hyperforce
Salesforce’s cloud infrastructure layer. Provides global deployment, compliance support, and scalable compute. All Salesforce AI services run on Hyperforce.
I
Integration Layer
The enterprise connectivity infrastructure. Built using MuleSoft and Informatica (acquired May 2025). Handles enterprise data movement and supports Model Context Protocol (MCP) and zero-copy integrations.
M
MCP (Model Context Protocol)
An open protocol defining how AI agents connect to external tools and data sources. Allows Agentforce agents to operate beyond Salesforce systems.
ML-Enabled APIs
Developer APIs exposing machine learning features. Examples include Einstein Vision, Einstein Language, Prediction Service, Bot SDK, and Discovery REST APIs.
Model Builder
Allows organisations to pull their own LLM providers into Salesforce. Supported models include OpenAI, Anthropic Claude, Google Gemini, and IBM Granite.
See also: BYOM, Multi-Model AI Layer
Multi-Model AI Layer
An architecture layer enabling multiple LLM providers. Organizations may choose models on the basis of use case, regulatory needs, or cost.
P
Prediction Builder
Point and click tool to create predictive artificial intelligence models without any coding. Uses supervised machine learning on data in Salesforce.
Prompt Builder
Tool to create reusable templates of prompts for LLMs. Prompts can incorporate Salesforce record fields to ground outputs in actual customer data.
R
RAG (Retrieval-Augmented Generation)
Technique in which AI uses retrieved relevant data prior to generating a response. Improves accuracy and reduces hallucinations. Agentforce enhanced RAG in version 2.0.
ReAct (Reasoning and Acting)
Reasoning framework used by the Atlas Reasoning Engine. Alternates between reasoning steps and action steps. This allows agents to evaluate results and self-correct before completing tasks.
See also: Atlas Reasoning Engine, System 2 Reasoning
S
System 2 Reasoning
Borrowed from Daniel Kahneman’s cognitive psychology model. System 1: fast, intuitive, error-prone. System 2: slow, deliberate, accurate. Agentforce uses System 2 reasoning to plan actions and verify results before executing tasks.
Z
Zero-Copy Architecture
Allows Data Cloud to access external data lakes (Snowflake, Databricks, BigQuery etc.) without the duplication of data. Benefits include lower storage costs, real-time data access, and fewer synchronization problems.
Zero Data Retention (ZDR)
Prevents the storage of Salesforce customer data by LLM providers. Prompts and responses are temporarily processed and discarded so that they are not retained for training.
This glossary is a companion to “What Is Salesforce AI?” by DataGroomr. Terminology is accurate as of February 2026. Salesforce’s branding changes frequently — check help.salesforce.com for the latest official documentation. For questions or feedback, visit datagroomr.com.
