Get in touch

Salesforce AI Glossary for Healthcare

Definitions of 52 Salesforce AI terms used across pharma, medtech and healthcare, from Agentforce to zero copy. The terms are grouped into three levels, from the basics to the builder's vocabulary. Every term has its own link.

All 52 Terms

Every definition is below. Jump to a level or search.

No terms match that. Try another word, or ask the Wisest Wolf.

Level 1: Pup

The basics: what AI assistants, agents and prompts are, plus Salesforce's AI and industry products.

AI assistant

The helper you chat with, like ChatGPT, Siri or Claude.

An AI assistant is an app you type or talk to, and it answers you or gets something done. ChatGPT, Siri and Claude are all AI assistants. Claude is made by Anthropic.

The new part: assistants like Claude can now work with your Salesforce data, using your own access.

Why it matters: It's the front door to almost everything else on this page.

Prompt

Whatever you type or say to the AI.

A prompt is the instruction you give an AI. It tells the AI what to do and how to go about it.

Specific prompts get specific answers. "Tell me about my accounts" gets you a wall of names. "Which hospital accounts haven't had a visit in 60 days?" gets you a plan for the week.

Business apps often add a second, hidden prompt behind yours that sets the ground rules, like tone and compliance limits.

Why it matters: Better prompts, better answers. It's the cheapest AI skill you'll ever learn.

See it in action →

LLM (large language model)

The engine under every AI assistant.

An LLM is an AI model trained on huge amounts of text, so it can understand your question and answer the way a person would.

LLMs write summaries, draft emails, create content and even write code. Anthropic's Claude runs on one. Salesforce even trained its own reasoning model, Koa.

Why it matters: When a vendor says "powered by AI," there's usually an LLM underneath. Now you know what to ask about.

AI agent

AI that doesn't just answer. It does the work.

A chatbot follows a script. An AI agent figures out the steps on its own: it breaks a goal into tasks, picks the right tools and gets it done, like updating a record or booking an appointment.

Autonomous doesn't mean unsupervised. Agents work inside the rules and guardrails you set.

Why it matters: Agents are where AI stops being a novelty and starts giving your team hours back.

Human in the loop

A person approves before the AI's work counts.

Human in the loop means a real person reviews and approves what the AI does before it becomes final.

Salesforce builds this in. In Salesforce in Claude, nothing is saved to Salesforce until you approve it, and no email goes out without you. Every session starts with a real person signing in, and every change is logged under that person's name.

It's how we build AI for regulated teams too. On our CancerNavigator project, clinical staff review and approve every AI suggestion.

Why it matters: In pharma and medtech, a human signs off on everything that matters. Your AI should work the same way.

See it in action →

Computer autonomy

The thing that destroys humanity. Probably.

Computer autonomy is AI running by itself with no human in the loop. No approvals, no checks, nobody at the wheel. It's the plot of every robot movie ever made.

Salesforce designed its hosted AI connections to avoid it: no shared robot accounts, no machine-to-machine logins and no AI acting outside a real person's session.

Why it matters: When someone pitches you "fully autonomous" AI for your field team, ask who approves the work.

See it in action →

Agentforce

Salesforce's platform for AI agents.

Agentforce is Salesforce's AI agent platform. It used to be called Einstein Copilot.

Agentforce agents can answer questions using your Salesforce data, schedule appointments, send reminders and hand tricky issues to a person.

Why it matters: If your company runs on Salesforce, you'll hear "Agentforce" in every roadmap meeting.

AIforce

Salesforce, wherever you already work.

AIforce is what Salesforce unveiled at Dreamforce 2026. It brings your Salesforce data, workflows and permissions to the AI tools your team already uses, like Claude, Slack and Lightning.

It launched with three pieces: Claudeforce, Slackforce and Agentforce Coworker. Salesforce says more are coming.

Why it matters: Your team may soon do its Salesforce work without ever opening Salesforce.

Claudeforce

The Salesforce and Anthropic partnership.

Claudeforce is the partnership between Salesforce and Anthropic, the company behind Claude. It was announced on August 26, 2026, and became one of the first pieces of AIforce at Dreamforce.

It runs both ways. Salesforce comes into Claude through Salesforce in Claude. Claude comes into Salesforce, powering Agentforce Coworker by default and working as a reasoning model inside Agentforce. It's also the default AI model in Slack.

Don't mix these up: Claudeforce is the partnership. Salesforce in Claude is the app your team actually opens.

Why it matters: Claudeforce is the big picture. The app you'll actually use has its own name.

Salesforce in Claude

The app that puts Salesforce inside Claude.

Salesforce in Claude is the first product to come out of Claudeforce. It connects Claude to your Salesforce, so you can ask for meeting prep, catch up on an account or review your pipeline from a Claude chat.

It comes with 37 ready-made sales skills, from meeting prep to deal health reviews. An admin connects it once and everyone on the team gets access. It's in beta for all customers.

Don't mix these up: Claudeforce is the partnership. Salesforce in Claude is the app your team actually opens.

Why it matters: It's the most direct way to put Salesforce in your field team's pocket today.

See it in action →

Headless 360

Salesforce, without logging into it in a browser.

Headless 360 lets AI assistants like Claude and ChatGPT work with your Salesforce data and processes, so nobody has to log in to get something done.

Why "headless"? The head is the screen you click through. Take it off and everything underneath still works. Odd name. Handy idea.

Salesforce first announced it at TDX 2026, according to CIO, and expanded it in August. Its connection for AI assistants is in open beta.

Why it matters: It's why your team can ask a question on their phone and get a real Salesforce answer back.

See it in action →

Skills

Ready-made routines the AI knows by heart.

A skill is a ready-made routine for the AI. You kick it off with a short request, like "prep me for my 2 o'clock," and the AI runs the same proven steps every time.

Salesforce in Claude ships with 37 sales skills, including a daily briefing, call prep, follow-up drafts and pipeline reviews. On our blog, we call them routines.

Why it matters: Skills mean your whole team gets the same good answer, not 40 versions of it.

Slackbot

Your AI assistant inside Slack.

Slackbot is the personal AI assistant built into Slack. With Slackforce, it reads across your Slack conversations and your Salesforce data at the same time.

It can spot accounts going quiet, figure out why, reassign an owner, create a follow-up task and draft the win-back email, all without leaving Slack. Claude is the default AI model behind it.

Why it matters: If your team lives in Slack, this is AI that meets them there.

Slackforce

Salesforce, inside every Slack conversation.

Slackforce is one of the first three pieces of AIforce. It brings Salesforce data and actions into Slack.

Log notes from a call, update records or spin up a new account with a simple prompt, right from Slack. Teams can also pull live data into shared views that everyone can filter and comment on.

Why it matters: Fewer apps to switch between. Fewer "I'll update Salesforce later" moments.

Agentforce Coworker

An AI teammate that lives inside Salesforce.

Agentforce Coworker is an AI teammate built right into Salesforce. It reasons across accounts, activity and history, surfaces insights and takes action.

It works within your existing permissions and business rules, and it can call the Agentforce agents you've already built. Salesforce says 100,000 people turned it on in its first 35 days.

Why it matters: It's AI for the people who do open Salesforce every day.

Health Cloud

Salesforce, built for healthcare.

Health Cloud is Salesforce's platform for healthcare. It brings clinical and nonclinical data together in one place.

It's built for payers, providers and public health agencies, and it connects to other health systems using industry-standard healthcare data connections. Salesforce says it's built to meet rules like HIPAA and GDPR.

Why it matters: If your customers are providers or payers, this is the version of Salesforce built for their world.

Life Sciences Cloud

Salesforce, built for pharma and medtech.

Life Sciences Cloud is Salesforce's AI-first platform for pharma, medtech, consumer health and animal health companies. It connects clinical, medical, commercial and patient services, with compliance built in.

For pharma, that includes patient services, savings programs and adverse event management. For medtech, it covers field teams, inventory planning and contract compliance.

Why it matters: If you sell drugs or devices, this is the Salesforce made for your team.

Level 2: Teen Wolf

Putting AI to work: connections, data, pricing, security and the tools that shape AI answers.

MCP (Model Context Protocol)

The universal plug between AI and your apps.

MCP is an open standard for connecting AI apps to the other systems you use. Build one connection and any AI assistant that speaks MCP can use it.

Salesforce offers ready-made MCP connections that it hosts for you. They've been generally available since April 2026, and they only let the AI do what the signed-in person is allowed to do.

Curious about the standard itself? It lives at modelcontextprotocol.io.

Why it matters: MCP is why connecting Claude or ChatGPT to Salesforce is a setup task, not a months-long integration project.

Data 360

Salesforce's data engine, formerly Data Cloud.

Data 360 is the new name for Data Cloud. It's the real-time data engine behind the Salesforce platform and the data foundation for Agentforce.

It brings data from all your systems into one place, so AI and your team work from the same view of each customer. Existing Data Cloud contracts carry through to the end of their term.

Why it matters: AI is only as good as the data it can reach. This is where that data comes together.

Data readiness

Is your data good enough for AI yet?

Data readiness is how prepared your data is for AI. It starts with classic data quality: is it complete, consistent, correct, current and free of duplicates?

For AI there's one more question: can the AI reach only the data it's allowed to see? Salesforce cites a Capgemini report that found fewer than one in five companies has a high level of data readiness.

Want to know where you stand? Our AI Readiness Evaluation scores it for you.

Why it matters: Clean data is the difference between an AI demo and AI your team actually trusts.

Data standardization

Getting all your data to speak the same language.

Data standardization, also called harmonization, means getting data from different systems into the same format. "USA," "U.S." and "United States" should be one country, not three.

Data 360 does this by mapping every source to one shared customer data model.

Why it matters: If your CRM thinks one hospital is four hospitals, so will your AI.

Consumption-based pricing

Pay for what the AI does, not per seat.

Consumption-based pricing means you pay for how much AI you use, instead of a fixed price per person. And yes, it's very much still a thing.

Salesforce sells Agentforce three ways: by consumption with Flex Credits, by conversation or per user. The newer editions bundle a set amount of credits in.

Why it matters: It changes how you budget: a busy month costs more, a quiet one less. Your Salesforce account executive can model it with you.

Flex Credits

The currency Salesforce AI runs on.

Flex Credits are how Salesforce meters AI usage. Each action an agent takes uses credits, and voice actions use a few more.

Salesforce's hosted AI connections are also meant for customers with Flex Credits, so that usage can be billed too.

Why it matters: Before a pilot, ask your Salesforce account executive how many credits your plan includes.

Agentforce editions

Core, Advanced and Max: Salesforce's all-in-one AI packages.

In September 2026, Salesforce launched Core, Advanced and Max editions for Agentforce Sales, Agentforce Service and Agentforce Industries. Each bundles AI, security, analytics, collaboration and support into a single purchase.

Each edition includes a set amount of Flex Credits, and the bigger the edition, the more credits you get.

Why it matters: Fewer line items to decode. Your Salesforce account executive can tell you which edition fits.

Beta vs. generally available

How finished is it?

Beta means a product works but is still changing. Open beta means any customer can try it. Generally available, or GA, means it's finished and fully supported.

Right now, Salesforce in Claude is in beta, Headless 360's connection for AI assistants is in open beta, and Salesforce's hosted MCP connections have been generally available since April 2026.

Why it matters: Beta is perfect for a pilot. It's not for a company-wide rollout.

Zero Data Retention

The AI answers, then forgets your data.

Zero Data Retention means your business data is used to answer the question at hand and isn't kept by the AI provider.

Salesforce builds it into AIforce and into its Einstein Trust Layer, which also masks sensitive data in prompts.

Why it matters: It's one of the first questions your compliance team will ask. Now you have the answer.

Access permissions

Who can see and change what in Salesforce.

Access permissions are the rules that decide who can see and change what in Salesforce: which objects, which fields and which records.

When an AI assistant works in Salesforce for you, it gets your exact permissions. If you can't do something in Salesforce, the AI can't either.

Why it matters: Your permissions are now your AI guardrails. If everyone can see everything today, fix that before you turn AI on.

Audit trail

The record of who changed what.

An audit trail records the changes made in Salesforce and who made them.

When an AI assistant updates a record for you, the change is logged under your name. Admins can filter Salesforce's logs to see exactly what AI assistants did.

Why it matters: Accountability doesn't disappear when AI shows up.

Sandbox

A safe copy of Salesforce to test in.

A sandbox is a copy of your Salesforce setup where you can build, test and train without touching live data or live users.

There are four types, from lightweight developer copies to full copies that include your data. Salesforce recommends trying AI changes in a sandbox before production.

Why it matters: Test your AI pilot where mistakes can't reach customers.

PHI and HIPAA

Health data the law protects.

PHI, or protected health information, is health information that identifies a person, or could reasonably be used to. It covers someone's health, their care and payment for that care.

HIPAA is the U.S. federal law behind the rules for protecting it. The core definitions live in 45 CFR 160.103.

Why it matters: Field sales rarely needs PHI. Keeping it out of AI prompts keeps your compliance team happy.

Prompt Builder

Reusable prompts, grounded in your own data.

Prompt Builder lets admins create reusable prompt templates that pull in your own Salesforce data, so the AI's answers fit your business.

Templates can use record data, data from Data 360 or outside data, and they can power Agentforce actions like writing summaries or filling in fields.

Why it matters: One great prompt, used by the whole team, beats a hundred improvised ones.

Grounding

Feeding the AI your facts, not the internet's.

Grounding means giving the AI the information you want it to use when it answers, like your Salesforce records, PDFs or emails.

AI models are trained on general knowledge, not your business. Grounding is how you get accurate, relevant answers about your accounts.

Why it matters: Ungrounded AI guesses. Grounded AI knows your accounts.

Tableau Next

Analytics built for the age of agents.

Tableau Next is Salesforce's newest analytics experience. It's built on Data 360 and works natively with Agentforce.

It runs on Tableau Semantics, which translates raw data into your business language.

Why it matters: Your dashboards and your AI agents finally work from the same data.

AgentExchange

Salesforce's marketplace for apps and agents, formerly AppExchange.

AgentExchange is the new name for AppExchange. It's Salesforce's marketplace for apps, agents and Slack solutions.

It's where you can grab Salesforce in Claude, and builders can plug in third-party agents and MCP connections right from Agentforce Builder.

Why it matters: Before you build it, check whether someone already has.

Plugin

An add-on that teaches Claude new tricks.

A plugin is an add-on package for an AI assistant. It bundles skills together with a connection to another system.

Salesforce in Claude is a plugin: a bundle of sales skills plus a secure connection to your Salesforce.

Why it matters: When someone says "just install the plugin," this is what they mean.

Level 3: Pack Leader

Building with AI: data architecture, developer tools, security architecture and the models behind Agentforce.

API

How software talks to software.

API stands for application programming interface. It's a set of rules for how one app asks another for data or actions.

The classic example: at a restaurant, you order from the waiter (the API), who takes it to the kitchen (the other system) and brings back your meal. Salesforce has been offering APIs for more than 27 years.

Why it matters: Every AI connection on this page runs on APIs underneath.

Data graph

A pre-built, connected view of your data.

A data graph links related objects in Data 360, like an account, its contacts and its orders, into one pre-combined record.

Because the combining is done ahead of time, apps and agents read it in a single fast lookup. A real-time version is available as a separate purchase.

Why it matters: Agents get connected customer context in one read instead of dozens.

Identity resolution

Turning five versions of a person into one.

Identity resolution matches records that belong to the same person or company and merges them into one unified profile.

In Data 360, you set the matching rules. "M. Gonzalez," her full name and her email address stop being three strangers.

Why it matters: Duplicate records confuse people. They confuse AI even more.

Semantic model

Your business, described so AI understands it.

A semantic model defines how your business data is organized and what it means for a specific use: the objects, how they relate and how metrics are calculated.

Tableau Semantics is Salesforce's semantic layer inside Data 360. It translates raw data into your business language.

Why it matters: When AI and your team define "active account" the same way, answers stop contradicting each other.

Zero copy

Use data where it lives. No copying.

Zero copy lets Data 360 use data from other systems, like a data warehouse, without duplicating or moving it.

It works both ways: Data 360 can read external data, and external systems can read Data 360 data.

Why it matters: Fewer copies means fewer versions of the truth to keep in sync.

Agentforce Builder

Where you build your own agents.

Agentforce Builder is Salesforce's low-code tool for building AI agents. You write the agent's instructions in everyday language on Canvas, a document-style workspace.

You set up subagents and a library of actions the agent can choose from, connected to your Flows, Apex, APIs and prompt templates. Claude is one of the models available in it.

Why it matters: It's how your admins go from "we should have an agent for that" to an actual agent.

Agentforce Vibes

Describe it. Salesforce codes it.

Agentforce Vibes is Salesforce's vibe coding tool for developers: you describe what you need, and it builds, tests and deploys the code.

It suggests code as you type, in its own editor and in VS Code, and Claude powers it by default.

Why it matters: Faster Salesforce builds for your development team.

Claude Code

Anthropic's AI coding tool.

Claude Code is Anthropic's agentic coding tool. It reads a codebase, edits files, runs commands and works with your development tools.

It runs in the terminal, in code editors, in a desktop app and on the web. Salesforce offers a Salesforce Development plug-in for Claude Code with more than 40 skills. Claude Code overview.

Why it matters: Your builders can use the same AI your sales team uses, pointed at code.

Headless Toolkit

Every piece of Salesforce, open to builders.

The Headless Toolkit is the open architecture that powers AIforce. It exposes every part of the Salesforce platform to builders through MCP, APIs, plug-ins, skills and developer tools.

Partners use it to bring Salesforce into their own products, from AI apps to cloud platforms.

Why it matters: If you're building AI experiences on Salesforce, this is your toolbox.

External Client App

The front-door key for outside apps.

An External Client App is how an admin registers an outside app, like Claude or ChatGPT, so it can securely sign in to your Salesforce.

Salesforce recommends one per AI tool. That way you control which users can use which tool, track usage separately and switch one off without touching the others.

Why it matters: One key per tool keeps your security team sane.

Customer 360

The business logic your agents lean on.

Customer 360 is Salesforce's family of apps across sales, service, marketing, commerce and more.

In Salesforce's AI architecture, it gives agents the business logic, processes, permissions and actions they need to get work done.

Why it matters: It's how your business already runs in Salesforce, made available to your agents.

Atlas Reasoning Engine

The brain of Agentforce.

The Atlas Reasoning Engine is what Salesforce calls the brain of Agentforce. It breaks a request into smaller tasks, checks its work at each step and proposes a plan.

It loops through reason, act and observe until the goal is done. Claude is one of the reasoning models it can use.

Why it matters: Knowing how agents plan helps you design agents that behave.

Reasoning model

An AI model that thinks before it answers.

A reasoning model works through a problem step by step before answering, which helps with multistep work.

Salesforce's own reasoning model, Koa, is built for CRM work. Claude also works as a reasoning model inside Agentforce.

Why it matters: Multistep business work needs a model that plans, not one that guesses.

Koa

Salesforce's own CRM reasoning model.

Koa is Salesforce's first CRM reasoning model for Agentforce, announced with NVIDIA at Dreamforce 2026.

Salesforce built it by training NVIDIA's Nemotron model on simulated business scenarios, so it applies consistent business rules and knows when to hand off to a human. It's in customer pilots for service, sales and commerce agents.

Why it matters: A model trained on how businesses actually run, including when to call in a human.

Long-horizon agents

Agents that work for weeks, not minutes.

Long-horizon agents pursue a goal across days and weeks, instead of finishing one task or one conversation.

Salesforce introduced a long-horizon runtime for Agentforce at Dreamforce 2026. An agent called Hunter is the first to run on it, and customers will be able to build their own.

Why it matters: Some deals take months. Soon your agents can keep up.

Salesforce Trust Boundary

Salesforce's secure perimeter for your data and AI.

The Salesforce Trust Boundary is the secure perimeter that keeps your data and AI workloads protected inside Salesforce.

Through Amazon Bedrock, Claude runs inside it, and Claude is the first AI model provider fully integrated within it. That lets regulated companies use AI while their data stays inside the boundary.

Why it matters: In healthcare, where your data goes matters as much as what the AI says.

AI Control Plane

Mission control for all your AI.

The AI Control Plane is Salesforce's single place to see, manage and control agents and AI as they spread across your company.

Salesforce announced it at Dreamforce 2026, alongside new capabilities for context, action, governance, security and models.

Why it matters: When you have dozens of agents, somebody needs one screen to watch them all.

Get in touch with our pack of HLS Salesforce experts today!

We'll answer your questions, look at where your team stands and show you what AI on Salesforce could do for you.

Wise Wolves helps providers, pharmaceutical companies, medical technology companies and more design Salesforce solutions that improve patient experience and drive revenue. Last updated September 29, 2026.
Ask me more about Salesforce AI