The AI & Salesforce glossary

Agentforce in Slack

Agentforce in Slack is the deployment of a Salesforce Agentforce agent directly inside Slack, where it runs as a conversational participant — reachable by @-mention in channels, DMs, and threads — rather than inside a Salesforce screen. The agent reads the context of the conversation it is in, grounds on CRM and Data Cloud data, and can both answer and take action where employees already work, governed by Slack identity and channel permissions on one side and Salesforce permission sets on the other. In practice it is the primary employee-facing surface for Agentforce, distinct from the customer-facing surfaces that live on a website or an Experience Cloud portal.

Salesforce CPQ

Salesforce CPQ (Configure, Price, Quote) is the part of Salesforce that turns a sales rep's product selections into a valid, correctly priced, approved quote. It encodes the company's product configurations, pricing rules, discounting policy, and approval thresholds as data and logic on the platform, so every quote follows the same rules instead of living in a rep's spreadsheet.

Salesforce Experience Cloud

Salesforce Experience Cloud is the platform for building external-facing digital experiences — customer portals, partner communities, help centers, and microsites — directly on top of your Salesforce data and security model. Each external user authenticates against Salesforce records and is governed by the same sharing rules as your internal org, so it acts as the governed front door through which customers and partners see and act on their own data. It was formerly called Community Cloud.

Salesforce Field Service

Salesforce Field Service is the part of Salesforce that manages work done in the physical world — dispatching technicians, scheduling appointments, and tracking jobs from request to completion. It models work orders, service appointments, technician skills, territories, and parts inventory as data, and uses a scheduling engine to match the right person to the right job at the right time. A mobile app lets technicians work the schedule, capture results, and update the record from the field, often offline.

Salesforce Health Cloud

Salesforce Health Cloud is an industry edition of Salesforce that ships a healthcare- and life-sciences-specific data model on top of the core platform, so patient, member, provider, and care-coordination concepts are standard objects rather than custom fields bolted onto generic CRM. Its purpose is a single, governed view of a person across clinical, administrative, and engagement data, aligned to HL7 FHIR. It is the layer Salesforce uses to deliver patient engagement, care management, and provider relationship workflows under HIPAA-eligible controls.

Agentforce

Agentforce is Salesforce’s platform for building AI agents — software that reasons over your business data, makes decisions, and takes actions inside Salesforce, governed by your existing permissions and audit trail. Unlike a chatbot that only replies, an agent can complete a task end to end.

AI Agent

An AI agent is software that pursues a goal by deciding its own next steps: it reasons with a language model, calls tools or APIs to act on the world, reads the result, and decides again, looping until the goal is met or it stops. Unlike a chatbot that only returns text, an agent takes actions. Unlike a fixed automation, it chooses the path at runtime instead of following one you scripted in advance.

AI Guardrails

AI guardrails are the controls that constrain what an AI system is allowed to say and, more importantly, do — enforced at the input, the model, the tool/action layer, and the output. They are not a single content filter but a layered control system that decides which actions an agent can take, on which data, under whose authority. For agents that write to systems of record, the action-layer guardrails — permissions, scopes, caps, approvals — matter far more than the word-filtering ones, because they are the only controls that still hold when the model is wrong.

Grounding (RAG)

Grounding is the practice of feeding a language model the specific, retrieved facts it needs at answer time, so its response is based on your data instead of its training memory. Retrieval-augmented generation (RAG) is the most common way to do this: fetch the relevant records or passages first, then have the model answer using only what was fetched. The point is not to make the model smarter but to make every answer traceable to a source you control.

Large Language Model (LLM)

A large language model (LLM) is a neural network trained on vast amounts of text to predict the most likely next unit of text (a token) given everything before it. By doing that one task at enormous scale, it learns to generate fluent language, answer questions, write code, and follow instructions. An LLM holds no memory between requests and no live access to your data unless you supply both at runtime, in the prompt.

Managed AI Workforce

A managed AI workforce is a set of AI agents that someone operates, monitors, and stays accountable for as an ongoing service — not just builds and hands off. The defining feature is operational ownership: a named party is responsible for keeping the agents accurate, safe, and producing a measurable business result over time, the way a managed service runs your infrastructure rather than just installing it.

AI ROI

AI ROI is the net financial return from an AI system measured against everything it costs to build, run, and govern over its useful life: ROI = (value created − total cost of ownership) ÷ total cost of ownership. It differs from traditional software ROI in two ways: the costs are recurring and usage-based (per-token inference, monitoring, error correction) rather than a fixed license, and the value depends on the model's hit rate on a real task and what the misses cost — not on whether the feature shipped.

Speed-to-Lead

Speed-to-lead is the elapsed time between a prospect raising their hand — a form fill, a demo request, an inbound call — and your first meaningful response to them. It matters because the value of an inbound lead decays fast: the further you are from the moment of intent, the lower your odds of reaching and converting that buyer. It is one of the few revenue levers a company controls entirely on its own side of the table.

Customer 360

Customer 360 is a single, unified view of everything an organization knows about a customer — identity, transactions, support history, web and product behavior, and consent — resolved from otherwise siloed systems into one trustworthy record per real person or account. It is a data-integration outcome, not a product or a screen. The bar that matters is whether every team and system, including an AI agent, can read the same resolved profile and act on it.

Customer Data Platform (CDP)

A Customer Data Platform (CDP) is software that ingests customer data from every source a company has — web, app, CRM, support, billing, ad platforms — resolves which records belong to the same person, and serves the result as one persistent, governed profile that operational tools can act on in real time. The hard part isn't storage; it's the identity decision that says the anonymous web visitor, the email subscriber, and the support ticket are the same human. A warehouse can hold the data. A CDP is judged by how well it unifies and activates it.

Data Unification

Data unification is the process of matching, merging, and reconciling records that describe the same real-world entity — a person, account, or product — scattered across separate systems into one consistent, queryable profile. It is not a one-time data dump; it is a continuous pipeline that ingests, standardizes, resolves identity, and reconciles conflicts as source data keeps changing. The output is a single version of an entity that downstream systems — analytics, marketing, and AI agents — can act on without guessing which record is true.

Identity Resolution

Identity resolution is the process of deciding which records scattered across your systems refer to the same real-world person or account, then linking them into one profile. It is fundamentally a probability problem: each candidate pair gets a match score, and a threshold turns that score into a yes-or-no decision. Set the threshold wrong and you either fragment one customer into many profiles or collapse two customers into one.

Salesforce Data Cloud

Salesforce Data Cloud is the layer that ingests, unifies, and resolves identity across your data sources into a single real-time customer profile that the rest of Salesforce — including Agentforce AI agents — can reason and act on.